Applied Energy 313 (2022) 118807

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

The value of in-reservoir energy storage for flexible dispatch of geothermal

power
Wilson Ricks a ,∗, Jack Norbeck b , Jesse Jenkins a
a Princeton University, 86 Olden St., Princeton, 08540, NJ, USA
b
Fervo Energy, 708 Main St., Houston, 77002, TX, USA

ARTICLE INFO ABSTRACT

Keywords: Geothermal systems making use of advanced drilling and well stimulation techniques have the potential to
Geothermal provide tens to hundreds of gigawatts of clean electricity generation in the United States by 2050. With
Flexibility near-zero variable costs, geothermal plants have traditionally been envisioned as providing ‘‘baseload’’ power,
Storage
generating at their maximum rated output at all times. However, as variable renewable energy sources (VREs)
EGS
see greater deployment in energy markets, baseload power is becoming increasingly less competitive relative to
Wind
Solar
flexible, dispatchable generation and energy storage. Herein we conduct an analysis of the potential for future
geothermal plants to provide both of these services, taking advantage of the natural properties of confined,
engineered geothermal reservoirs to store energy in the form of accumulated, pressurized geofluid and provide
flexible load-following generation. We develop a linear optimization model based on multi-physics reservoir
simulations that captures the transient pressure and flow behaviors within a confined, engineered geothermal
reservoir. We then optimize the investment decisions and hourly operations of a power plant exploiting such
a reservoir against a set of historical and modeled future electricity price series. We find that operational
flexibility and in-reservoir energy storage can significantly enhance the value of geothermal plants in markets
with high VRE penetration, with energy value improvements of up to 60% relative to conventional baseload
plants operating under identical conditions. Across a range of realistic subsurface and operational conditions,
our modeling demonstrates that confined, engineered geothermal reservoirs can provide large and effectively
free energy storage capacity, with round-trip storage efficiencies comparable to those of leading grid-scale
energy storage technologies. Optimized operational strategies indicate that flexible geothermal plants can
provide both short- and long-duration energy storage, prioritizing output during periods of high electricity
prices. Sensitivity analysis assesses the variation in outcomes across a range of subsurface conditions and cost
scenarios.

1. Introduction Through technology innovation, drilling cost reductions, and im-

proved exploration techniques, it may be possible to significantly in-
Firm, low-carbon resources have been identified as critical for cost- crease the economically viable resource base for geothermal energy.
effective deep decarbonization of electricity systems [1,2]. Geother- For example, studies by the USGS have indicated that up to 30 GW of
mal power is one such resource, with added benefits of full renewa- undiscovered hydrothermal resources may exist in the US [6,7], which
could be identified using novel geophysical exploration techniques. In
bility and minimal land and resource use relative to other sources
addition, recent innovations in horizontal drilling, reservoir stimulation
of electricity [3]. Despite these advantages, geothermal deployment
techniques, and other Enhanced Geothermal System (EGS) technolo-
has historically been constrained to a very select set of sites where
gies can enable geothermal development in formations that would
naturally-occurring hydrothermal reservoirs can be exploited for elec-
otherwise be unsuitable or uneconomic [8]. Relatively shallow EGS
tricity generation. Due in large part to this lack of resource availability,
resources underlie much of the western United States, and successful
geothermal power currently supplies only 0.4% of annual electricity
development of this technology could unlock hundreds to thousands of
demand in the United States [4], with a total installed generating
gigawatts of geothermal resource potential nationwide, with up to 120
capacity of under 4 GW [5].
GW deployable by 2050 [7,9,10].

∗ Corresponding author.
E-mail address: [email protected] (W. Ricks).

https://doi.org/10.1016/j.apenergy.2022.118807
Received 11 November 2021; Received in revised form 30 January 2022; Accepted 20 February 2022
Available online 7 March 2022
0306-2619/© 2022 Elsevier Ltd. All rights reserved.
W. Ricks et al. Applied Energy 313 (2022) 118807

Still, the economic outlook for geothermal power is unclear in a 1.2. Geothermal energy storage
rapidly evolving electricity market. Geothermal plants typically operate
as ‘‘baseload’’ power, generating at their maximum rated capacity at all While conventional, stand-alone geothermal plants are therefore
times [11]. This has historically been a viable operating strategy, but limited in their ability to extract additional value from flexible oper-
shifting electricity market conditions are eroding the economic value of ations, a range of concepts have been proposed for improving plant
baseload power relative to more flexible alternatives. As electricity sys- value by using a geothermal reservoir as a medium for thermal or
tems move toward complete decarbonization, it is generally accepted geomechanical energy storage. Thermal energy storage can be enabled
that variable renewable energy (VRE) sources will supply an increas- by coupling a geothermal plant with another high-temperature ther-
ingly significant share of total generation [12–14]. However, increased mal energy source such as a solar thermal or nuclear power plant.
VRE penetration in electricity markets drives greater volatility in net Thermal energy from the coupled plant can be used during times
load and electricity prices [15], can lead to overgeneration [12], and of energy overabundance to heat the geothermal reservoir, allowing
has been associated with negative pricing episodes of increased length for greater energy production at later times [28,29]. This hybrid ap-
and severity [16]. Under such conditions, greater system flexibility proach is promising, but depends on the mutual cost-effectiveness and
is needed in order to maintain supply–demand balance in the grid, co-location potential of multiple generating technologies. It cannot
limiting the system value inflexible baseload resources [17–24]. From improve the performance of a stand-alone geothermal plant with no
an economic perspective, fast-ramping generators with low fixed costs, outside source of thermal energy. Another thermal energy storage
concept is CO2 plume geothermal, which charges and discharges by
which can save money by only generating when electricity prices are
moving a CO2 working fluid between two isolated subsurface reservoirs
high, and energy storage devices that shift generation to valuable
at different temperatures [30]. The reliance of this latter concept on
periods, can have a competitive advantage over baseload generators
unique geologic conditions likely limits its potential for large-scale
in a grid with significant VRE penetration [2].
adoption.
Geomechanical geothermal energy storage has been explored in
1.1. Flexibility in existing geothermal plants the context of sedimentary basin CO2 sequestration, particularly by
Buscheck et al. [31]. In the CO2 -Bulk Energy Storage (CO2 -BES) con-
To improve the future economic viability of geothermal power, cept, concentric rings of CO2 and brine injection and production wells
next-generation geothermal power plants must be designed with flex- create hydraulically confined regions where injected fluid can be effi-
ibility in mind. From a surface facilities perspective, most geothermal ciently recovered rather than being lost to leakoff. By injecting fluid
power plants are already capable of operating with a high degree of and reducing production during times of energy overabundance, such
flexibility, which is achieved by altering geothermal fluid production systems can build up pressure and accumulate working fluid in the
rate or production fluid enthalpy. Surface generators designed with reservoir. Increased pressure can then be used to drive artesian pro-
flexibility in mind, particularly binary-cycle plants, can ramp efficiently duction flow during periods of energy scarcity, effectively providing
between 10% and 100% output at rates of up to 30% nominal power grid-scale energy storage. Ogland-Hand et al. [32] found that such sys-
per minute [25]. Some binary plants today take advantage of this high tems could theoretically provide efficient energy storage for durations
degree of flexibility to provide ancillary services to the grid [11], but of up to a week.
the economic benefits are limited because reducing production does Although there may be value in CO2 -BES energy storage in the
not reduce operating costs for such plants. In general, the high fixed context of CO2 sequestration operations, sedimentary basins tend to be
costs and near-zero variable costs of geothermal power plants strongly poorly suited for stand-alone geothermal power production. The vast
disincentivize deviation from a baseload operating strategy. majority of high-heat (and therefore economically attractive) geother-
One strategy by which geothermal plants can derive benefits from mal resources in the United States exist elsewhere, particularly in
flexibility is to ramp down production during periods of negative elec- low-permeability settings where EGS would be the most viable ex-
traction method [10]. Though EGS plants operate under subsurface
tricity prices, thereby avoiding losses. Millstein et al. [26] optimized
conditions that differ greatly from those in sedimentary basins, they
geothermal plant operations against historical real-time electricity price
may actually be well-suited to providing a form of geomechanical
series and found that existing binary-cycle plants in the U.S. could
energy storage similar to that proposed by Buscheck et al. [31]. In
improve their energy value (measured as the average price of electricity
typical EGS designs, the low-permeability rock matrix surrounding an
per unit geothermal generation) by an average of 5.5% by operating in
engineered geothermal reservoir can provide natural hydraulic confine-
this manner. While not insignificant, these improvements are ultimately
ment without the need for complex well configurations. This behavior
the result of cutting losses during negative pricing episodes, and this
was demonstrated by the engineered reservoir at Fenton Hill, a Depart-
mode of operation does nothing to increase the value of the geothermal
ment of Energy EGS test site. At this site, fractures created during a
plant outside of those periods.
stimulation treatment program provided high-conductivity flow paths
There is some evidence that flash steam geothermal plants, which between an injection well and production well, and the relatively
rely on reservoir pressure rather than pumping to drive production low-permeability formation surrounding the fractures prevented fluid
flow, may be able to derive additional benefits from flexibility outside leakoff, causing confined reservoir behavior [33]. Periodic reductions
of avoiding negative prices. In an analysis of curtailment events at the in production flow rate with no changes to injection rate were shown
Geysers geothermal field, Goyal [27] found that multi-week reductions to result in increased reservoir pressure, which was effectively retained
in steam production were routinely followed by short-term increases in over many hours due to the confined nature of the reservoir. When
production flow, driven by increased reservoir pressure. This behavior the production wellhead backpressure was subsequently reduced, the
could in theory be exploited to ‘‘shift’’ plant generation from times accumulated reservoir pressure drove production flow at a higher rate
with lower electricity prices to times with higher prices. However, flash than the standard injection rate for a period of several hours.
steam plants are generally more limited in their ability to curtail pro- The results of the Fenton Hill EGS project demonstrated the po-
duction than binary plants [26], and it should be noted that the ‘‘puff’’ tential for in-reservoir energy storage (IRES) in such systems, wherein
events observed by Goyal [27] only recovered roughly 15% of curtailed accumulated geofluid and reservoir pressure are used to shift the output
generation due to pressure leakoff in the reservoir. Analysis by Millstein of a geothermal plant from one time to another. Importantly, the ability
et al. [26] indicates that the benefits of flexibility for a plant with these to store energy in this manner is an inherent property of an EGS
characteristics would be less than those for a fully-curtailable binary reservoir requiring no additional capital expenditure. Given the critical
plant that could not induce periods of flush production. importance of EGS to large-scale geothermal deployment [10], and

2
W. Ricks et al. Applied Energy 313 (2022) 118807

the value of both flexible generation and energy storage in electricity

systems [34–36], the ability for EGS to provide both these services
has the potential to significantly increase the value and deployment
potential of geothermal power as a whole.

1.3. Contribution of this work

Successful deployment of flexible EGS power with IRES will depend

on (1) developing a broader understanding of the subsurface char-
acteristics that promote effective energy storage, and (2) developing
innovative operational strategies that maximize the economic benefit
of operating flexibly. To address these issues, we first use a numerical Fig. 1. Three-dimensional model of the triplet well system, wherein two production
geothermal reservoir model to evaluate the technical feasibility of a wells are drilled at opposite orientation to a single horizontal injection well. The wells
flexible geothermal operational strategy in a synthetic system similar to are connected hydraulically by a series of evenly-spaced vertical fractures that provide
the primary flow paths. The fractures are able to deform mechanically in response to
the Fenton Hill EGS test site. We then employ plant-level optimization
changes in pressure and stress, providing fluid storage volume in the subsurface.
modeling to assess the impact of flexible operations and IRES on the
economic value of a single next-generation geothermal power plant. We
develop a linear optimization model capturing the transient pressure
and flow behavior within a geothermal reservoir system designed for well as uncertainties in technology costs and performance, we include
flexible operations, and use this model to co-optimize plant investments a set of sensitivity cases covering the likely ranges of these parameters.
and operations against both historical and modeled future electricity In the configuration described above, reservoir pressure alone is
price series. We further characterize the optimal operating modes of used to drive artesian geofluid flow in the production wells. Produced
a flexible geothermal power plant and assess the sensitivity of model fluid is run through a binary-cycle surface plant before being re-
outputs to variations in critical uncertain parameters. The goal of this injected into the reservoir. The use of a binary-cycle surface plant
work is to quantify the economic value that can be derived from flexible allows for highly-flexible generation with no additional operational and
maintenance costs [11].
operations and IRES for geothermal power plants, and to identify
primary areas of focus for ongoing research and development of this
2.2. Modeling approach
technology.

We employ a bottom-up approach to modeling the performance of

2. Methods
a flexible geothermal plant, using multi-physics numerical reservoir
simulations to develop a linear optimization model, which is used to
2.1. Representative plant design
evaluate the economics of flexible operations. Reservoir simulations
are done using a commercial geothermal reservoir simulation software
EGS is an emerging technology, and most proposed plans for its package that includes detailed calculations of wellbore friction, fracture
commercialization hinge on using early ‘‘near-field’’ projects to acceler- deformation and pressure response, fluid leakoff, and heat transfer [41,
ate technological learning [10]. These projects would target the hot but 42]. These simulations provide all of the information necessary for eval-
low-permeability formations surrounding known hydrothermal sites, uation of reservoir performance under different operating conditions
where minimal exploration and drilling is necessary for development. and strategies.
Sufficient cost reductions in these early phases could enable the eco- To allow for techno-economic optimization of plant investments
nomical development of ‘‘deep EGS’’ resources, those at depths of 3 and operations, we develop a secondary linear optimization model
km or more located in low-permeability basement rock [37]. It is this based on the numerical reservoir simulation results. We simulate the
deep resource that represents the vast majority of geothermal potential transient pressure response at the injection and production well bottom-
worldwide [10]. In this paper we focus on the transitional point in this holes to stepwise changes in injection and production rates. Piecewise
approach, considering an EGS plant mining a 218 ◦ C low-permeability linearization and subsequent linear superposition of this pressure re-
geothermal resource at a depth of 3 km. We use this representative case sponse function allow us to capture the transient reservoir pressure
to analyze the impact of flexibility and energy storage on the economics response to variable pumping rates within a linear programming (LP)
of early EGS projects. framework. The LP model also considers the dependence of production
In our analysis we consider a triplet well design similar to those rate on reservoir pressure and variable ‘‘parasitic’’ load due to injection
discussed in Gringarten et al. [38], Olson et al. [39], Li et al. [40], pumping. The reservoir simulation and optimization modeling method-
and others. An injection well is drilled to the target depth and deviated ology is described in further detail in Appendix A. Fig. 2 illustrates the
to produce a 1.5 km horizontal section. Hydraulic stimulation is used tight agreement between the simplified LP model and the multi-physics
to create regularly-spaced vertical fractures along this interval, and simulation results.
two production wells are drilled in parallel and at opposite orientation
to the injection well, intersecting these fractures and creating a long 2.3. Techno-economic optimization
horizontal reservoir. The vertical fractures serve as flow pathways that
create a strong hydraulic connection between the injection and produc- The LP model introduced in Section 2.2 captures the effects of
tion wellbores, as well as the reservoir contact area, to enable sustained changes to production and injection rates at hourly timesteps. These
heat recovery over the system life. The configuration assessed in this two flow rates are the model’s operational decision variables. Reduc-
paper is designed to produce 10.1 MW of electric power per well triplet. tions to production rate (or increases to injection) lead to increased
This well completion and reservoir engineering design is illustrated in reservoir pressure. The production well bottomhole pressure (BHP)
Fig. 1. The rock matrix surrounding the engineered reservoir is assumed determines the maximum achievable production flow rate at each
to have low but non-zero permeability, representative of the hot, low- model timestep. The maximum usable production flow rate is further
permeability formations that make up the majority of developable EGS constrained by the generating capacity of the surface plant, and the
resources worldwide. Recognizing the significant locational variation size of the plant’s grid interconnection. It is assumed for the purposes
in subsurface characteristics across potential EGS sites worldwide, as of this work that the binary organic Rankine cycle (ORC) surface plant

3
W. Ricks et al. Applied Energy 313 (2022) 118807

long-term power purchase agreements, renewable energy credits, ca-

pacity payments, and other factors. Rather, we use wholesale electricity
prices as a means to measure the market value of generation at each
hour, and evaluate the overall improvement in a geothermal plant’s
energy value that can be derived from flexible operations and energy
storage. This approach is designed to assess geothermal’s overall appeal
to potential power offtakers, which will be determined in large part by
the value of geothermal energy in wholesale electricity markets. The
use of a fixed electricity price series in this work assumes that the flex-
ible geothermal plant in question operates as a price-taker, i.e. that it
does not have enough market power to influence the price of electricity
through its operational decisions. This is a good approximation of the
conditions that would be experienced by the first flexible EGS plants to
enter a regional electricity market.
In this paper we optimize flexible geothermal investments and oper-
ations against eight historical and eight modeled future electricity price
series. The historical series consist of real-time locational marginal
Fig. 2. The LP optimization model and the numerical reservoir simulation executing
an identical sequence of operational decisions. Injection rate is held constant, while price data for the year 2019 taken from seven grid nodes in the
production flow is partially shut in for a period of 240 h, causing the injection and Western Interconnection and one in the Texas Interconnection, sourced
production well bottomhole pressures to rise. The production well is subsequently from the CAISO OASIS tool [44] and ERCOT’s online market prices
opened to allow up to a specified maximum flow rate, and both pressures and flow database [45], respectively. The nodes in the Western Interconnection
decline back to their steady-state values. Simulation results are shown as point data,
are located in states and regions with identified geothermal resources,
while model results are shown as continuous data.
and are intended to represent a diversity of possible electricity market
Table 1
conditions that might be encountered by an early flexible geothermal
Parameter values for the base case model. plant. The node from Texas is not located in an area with significant
Parameter Symbol Value geothermal potential but is instead selected to represent an electricity
Surface plant capital cost 𝐶𝑝𝑙𝑎𝑛𝑡 $2050/kW𝑛𝑒𝑡
market with high wind power penetration.
Interconnection capital cost 𝐶𝐼𝐶 $130/kW In addition to these historical price series, we also consider synthetic
Injection pump capital cost 𝐶𝑝𝑢𝑚𝑝 $400/kW national price series for the year 2030 created using the GenX electric-
Reservoir temperature 𝑇𝑟𝑒𝑠 218 C ity system optimization model [46]. We consider 2030 as a target year
Production temperature 𝑇𝑝𝑟𝑜𝑑 210 C
for full commercial deployment of flexible EGS, and create eight price
Injection temperature 𝑇𝑖𝑛𝑗 70 C
Maximum increase to reservoir pressure 𝛥𝑃𝑚𝑎𝑥 3.5 MPa series for this year in GenX to reflect different potential market sce-
Number of reservoir fractures 𝑁𝑓 100 narios. These include a set of four series assuming various technology
Fracture conductivity 𝐾𝑓 4.5 × 10−12 m3 deployment and load scenarios but no policy changes, and an additional
Matrix permeability 𝑘𝑚 9.9 × 10−19 m2 four considering the same scenarios under a national $60/ton carbon
Matrix porosity 𝜙𝑚 0.05
tax. Each set of four includes one baseline scenario, which represents
the results of a simple capacity expansion optimization with no changes
to base system inputs. The second scenario uses an alternate load
can handle geofluid flow rates up to 10% above its design point with profile peaking in winter, which is more representative of northeastern
minimal loss of efficiency and without any additional capital expendi- states under increased electrification. The third and fourth scenarios
ture [43]. Beyond this point, the surface facilities must be oversized alter wind and solar input parameters to prioritize deployment of
(with respect to a plant designed for constant flow from the same one of these resources over the other. Though these modeled price
reservoir) to take advantage of increased production flow. Surface plant series are spatially unresolved, and do not capture the same level of
and grid interconnection oversizing are included as investment deci- volatility seen in real electricity markets, they allow us to evaluate
sions, and are penalized in the objective function to reflect increased the performance of flexible geothermal in systems with significantly
construction, maintenance and insurance costs. Injection pumps may higher VRE penetration than exists in most markets today. For example,
also be oversized at a specified cost in order to accommodate greater the ‘‘high solar’’ and ‘‘high wind’’ carbon tax scenarios each see VREs
injection pressures and flow rates. Financial assumptions for surface making up just under 50% of installed generating capacity nationwide.
facilities oversizing are outlined in Appendix B.6. Costs used in the
baseline model are presented in Table 1, along with baseline values for 3. Results and discussion
other important model parameters. Due to the very high flexibility of
ORC units, the model does not impose any hourly ramping or minimum 3.1. Base case results
generation constraints.
The optimization model takes an hourly electricity price series as Optimization results under baseline reservoir performance and plant
input, and co-optimizes investment and hourly operational decisions to costing assumptions for historical price series are given in Table 2,
maximize plant revenue over a single weather year. Only investment and results for modeled 2030 price series are given in Table 3. All
costs corresponding to surface facilities oversizing are modeled, and all energy value and component-sizing results are presented with respect
costs associated with the reservoir, well field, and conventional surface to the corresponding values for a baseload geothermal plant. The
facilities are considered sunk. Though the wellfield and subsurface average hourly energy value per MW of net generating capacity for
reservoir at a given plant could theoretically be engineered to maximize such a plant is equal to the average hourly price of electricity for each
the value of the facility when operating flexibly, such an optimization is series. Relative improvement over this ‘‘baseload’’ average energy value
beyond the scope of the current work. Our model thus assesses only the from a curtailment-only operating strategy similar to that discussed
relative costs and benefits associated with flexibility and energy storage in Millstein et al. [26] is given for both sets of price series. This number
at a plant otherwise designed for baseload operations. represents the degree to which value could be improved if the plant
It should be noted that this work is not intended to predict true curtailed generation during negative pricing episodes but did not store
revenues for geothermal power plants, which are often determined by the lost energy. Value improvement from full flexibility represents the

4
W. Ricks et al. Applied Energy 313 (2022) 118807

average energy value of the same plant under flexible operations with oversizing relative to the benefits that can be extracted. Interconnection
IRES, after subtracting the annuitized cost of any oversizing of surface oversizing (e.g. construction of greater transmission interconnection
facilities. For all flexible cases, energy value is given with respect to the capacity than baseload surface plant net generating capacity) is fairly
total annual generation of a baseload plant. This allows for meaningful significant and consistent across all price series, due to very low capital
comparison of results in cases where annual flexible generation is less costs relative to other components. Even if the gross surface plant
than baseload (e.g. due to increased pumping loads). The optimal capacity itself is not expanded, oversizing the interconnection allows
degree of plant, interconnection, and injection pump oversizing for a the geothermal facility to deliver more net power at certain times
fully flexible plant is given with respect to the sizes of these components by reducing parasitic load from pumping. Pump oversizing, which
in an inflexible baseload plant. is required for operation at higher injection rates and BHPs, shows
Results for historical price series indicate that flexible operation and significant variation between price series. It is most extreme for series
energy storage produce greater energy value than baseload operation with many hours of zero or negative pricing, which allow the plant
in all cases, with a minimum relative value improvement of 6% and to temporarily boost its pumping rate without incurring an economic
a maximum of 44%. The greatest level of improvement occurs at penalty. Given the variance in optimal component sizing across differ-
the southern California node, which represents an electricity market ent electricity market conditions, geothermal developers should make
with the highest level of solar penetration in the United States [4]. efforts to forecast the evolution of a local electricity market over a
Significant improvement is also observed at the Arizona and Nevada flexible plant’s operational lifetime before beginning surface facilities
nodes, as well as at the wind-heavy Texas node. In these markets, high
construction. Doing so will minimize the risk of suboptimal plant and
VRE penetration drives greater electricity price variability, increasing
interconnection sizing. It is possible that in the case of an undersized
flexible geothermal’s relative advantage over baseload by providing
plant, extra ‘‘peaking’’ turbines, interconnection reinforcements, or sup-
more opportunities for arbitrage via energy storage. Conversely, bene-
plementary pumps could be added later in its lifetime to enable greater
fits from flexibility and energy storage are greatly reduced in locations
flexibility.
with more stable electricity prices. The northern California node, for
example, is located near The Geysers geothermal installation, where a
steady supply of baseload power leads to relatively stable electricity 3.2. Optimal operational strategies
prices and reduces benefits from flexibility. Though simple curtailment
can occasionally provide appreciable value improvements, improve- Analysis of the operational decisions made by the optimization
ments from full flexibility are much larger in all cases, even taking into model in response to specific price structures provides insight into the
account the extra costs associated with surface facilities oversizing. sources of value for fully flexible geothermal power. Fig. 3 presents
Results for modeled future price series indicate similar trends to pressure and flow behaviors during a selected period of 200 h, for the
those observed for historical price series. Flexibility and energy storage base case model run against the southern California historical price
offer greater relative benefits in a market with a large carbon tax, which series, as well as hourly electricity prices during the same period.
results in greater VRE deployment and higher marginal prices during As expected, the model tends to reduce production rate and increase
periods of low VRE output. In cases without a carbon tax, energy value injection rate during periods of low electricity prices, which causes
improvements are relatively consistent across the baseline, high solar reservoir pressure to rise. When prices are high, the model will reduce
and high wind cases. The high-VRE cases see a much larger relative injection while boosting production to the maximum level allowed by
improvement over the baseline when a carbon tax is also in place. With the capacity of the surface facilities. This process occurs on a diurnal
and without a carbon tax, value improvements are greater for the high basis, following the daily peaks and troughs in electricity prices, but
wind case than for the high solar case. The ‘‘winter peaking’’ case sees also occurs over much longer timescales. For the 200 h period refer-
very reduced benefits under both policy scenarios, but this case is also enced in Fig. 3, electricity prices are generally higher in the second set
not very representative of areas in the US with significant geothermal of 100 h than in the first. The model responds to this price structure by
potential [10]. Because the modeled price series used in this paper injecting more and producing less over the first 100 h period, causing
reflect a perfectly planned and operated system with minimal volatility
the reservoir pressure to become elevated over this entire period.
and low geospatial resolution, the relative benefits of flexibility in
During the second period, this accumulated pressure allows the plant
similar real-life systems are likely to be greater than those reported
to maintain its maximum production while reducing injection, thereby
here.
limiting parasitic load and maximizing net generation. This behavior
Analysis of the 16 historical and modeled future price series (see
demonstrates that with the right subsurface conditions, fully flexible
Appendix D.1) shows that the level of energy value improvement
geothermal is capable of providing long-duration energy storage over
from flexible operations and energy storage cannot be easily deduced
periods of hundreds of hours. This is possible because a large amount
analytically. For some price series, very high peak prices appear to be
of geofluid can be stored within the reservoir before the maximum
the primary source of value. This is the case for the Texas historical
allowed pressure is reached. The energy storage capacity of a flexible
price series, which features a very high price cap of $9000/MWh. For
geothermal reservoir is discussed further in Section 3.4.
others, frequent negative pricing rewards energy storage. In general,
the number of hours of the year for which prices are at or below Figs. 4 and 5 show behaviors in response to price series with
$0/MWh appears to be a good, but imperfect indicator of the level different characteristics. Fig. 4 illustrates the optimal operating strategy
of benefit that can be extracted from full flexibility. Zero or negative in response to the historical Texas price series, which features mostly
prices allow flexible geothermal plants to boost injection rates without low, stable prices punctuated by extreme price spikes. In this case,
regard for increased parasitic loads, increasing generating capacity the model builds extra generating capacity to take advantage of these
during subsequent periods of positive pricing. spikes, and organizes all operations around maximizing net generation
Across all price series, optimal surface plant oversizing (e.g. con- during these periods. Fig. 5 illustrates behavior for a section of the 2030
struction of additional surface plant gross generating capacity beyond carbon-tax high-wind price series featuring a long period of near-zero
that of a baseload plant) ranges from 0%–61% of baseload surface plant prices followed by intermittent but fairly stable periods of high prices.
gross capacity. The greatest levels of oversizing are observed for the The plant curtails generation completely during the low-price period,
historical Texas series and the future carbon tax scenarios. For the allowing it to maintain maximum generation for more than 20 h during
Texas series, oversizing allows for increased generation during short the high-price periods. The facility takes advantage of short intervals
periods of extremely high prices. In the future carbon tax scenarios, it of near-zero prices to quickly ‘‘recharge’’ the reservoir, allowing it to
is likely that higher average electricity prices decrease the cost of plant maintain high net generation.

5
W. Ricks et al. Applied Energy 313 (2022) 118807

Table 2
Energy value and optimal investment results for historical price series in the base case model.
Node Baseload Value improvement: Value improvement: Plant oversizing Interconnection Pump oversizing
location energy value Curtailment only Full flexibility (% of baseload oversizing (% of (% of baseload
($/MWh Avg.) (% of baseload) (% of baseload) plant capacity) baseload IC capacity) pump capacity)
AZ 31.3 5 20 23 48 129
N-CA 34.8 0 9 5 26 62
S-CA 25.9 19 44 16 40 636
ID 29.6 0 8 0 21 42
NV 30.5 6 18 8 31 95
OR 27.3 0 6 0 20 37
TX 34.6 0 22 61 93 85
UT 27.8 2 10 0 21 45

Table 3
Energy value and optimal investment results for modeled future price series in the base case model.
Scenario Baseload Value improvement: Value improvement: Plant oversizing Interconnection Pump oversizing
energy value Curtailment only Full flexibility (% of baseload oversizing (% of (% of baseload
($/MWh Avg.) (% of baseload) (% of baseload) plant capacity) baseload IC capacity) pump capacity)
BAU baseline 29.6 2 14 6 25 61
BAU winter peak 29.4 0 3 0 18 18
BAU high solar 28.9 0 11 4 24 46
BAU high wind 32.2 1 15 13 35 77
CO2 tax baseline 49.3 0 15 20 45 148
CO2 tax winter peak 51.1 0 11 8 29 82
CO2 tax high solar 46.6 0 24 33 60 189
CO2 tax high wind 46.5 0 26 29 55 203

Fig. 3. Optimal flexible operational strategy during a selected period of 200 h for the Fig. 4. Same as Fig. 3, for the Texas historical price series.
southern California historical price series in the base case model. The upper plot gives
injection and production well BHPs and injection and production flow rates at hourly
timesteps during this period. The middle plot gives net generation under both baseload
and flexible operating strategies, and lower plot gives hourly electricity prices. Model and rock matrix permeability 𝑘𝑚 . We include high and low bounding
timestep 1 corresponds to 00:00 on January 1. cases for all parameters except 𝑂𝑀𝑠 , as well as mid-high and mid-low
cases for both 𝐾𝑓 and 𝑘𝑚 . Assuming that flexible operations would
only increase maintenance costs, we include only a high bounding case
3.3. Sensitivity analysis for the 𝑂𝑀𝑠 parameter. The parameter variations associated with each
sensitivity case are outlined in Table 4. Fig. 6 shows the variation in
In addition to the base case, we also investigate 15 sensitivity total annual energy output for a baseload geothermal plant across all
cases designed to cover the range of parametric variation that might sensitivity cases. Flexible operations and investments are optimized for
occur across different geologic, regulatory, and technology advance-
each sensitivity case against all 16 historical and modeled future elec-
ment scenarios. Individual sensitivity cases adjust a single parameter
tricity price series. Fig. 7 presents the relative improvement in energy
value relative to the base case, allowing us to evaluate the sensitivity
value from full flexibility for each of the 240 total cases analyzed.
of the results outlined in Section 3.1 to variations in each of these
parameters. Sensitivity parameters include the cost of surface plant Results indicate that a reduction in surface plant cost generally
oversizing 𝐶𝑝𝑙𝑎𝑛𝑡 , the effect of flexible operations on surface facilities improves the value of flexible operations, especially for price series
maintenance costs 𝑂𝑀𝑠 , the maximum allowable pressure increase in with a high level of optimal plant oversizing. Reduced plant cost
the reservoir 𝛥𝑃𝑚𝑎𝑥 , fracture conductivity 𝐾𝑓 , number of fractures 𝑁𝑓 , allows for greater surface oversizing, which in turn enables greater

6
W. Ricks et al. Applied Energy 313 (2022) 118807

Table 4
Sensitivity case names and parametric variations.
Scenario Description
−
𝐶𝑝𝑙𝑎𝑛𝑡 Surface plant capital cost reduced by 33%.
+
𝐶𝑝𝑙𝑎𝑛𝑡 Surface plant capital cost increased by 50%.
𝑂𝑀𝑠+ Surface maintenance costs increased by 10% when operating flexibly.
−
𝛥𝑃𝑚𝑎𝑥 Maximum allowable reservoir pressure reduced by 1.5 MPa.
+
𝛥𝑃𝑚𝑎𝑥 Maximum allowable reservoir pressure increased by 1.5 MPa.
𝑘−−
𝑚 Matrix permeability reduced to 1/100 of base case value.
𝑘−𝑚 Matrix permeability reduced to 1/10 of base case value.
𝑘+𝑚 Matrix permeability increased to 10x base case value.
𝑘++
𝑚
Matrix permeability increased to 100x base case value.
𝐾𝑓−− Fracture conductivity reduced to 1.1 × 10−12 m3 .
𝐾𝑓− Fracture conductivity reduced to 2.3 × 10−12 m3 .
𝐾𝑓+ Fracture conductivity increased to 1.2 × 10−11 m3 .
𝐾𝑓++ Fracture conductivity increased to 3.2 × 10−11 m3 .
𝑁𝑓− Number of fractures reduced to 50.
𝑁𝑓+ Number of fractures increased to 150.

determining parasitic power requirements for injection pumping. While

reduced fracture conductivity reduces the energy output of a baseload
Fig. 5. Same as Fig. 3, for the 2030 CO2 tax high wind price series. plant by increasing parasitic load from injection, it increases the rela-
tive benefits of flexibility. The opposite is true for cases with increased
fracture conductivity, which see greater baseload energy output but
generation during peak-price periods. Increasing surface plant cost has reduced relative benefits from flexibility. This occurs because shifting
the opposite effect, though often to a lesser extent. injection pumping to times with lower electricity prices has a greater
It is possible that frequent cycling of generators and injection pumps relative impact in scenarios where injection load consumes a greater
under flexible operation could increase the maintenance costs of these proportion of net plant generation. Flexible operations can thus reduce
facilities. These additional costs would affect the entire facility, not just the negative impact of lower-than-desired fracture conductivity on the
the oversized portion. Lew et al. [47] found that increased cycling in economics of a geothermal plant.
response to greater VRE penetration raised the O&M costs per MWh Increasing or reducing the number of fractures present in the en-
of fossil-fueled thermal power plants by 2%–5%. Here, we find that a gineered reservoir has several effects. A smaller number of fractures
10% increase in maintenance costs for both the ORC power plant and will reduce the overall transmissivity of the reservoir, leading to higher
injection pump under operational flexibility reduces the relative value parasitic load from injection pumping. It will also reduce the surface
of IRES only marginally. Given the high flexibility of ORC units relative area over which the fracture system interacts with the rock matrix,
to conventional thermal generators, it is likely that true surface O&M somewhat limiting the exchange of fluid and therefor increasing the
penalties from IRES will be well below this threshold. pressure response to pumping. We find that reducing the number of
𝛥𝑃𝑚𝑎𝑥 limits the allowable pressure increase within the reservoir, fractures reduces baseload energy output somewhat, and that increas-
and effectively represents the maximum ‘‘charge capacity’’ of the sys- ing the number of fractures does the opposite. The effect on relative
tem. Anticipating that excessive reservoir pressurization could produce benefits of flexibility over baseload operations is unpredictable, but is
deleterious effects, including induced seismicity [48], we limit 𝛥𝑃𝑚𝑎𝑥 generally fairly small. Overall, results are fairly insensitive to changes
to 3.5 MPa in the base case model. We find that increasing this in the number of fractures. A larger number of fractures will therefore
value by 1.5 MPa does not produce major benefits in any scenario. be desirable, as this will increase the heat extraction efficiency of the
Reducing 𝛥𝑃𝑚𝑎𝑥 by the same amount limits the benefits of flexibility system with minimal negative effects.
in only a small number of cases. Overall, it seems that the 3.5 MPa
pressurization limit used in this paper unlocks nearly all of the benefits
of flexibility, and that a 2 MPa limit still captures most of these benefits. 3.4. Energy storage efficiency and capacity
The seismic risk associated with this level of pressurization should be
further characterized through field testing. A fully flexible geothermal power plant can be thought of as an
We find that reduced matrix permeability has the most significant energy storage device stacked on top of an inflexible baseload genera-
effect on flexible operations, cutting relative benefits by up to a third tor. If the plant is at zero ‘‘state of charge’’, it will produce a constant
in the lowest-permeability case. Reduced matrix permeability limits the amount of power equal to its baseload capacity. Negative deviation
migration of geofluid between the fractures and the matrix, leading from this steady-state output, either through increased injection load
to more significant pressure changes in response to pumping. This in- or curtailed production, can be thought of as equivalent to ‘‘charging’’
creased pressure response causes the reservoir to reach its pressure cap the geothermal energy storage. From the perspective of the electricity
more quickly, effectively reducing the energy storage capacity of the grid, a reduction in generation from the geothermal plant appears
system. Increased matrix permeability leads to greater baseload energy identical to an increased load from energy storage charging at the
production, due to greater natural flow rates from the reservoir, but same location. Likewise, positive deviation from the plant’s baseload
does not appreciably improve the relative benefits of flexible operation generation represents discharging of the geothermal energy storage.
over baseload. These results suggest that it will be necessary to fully
Based on these interpretations, the average round-trip efficiency of
characterize the subsurface properties of a potential geothermal site in
flexible geothermal energy storage over a year of operations can be
order to accurately assess the local potential for flexible operation and
calculated via the following equation:
IRES. ∑
Changes to the fracture conductivity 𝐾𝑓 have a significant effect on | 𝑡∈[1∶8760] 𝛥𝐺+ |
| 𝑡 |
𝜂𝑎𝑣𝑔 = | ∑ |, (1)
baseload energy output, as fracture conductivity is the primary factor | 𝑡∈[1∶8760] 𝛥𝐺− |
| 𝑡 |

7
W. Ricks et al. Applied Energy 313 (2022) 118807

Fig. 6. Annual energy production for a baseload geothermal plant in each sensitivity case, normalized with respect to the energy production of a base-case plant.

Fig. 7. Relative energy value improvement over baseload for a fully flexible geothermal plant across 16 sensitivity cases and 16 electricity price series.

where 𝛥𝐺𝑡± is the deviation of net plant generation at hour 𝑡 from In addition to price series analysis, we conduct a set of controlled
the steady-state baseload output, either positive or negative. Round- tests using the same optimization model to quantify the total energy
trip efficiency is calculated for each of the 240 cases analyzed in storage capacity of IRES. To optimize energy capacity, we provide the
this paper. These results are given in Fig. D.15. Optimal efficiency is model with an artificial input price series featuring near-zero positive
variable across price series and sensitivity cases, but is typically in prices for a period of 240 h, followed by an extended period of
excess of 80%. This value is comparable to the round-trip efficiencies extremely high prices. Plant investment costs are not taken into con-
of both pumped-hydro energy storage and lithium-ion batteries, two sideration, and both maximum injection and production flow rates are
leading grid-scale energy storage technologies [49]. As a general rule, capped at 4/3 the corresponding steady-state flow rates for a baseload
flexible geothermal plants with energy storage have lower annual plant. The model’s objective is therefore to store as much energy as
generation than their baseload counterparts. Despite this, the fact that possible in the 240 h ‘‘charging period’’ so that it can maximize the total
the flexible plants prioritize generation at times of high electricity net generation during the subsequent ‘‘discharging period’’. The results
prices significantly improves the overall value of their energy relative of this experiment can be used to set an upper bound on the energy
to baseload plants. Notably, analysis of model results indicates that only storage capacity of a flexible geothermal plant. Fig. 8 shows behavior
a small portion of the reduced net generation from flexible operations of the base case model operating in this controlled scenario. Energy
is attributable to reduced gross generation. Instead, the vast majority storage capacity is calculated as the deviation from baseload generation
of ‘‘round trip losses’’ occur due to increases in parasitic load from integrated over the entire discharge period, and is normalized with
injection pumping during flexible operation. This extra pumping power respect to the net generating capacity of the baseload surface plant (ig-
is a consequence of maintaining the reservoir at an elevated pressure, noring injection pumping load), which scales in size with the subsurface
but also of increasing injection rate above its steady-state level dur- reservoir. Fig. 9 shows the maximum energy storage capacity per MW
ing certain periods. Injection pumping power increases approximately of net baseload plant capacity for each sensitivity case.
quadratically with injection rate, meaning that there is a tradeoff These results indicate that under baseline assumptions the subsur-
between faster ‘‘charging’’ and overall storage efficiency. The observed face reservoir of a fully flexible plant can store and discharge up to
variation in optimal efficiency can be attributed to the model ignoring 159 MWh of electricity per MW of baseload surface capacity. Assuming
or even encouraging losses during periods of very low or negative that the ‘‘discharge capacity’’ of IRES is roughly equivalent to the extra
electricity prices, while instead prioritizing fast charging using cheap surface capacity built, this implies a storage duration in excess of 450 h.
energy. This figure falls well within the range described by Sepulveda et al. [49]

8
W. Ricks et al. Applied Energy 313 (2022) 118807

Nevertheless, the results presented in Fig. 9 represent good ballpark

estimates of the maximum energy storage capacity provided by an EGS
reservoir under different subsurface conditions.

3.5. Discussion of in-reservoir energy storage

As an energy storage technology, fully flexible geothermal power

is fairly unique. It has significant charging power capacity, equiva-
lent to the capacity of the base geothermal plant plus any additional
load from increased injection pumping. This parasitic load can be
increased significantly during times of very low or negative electric-
ity prices, allowing for rapid pressurization of the reservoir at the
cost of round-trip storage efficiency. Charging rate is limited only
by the capabilities of the injection pumps and pressure ratings of
surface piping and wellheads, all of which are fairly low-cost relative to
other plant components [50,51]. The storage discharge power capacity
Fig. 8. Operational decisions and net electrical output under an capacity-optimized is more constrained, being limited by the size of the surface plant
charge–discharge cycle, for the base case model. and grid interconnection as well as by the current level of reservoir
pressurization. Binary-cycle geothermal plants have a relatively high
specific cost (on the order of $2000/kW) [50,51], which limits the
economically viable discharge capacity of a flexible geothermal plant.
As discussed in Section 3.1, optimal surface capacities for flexible plants
are generally not significantly larger than those for baseload plants.
This economic constraint means that flexible geothermal is ill suited
to delivering large injections of power over short timescales. However,
effective ‘‘discharge’’ capacity can also be increased (at generally lower
cost, depending on location) by oversizing transmission interconnection
capacity and cutting injection pumps to boost net generation during
high price periods, even without increasing gross generation capacity.
As illustrated above, economically optimized round-trip efficiencies for
flexible EGS plants are in the same range as those for today’s leading
grid-scale storage technologies. Table 5 compares the fundamental
parameters of geothermal IRES to those of a number of competing
grid-scale storage technologies.
What makes EGS energy storage unique is its large energy capacity,
which comes essentially free of charge as an inherent property of
an engineered geothermal reservoir. Low-cost energy capacity is the
Fig. 9. Maximum energy storage capacity per MW of baseload plant capacity for a
most important characteristic of effective long-duration energy storage
flexible EGS plant under various sensitivity cases.
technologies [49], and enables the hundred-hour-plus charging and
discharging behaviors observed in Fig. 3. The energy storage provided
by these plants is therefore extremely competitive, providing an option
as ideal for long-duration energy storage technologies in decarbonized that is in many ways superior to other grid-scale storage technologies.
electricity systems. Its primary downside is that it can also only be accessed via construc-
Sensitivity analysis indicates that storage capacity has a positive tion of a full-scale geothermal power plant. Its economic viability is
relationship with the maximum allowable reservoir pressure. Matrix therefore tied to that of next-generation geothermal power. However,
permeability has a strong effect on storage capacity, with very low if geothermal systems exploiting advanced drilling and stimulation
matrix permeability showing significantly reduced storage durations. techniques can be successfully developed in near-field locations and
This finding is in line with results presented in Section 3.3. Fracture harness cost reductions from both oil and gas sector spillovers, progres-
conductivity has a counter-intuitive effect on storage capacity, as lower sive learning-by-doing in well completion and stimulation [37], and
values for this parameter lead to larger capacities. In these low fracture economies of scale in binary ORC turbine production [52], geothermal
conductivity cases, changes in flow rate at one well have a significantly costs may become steadily lower over time. In this case, the ability to
muted effect on BHP at the other wells. This effect leads to a delinking take advantage of such effective energy storage can significantly im-
of injection and production BHPs, allowing the system to raise pro- prove the economics of future geothermal projects, as is demonstrated
duction BHP by a greater amount without violating the injection well in this paper, with increased revenue from flexible operation potentially
BHP cap. Finally, results indicate that storage capacity scales with the expanding the range of economically-viable early-stage projects.
number of fractures. While adding fractures increases total reservoir
transmissivity and thereby reduces the delinking effect noted above, 4. Conclusions and future work
the increased storage potential due to a larger contact area with the
rock matrix outweighs this effect. While geothermal power has significant potential to contribute to
These results are idealized in the sense that real electricity prices do a zero-carbon electricity grid, the technology’s historically inflexible
not necessarily incentivize capacity-optimized charge–discharge cycles generation profile limits its usefulness and economic appeal in systems
hundreds of hours long. It should also be noted that, when analyzing with high VRE penetration. In this paper, we demonstrate that by
storage capacity, it is difficult to quantify the state of charge for this exploiting the hydraulic and geomechanical properties of a confined
system in a consistent manner. Reservoir pressure is limited by the geothermal reservoir, an EGS power plant can provide flexible genera-
injection BHP cap, but a combination of production BHP and flow rate tion and energy storage services that significantly enhance its economic
histories determines the total discharge potential at any given time. value. We develop a novel modeling framework which accurately

9
W. Ricks et al. Applied Energy 313 (2022) 118807

Table 5 traditional EGS. The potential to engineer and optimize operations of

Comparison of EGS IRES cost and efficiency parameters with future projections for
flexible geothermal systems across a wider range of possible subsurface
leading grid-scale storage technologies. Competing technology parameters are taken
from Sepulveda et al. [49], and are subject to geological and geographic constraints. conditions should be explored in future work.
Lithium-ion battery costs are projections for the year 2040 taken from the NREL ATB A subject not investigated in this paper is the potential for flexible
2020 [53]. operating strategies to produce stresses that negatively affect well and
Technology Power Energy Round-Trip reservoir integrity over the lifetime of a geothermal plant. It is possible,
capacity cost capacity cost Efficiency
($/kW) ($/kWh) (%)
as discussed above, that pressurization of the reservoir and frequent
changes in flow rates could increase the risk of induced seismicity.
EGS IRES 1100–2600 <1 61–91
Pumped Hydro 600–2000 20+ 70–85 Well damage due to thermal cycling has also been cited as a potential
Compressed air 600–1150 1–30+ 42–67 downside of steam flow curtailment at The Geysers [54], though this
Power-H2 -Power (Fuel Cell) 440–3400 1–15+ 20–46 problem may be less severe for liquid-dominated systems like those
Firebrick resistance-heated 930–1150 5–10 49–54
investigated here. Finally, modulation of injection and production rates
energy storage
Vanadium redox flow battery 270–600 40–200 65–80 may produce unforeseen issues with mineral precipitation and scaling,
Lithium-Ion battery 82–180 94–206 85 increasing the costs associated with flexible operations and IRES.
Despite the high level of detail captured by the reservoir simulation
software used in this study, only at-scale field testing can fully confirm
represents pressure and flow behaviors within an EGS reservoir and the viability of IRES as an EGS operating strategy. Though the basic
optimizes operations and investments to maximize the revenue of a principles of IRES were demonstrated successfully at Fenton hill, the
geothermal power plant. The results presented here indicate that under risks of operating a new EGS field in a fully flexible mode without
the right conditions, an EGS reservoir can efficiently store hundreds prior field testing are extremely high. This work represents an initial
of megawatt-hours of energy per MW of surface capacity without screening effort designed to estimate the incremental value of IRES
significantly increasing reservoir pressure. This large energy storage under reasonable assumptions, and the significant potential benefits
capacity is an inherent property of a confined reservoir, and therefore demonstrated here suggest that further research and field demon-
comes at no extra cost relative to a baseload geothermal plant. By stration is warranted. Extensive field validation of flexible operating
taking advantage of this capability to provide flexible generation and strategies to assess the long-term impacts of flexible operations on
energy storage, EGS plants can improve their energy value by a large reservoir performance and surface plant O&M costs, as well as the
amount relative to baseload geothermal plants. If this additional value associated seismic risk, should be explored as a next step. Results from
can be translated into plant revenues via well-constructed energy and field testing can be used to further refine the modeling tools developed
capacity markets or power-purchase agreements, flexible operations here, leading to more accurate assessments of value. It is likely that
and energy storage can effectively raise the price point at which EGS multiple demonstrations in different lithologies will be required before
power can compete as an energy source. This extra financial breathing this technology can be reliably deployed at scale.
room could be crucial to the economic success of early EGS projects, The work presented in this paper is designed to assess the value of
which represent an important step on the experience curve that could flexible operations and energy storage for early EGS projects, and does
unlock development of much more significant ‘‘deep EGS’’ resources. not focus on electricity systems under deep decarbonization. Moving
We find that flexibility and energy storage are more valuable in elec- toward entirely carbon-free electricity would almost certainly drive
tricity markets with high VRE penetration, where large fluctuations in
greater deployment of variable wind and solar resources [12–14],
electricity prices provide opportunities for arbitrage. Sensitivity analy-
which would lead to even greater benefits from geothermal storage
sis shows that while baseload plant energy production varies somewhat,
and flexibility. However, it is likely that any successfully deployed
the degree of value added by full flexibility is fairly insensitive to
early EGS projects will have entered the market well before this level
variation in most subsurface parameters and surface plant costs. The
of decarbonization is achieved. Because the fixed price series analysis
greatest impact is observed in the very low matrix permeability case,
used in this work assumes very low levels of flexible EGS market
which significantly reduces both the total energy storage capacity of the
reservoir and the revenue improvements that can be achieved through penetration, it is of limited use in quantifying the long-term value and
flexible operation. These results indicate that flexible operations and deployment potential of flexible geothermal power under such scenar-
energy storage can improve the value of EGS projects across a wide ios. An accurate analysis of this long-term potential must capture the
range of subsurface conditions, but that these conditions will need ability of flexible geothermal plants deployed at scale to influence the
to be well-characterized in order to accurately assess a site’s flexible price of electricity through their operational decisions, as well as the
potential. diminishing marginal returns associated with increasing deployment
Our results are based on numerical reservoir simulations of a triplet of a specific energy technology, which apply especially strongly to
well configuration with a reservoir management strategy involving a energy storage technologies [36]. To this end, the optimization model
ratio of two production wells per injection well, and a uniform series developed for this paper is designed in such a way that it can be
of vertical fractures. It is likely that different reservoir designs will incorporated, with some reductions in complexity, into a larger long-
see different energy storage capacities and efficiencies. For example, term electricity system capacity expansion model. The use of capacity
reservoirs with larger stimulated volume per unit of flow rate will likely expansion modeling to analyze the long-term value and deployment
see larger storage capacities but slower charging times. However, as potential of flexible geothermal power with energy storage will be the
long as the engineered reservoir remains confined, the results presented primary focus of our research on this subject going forward.
here should remain qualitatively accurate. Extending this reasoning, it
may be possible to exploit IRES in geothermal systems that would not
CRediT authorship contribution statement
traditionally be classified as ‘‘EGS’’. In formations with much higher
natural permeability and no artificial fractures, a large number of
triplets could be organized so as to impose artificial flow barriers on Wilson Ricks: Writing – original draft, Methodology, Software, For-
the more central wells, creating an effectively confined system like that mal analysis, Investigation. Jack Norbeck: Conceptualization, Fund-
proposed in Buscheck et al. [31]. Given the observed positive effect ing acquisition, Writing – review & editing, Methodology, Resources,
of reduced matrix permeability on storage potential, high-permeability Validation. Jesse Jenkins: Conceptualization, Methodology, Funding
systems may derive even greater benefits from flexible operation than acquisition, Supervision.

10
W. Ricks et al. Applied Energy 313 (2022) 118807

Declaration of competing interest plant-level flexible geothermal model developed in this paper to be
easily incorporated into a larger electricity system capacity expansion
Jack Norbeck declares that he is a representative of Fervo Energy, a model (typically formulated as an LP or a mixed-integer LP), which is
geothermal energy project development company. There are no other a priority for future work.
conflicts to declare. The LP formulation used to optimize investments and operations
for a flexible geothermal power plant must accurately represent the
Acknowledgments pressure and flow behaviors observed in numerical simulations, ideally
being capable of reproducing simulation results when executing an
This work was supported by the US Department of Energy Office identical operational strategy. It must do so while remaining compu-
of Science SBIR program under Award No. DE-SC0020823, and by tationally lean and entirely linear, neglecting most of the complicated
Princeton University’s Carbon Mitigation Initiative. The authors thank physics included in the numerical simulations. The following subsec-
Prof. Michael Celia for useful discussions and the two anonymous tions describe how a formulation that meets these requirements was
reviewers for helpful comments and review. developed and validated.

Appendix A. Computational modeling and flexible geothermal B.1. Reservoir pressure behavior
reservoir simulation
Our approach to capturing pressure behavior in response to pump-
In this work, we use a computational geothermal reservoir sim- ing takes advantage of several principles from the field of groundwater
ulator to evaluate the technical feasibility of the in-reservoir energy hydrology. The first, and most important, is that the transient pressure
storage operational strategy. We use a commercial reservoir simulation response at any fixed point in a static subsurface reservoir to constant
software package called ResFrac, which is capable of solving the cou- pumping on a well penetrating that reservoir is a consistent function
pled physical processes involved in geothermal reservoir production, directly proportional to the pumping rate [55]. That is to say, pumping
including fluid flow in porous and fractured media, fluid flow in at double the rate on a well will result in double the transient pressure
complex wellbores, heat transfer, and mechanical deformation of the response in the reservoir. Thus, if we know the transient pressure
subsurface [41,42]. Here, we provide a brief overview of the governing response curve corresponding to pumping on one of the wells in a
equations and numerical methods used in the reservoir simulator. A flexible geothermal reservoir at a single constant rate, we can find the
more detailed description of the simulator is provided in the ResFrac pressure response to pumping on that well at any constant rate.
users manual [42]. To extract this information, we run specially-designed numerical
The simulator used in this study is a fully-compositional, thermal, simulations for every set of subsurface conditions examined in this
and geomechanical reservoir model. The simulator integrates fluid paper. In each simulation, the geothermal reservoir is allowed to op-
flow in the wellbores, fluid flow in the reservoir, heat transfer in the erate under steady-state conditions (i.e. at a constant injection rate
reservoir, and mechanical deformation of the pore volume and fractures with no production curtailment) for a period of five years in order to
in the subsurface. The governing equations involve momentum balance ensure that no transient behaviors are present. The production rate is
for flow in the wellbore, mass balance for flow in the reservoir, energy then instantaneously lowered by a pre-prescribed amount, and kept at
balance for heat transfer in the reservoir, and momentum balance for that level for a period of 240 h. The system is allowed to come back
mechanical deformation of the reservoir. into equilibrium, and the injection rate is then increased by the same
The system is discretized numerically using one-dimensional ele- amount for an identical period of time. The transient pressure responses
ments along the wellbore, three-dimensional volumetric elements for at the injection and production well bottomholes are recorded during
the reservoir matrix, and two-dimensional surface area elements for both of these periods. We create piecewise linearizations for each
the fractures. Fluid flow and heat transfer are calculated using a finite of these four pressure response functions, at hourly timesteps, and
volume method. Fracture deformation is calculated using a boundary normalize each function to a unit pumping rate.
element method. In this study, we use a Cartesian mesh for the reservoir These functions give the pressure responses to constant pumping
matrix volume and a fracture mesh of rectangular elements. Solving rates, but to have a useful model of a flexible geothermal system
this system of equations allows us to calculate fluid pressure along the we must accurately capture the response to variable pumping rates.
wellbore (along with flowing friction), the fluid pressure distribution in The technique we use to accomplish this is linear superposition of
the reservoir and fracture volume, the temperature distribution in the pressure transients. This technique’s basic principle states that the
reservoir and fracture volume, and the mechanical deformation of the transient pressure response to multiple stepwise changes in pumping
fractures. rate on a given well is simply the superposition of the individual
In our simulations, the reservoir is assumed to be fully saturated pressure responses to each of the stepwise changes taken in isolation.
with single-phase water. Thermoelastic and poroelastic stresses caused The superposition technique is outlined in Reilly et al. [56], and is
by temperature and pressure changes in the matrix are neglected in valid for groundwater problems where the governing equations for
this study, therefore fracture deformation is caused by changes in fluid flow and boundary conditions are universally linear. These conditions
pressure within the fracture as well as mechanical interaction between are generally satisfied by confined reservoirs like those investigated
fractures. in this paper. One possible nonlinearity comes from the dilation of
fracture apertures in response to increased reservoir pressure, which
Appendix B. Development and validation of the linear optimiza- leads to non-constant fracture conductivity. As will be demonstrated,
tion model this nonlinearity is small and does not meaningfully affect the accuracy
of the superposition approach.
While simulation results are useful for evaluating the performance Based on the principle of superposition, we create a set of con-
of a flexible geothermal reservoir under an exogenously determined straints to model the injection and production well bottomhole pres-
operational strategy, the simulation software is too computationally sures (BHPs) as functions of the injection and production rates at
intensive to allow for endogenous optimization of plant operations. previous timesteps. These are constraints (C.4a) and (C.4b) in Ap-
Such an optimization would ideally be performed within a linear pro- pendix C. They express BHP for each well as the linear superposition of
gramming (LP) framework, which enables quick solutions to complex the transient pressure responses to changes in injection and production
optimization problems with tens of thousands of decision variables and rates over n previous timesteps. For this work, we choose 𝑛 = 200, as the
constraints. An LP formulation is also desirable because it allows the second derivative of the pressure response function is negligible after

11
W. Ricks et al. Applied Energy 313 (2022) 118807

200 h. Limiting the number of timesteps referenced in this constraint

reduces the runtime of the overall LP. Beyond this point, we assume
that the pressure changes at a constant rate, denoted 𝜔𝑠𝑠 in the model
formulation. The same value of 𝜔𝑠𝑠 is used for both wells, as both
experience convergent long-term behavior.

B.2. Maximum production flow rate

Reservoir simulation outputs are also used to calibrate the relation-

ship between production BHP and the maximum achievable production
flow rate at a given time step. Physically, production flow rate as
a function of production BHP is given by Bernoulli’s equation for
turbulent pipe flow with friction:
(𝑃𝐵𝐻 − 𝑃𝑊 𝐻 ) 𝑣2
+ 𝛥𝑧 = + 𝐻𝑓 𝑟𝑖𝑐𝑡𝑖𝑜𝑛 , (B.1)
𝜌𝑔 2𝑔
where 𝑃𝐵𝐻 is the production bottomhole pressure, 𝑃𝑊 𝐻 is the mini-
mum allowable value of production wellhead pressure, 𝑣 is flow veloc- Fig. B.10. The LP optimization model and the numerical simulation executing a series
ity in the well, 𝑧 is the vertical length of the well, and 𝐻𝑓 𝑟𝑖𝑐𝑡𝑖𝑜𝑛 is the of daily production cycles. Simulation results are shown as point data, while model
major head loss due to friction, given by the Darcy–Weisbach Equation: results are shown as continuous data.

𝐻𝑓 𝑟𝑖𝑐𝑡𝑖𝑜𝑛 𝑓 𝐷 𝑣2
= , (B.2) governed by the Bernoulli equation as described in Appendix B.2, and is
𝐿 2 𝑔𝐷
found explicitly by fitting this equation to reservoir simulation results.
where 𝐿 is the length of the well, 𝑓𝐷 is the Darcy friction factor, and Because this function is highly nonlinear, a step function is developed
𝐷 is the diameter of the well. The maximum achievable production to represent it in the LP model. The range of possible injection pumping
flow rate 𝑄𝑝𝑟𝑜𝑑𝑚𝑎𝑥 for a given value of 𝑃𝐵 𝐻 is directly proportional to 𝑣 rates is divided into 53 l/s intervals, and it is assumed that pumping
as given by the above equations. Rather than explicitly calculating the at any point in each interval induces a 𝛥𝑃𝑖 equal to the value of the
friction factor and other quantities to find the dependence of 𝑄𝑝𝑟𝑜𝑑𝑚𝑎𝑥 nonlinear function at the uppermost point of that interval. The practical
on 𝑃𝐵𝐻 , we use these equations to derive a fit of the form: result of this modeling choice is that the value of injection WHP,
√
𝑄𝑝𝑟𝑜𝑑𝑚𝑎𝑥 = 𝑎 × 𝑃𝐵𝐻 + 𝑏, (B.3) which increases nonlinearly with injection rate, is overestimated in the
intervals between the calibration injection rates. This choice biases the
where 𝑎 and 𝑏 are unknown fit constants. We fit reservoir simulation model toward pumping at exactly those rates, where the distortion is
results to this function to find values for 𝑎 and 𝑏, thereby expressing minimal, but still allows for continuous modulation of injection rate
production flow rate as a function of production BHP. Eq. (B.3) is when such a strategy is optimal. The distortion could be minimized
nonlinear, so we cannot explicitly include it as a constraint in our by including more injection ‘‘segments’’, but including too many of
LP model. Instead, because the curve is highly linear in the region these would increase the model size significantly. In our model, we
with which we are concerned, we approximate it via a single linear choose to use six injection segments as a compromise between these
constraint. competing interests. Injection rate is capped at double the steady-state
injection rate, at which point injection wellhead pressure begins to
B.3. Validation of the pressure and flow formulations become excessive.
Not only is the physical relationship between injection WHP and
As stated above, the goal in developing a LP optimization model injection rate nonlinear, but the relationship between these quantities
for flexible geothermal was to be able to reproduce simulation re- and pumping power is as well. Pumping power is directly proportional
sults with a high degree of accuracy. To validate the LP pressure to the product of injection rate and injection WHP, a relationship which
and flow formulations described in Appendices B.1 and B.2, we force cannot be explicitly modeled in an LP formulation if both quantities are
our optimization model to perform the same set of operations as in variable. To capture the parasitic load from injection we employ a lin-
the original numerical simulations. Model outputs plotted alongside ear approximation which overestimates the pumping power to a small
simulation outputs are shown in Fig. 2 of the main paper, and in degree. The explicit constraint formulation is detailed in Appendix C.
Fig. B.10. There is strong agreement between the model and simulation Overall, our LP model is formulated in such a way that it always
results during both long and repeated short charging cycles. The fact returns a value for injection pumping power greater than or equal to
that agreement is strong even during periods of variable production the true value for the same injection rate and injection WHP. Because
rate serves as a validation of the superposition approach described in injection pumping power is a small portion of the plant’s net power
Appendix B.1 and the linearization in Appendix B.2, and confirmation output (∼10%), this overestimation has only a minor effect on model
that any nonlinearities present in the system are not very significant. results. The effect of this choice is that the linear optimal objective (and
therefore the calculated benefits attributable to flexibility), will be a
B.4. Injection pumping requirements fairly tight lower bound for the true nonlinear optimum.

The optimization model must endogenously calculate the parasitic B.5. Power plant operational constraints
load due to injection pumping, and by extension the injection WHP
required to pump at a given injection rate. This effect is considered Several abstractions are introduced in order to accurately represent
separately from the injection well BHP, which is calculated via the the Organic Rankine Cycle surface plant within a linear optimization
linear superposition approach described in the previous section. The framework. First, the plant is assumed to be capable of operating
injection WHP at a given timestep is calculated as the sum of the continuously from 0%–110% capacity at a constant efficiency. ORCs,
injection BHP and the pressure differential 𝛥𝑃𝑖 across the injection unlike large thermal generators, have very quick startup times. Thus,
well, the latter of which depends exclusively on the injection rate. they are capable of rapidly and efficiently adjusting their output contin-
The pressure differential is a nonlinear function of the injection rate, uously over the range of 10%–100% capacity and in a stepwise manner

12
W. Ricks et al. Applied Energy 313 (2022) 118807

from 0%–10%. Removing this lower discontinuity has a minimal effect Table C.6
Model indices.
on model outputs (the model very infrequently chooses to produce in
Notation Description
this region, typically going all the way to zero), while allowing us to
keep the model linear and continuous. The extra 10% peaking capacity 𝑡 ∈ [0, … , 8760] Where 𝑡 denotes an hour in the modeled weather year. Hour
0 sets initial conditions.
is justified by the fact that ORCs can operate at high efficiencies off- 𝑛 ∈ [1, … , 6] Where 𝑛 denotes one of six injection pumping levels.
design, and have been shown to be capable of accepting up to 10%
higher geofluid mass-flow rates with minimal efficiency loss [43].
Plant power output is determined based on the geofluid mass flow
rate, difference in enthalpy between production and injection condi- Table C.7
Model variables and expressions.
tions, and net plant thermo-electric conversion efficiency (including
Notation Description
plant parasitic loads, but neglecting wellfield pumping power). Plant
𝑦𝑐𝑎𝑝 Surface plant net electric generating capacity.
efficiency is calculated for given inlet and outlet temperatures based
𝑦𝑖𝑐𝑎𝑝 Surface plant grid interconnection capacity.
on results from genGEO, an open-source geothermal combined reser- 𝑦𝑝𝑐𝑎𝑝 Injection pump power capacity.
voir, well, power plant, cost, and financing simulator [57], presented 𝑥𝑝𝑟𝑜𝑑
𝑡 Volumetric production flow rate at hour 𝑡.
in Adams et al. [50]. For the conditions used in this paper, namely an 𝑥𝑖𝑛𝑗
𝑡,𝑛 Deviation from the steady-state injection rate at hour 𝑡 and injection
level 𝑛.
inlet temperature of 210 C and an outlet temperature of 70 C, net plant
𝑥𝑖𝐵𝐻𝑃
𝑡
Bottomhole pressure in the injection well at hour 𝑡.
efficiency is 13.80%. 𝑥𝑝𝐵𝐻𝑃 Bottomhole pressure in the production well at hour 𝑡.
𝑡
𝑒𝑖𝑛𝑗
𝑡 Volumetric injection flow rate at hour 𝑡.
𝑒𝑖𝑝𝑤𝑟
𝑡 Parasitic load due to injection pumping at hour 𝑡.
B.6. Cost and financing assumptions
𝑒𝑝𝑝𝑤𝑟
𝑡 Net surface plant generation at hour 𝑡.

Plant specific cost for the base case is calculated for the same
set of conditions based on genGEO results, at $1950/kW net. This
cost does not include production pumps, which are not used in the
design considered in this paper, and cost of the field gathering system. C.2. Objective function
We assume a field gathering system cost of $100/kW. Because only
some plant component costs scale with production flow rate, and costs The Objective Function in Eq. (C.1) maximizes plant revenue over
of others scale in nonlinear fashion, we assume that plant capacity a weather year by co-optimizing investment and operational decisions.
oversizing has a specific cost equal to 80% that of the base plant.
Injection pump specific cost is set based on genGEO parameters at (
$400/kW, where this cost is in relation to the power consumed by max (C.1a)
𝑥,𝑦
the pump. For surface facilities financing, we assume that interest ( )
∑
equal to 6% of plant overnight cost is accumulated during a one year (𝑒𝑝𝑝𝑤𝑟
𝑡 − 𝑒𝑖𝑝𝑤𝑟
𝑡 ) ⋅ 𝜌𝑡 − (C.1b)
construction period, and that during its 30 year capital recovery period 𝑡∈[1∶8760]
)
the plant has a nominal weighted average cost of capital (WACC) of ( )
𝑐𝑎𝑝 𝑐𝑎𝑝1 𝑝𝑙𝑎𝑛𝑡 𝑖𝑐𝑎𝑝 𝑐𝑎𝑝2 𝑡𝑟𝑎𝑛𝑠 𝑝𝑐𝑎𝑝 𝑐𝑎𝑝3 𝑝𝑢𝑚𝑝
(𝑦 −𝑝 )⋅𝑐 + (𝑦 −𝑝 )⋅𝑐 + (𝑦 −𝑝 )⋅𝑐
5.5%, and a real WACC of 2.93%. These assumptions result in an
annual capital recovery factor of 5.1%. In addition to this annuitized
(C.1c)
investment cost, it is assumed that surface facilities oversizing leads
to proportionately increased maintenance costs. These are calculated The method consists of maximizing the sum of all hourly operational
to be 1.8% of plant CAPEX annually, a value taken from the GETEM revenues over the year (C.1b), while minimizing the annuitized costs
geothermal development simulator produced by NREL [51]. It should of investments in surface plant, injection pump, and interconnection
be noted that these maintenance costs assume the presence of a down- oversizing (C.1c). Operational revenue for any hour 𝑡 is given by the
hole production pump, which is not used in our model. The annual product of the net electrical output of the plant, equal to the net power
cost of taxes and insurance is set at 0.75% of plant CAPEX, a value also plant generation less the parasitic power used for injection pumping,
taken from GETEM. It is assumed that surface facilities oversizing does and the price of electricity at that hour. Total cost of surface plant
not result in increased labor costs. We further account for the cost of oversizing is calculated as the difference between the oversized gross
additional grid interconnection capacity. The capital cost of geothermal plant capacity and the baseload gross plant capacity, multiplied by the
interconnection is set at $130/kW, with a capital recovery period of annuitized specific cost of surface plant oversizing. Total cost of injec-
60 years and a nominal WACC of 6.9%, as reported in Gorman et al. tion pump oversizing is the difference between the oversized injection
[58]. All financial information used in this paper is presented in 2019 pump maximum power and the baseload injection pump maximum
$USD. power, multiplied by the annuitized specific cost of the injection pump.
Total cost of interconnection oversizing is the difference between the
oversized interconnection capacity and the baseload interconnection
Appendix C. Optimization model formulation
capacity, multiplied by the annuitized specific cost of interconnection.

The following sections present the explicit formulation of the LP

C.3. Constraints
optimization model used in this work. All parameter values are for the
base-case model, and some vary between sensitivity cases.
The optimization function defined in Eq. (C.1) is subject to different
sets of constraints that define the feasible solution space for the decision
C.1. Model indices, variables, expressions, and parameters variables. Without these constraints, the model would converge to
unbounded levels of power production and the objective value would
See Tables C.6–C.8. be infinite.

13
W. Ricks et al. Applied Energy 313 (2022) 118807

Table C.8
Model parameters.
Notation Description
𝑝𝑐𝑎𝑝1 Baseload plant net electric generating capacity excluding injection pumping load, 11.1 MW
𝑝𝑐𝑎𝑝2 Baseload plant net electric generating capacity including injection pumping load, 10.1 MW
𝑝𝑐𝑎𝑝3 Baseload plant injection pumping power, 1.0 MW
𝑝𝑝𝑒𝑎𝑘 Peaking factor representing maximum usable geofluid flow, 1.1.
𝑝𝑖𝐵𝐻𝑃 𝑚𝑖𝑛 Minimum and starting injection bottomhole pressure, 31.66 MPa.
𝑝𝑖𝐵𝐻𝑃 𝑚𝑎𝑥 Maximum injection bottomhole pressure, 35.16 MPa.
𝑝𝑝𝐵𝐻𝑃 𝑚𝑖𝑛 Minimum and starting production bottomhole pressure, 28.63 MPa.
𝑝𝑖𝛥𝑃 Pressure differential across injection well at steady-state injection rate, 26.17 MPa.
𝑝𝑖𝑠𝑒𝑔 Range of each injection pumping interval, 53 l/s.
𝑝𝑖𝑊 𝐻𝑃 𝑚𝑖𝑛 Minimum injection wellhead pressure due to pumping in the 0–53 l/s interval, 2.34 MPa.
𝑝𝑖𝑊 𝐻𝑃 2 Additional injection wellhead pressure due to pumping in the 53–106 l/s interval, 1.18 MPa.
𝑝𝑖𝑊 𝐻𝑃 3 Additional injection wellhead pressure due to pumping in the 106–159 l/s interval, 3.14 MPa.
𝑝𝑖𝑊 𝐻𝑃 4 Additional injection wellhead pressure due to pumping in the 159–212 l/s interval, 5.90 MPa.
𝑝𝑖𝑊 𝐻𝑃 5 Additional injection wellhead pressure due to pumping in the 212–265 l/s interval, 9.44 MPa.
𝑝𝑖𝑊 𝐻𝑃 6 Additional injection wellhead pressure due to pumping in the 265–318 l/s interval, 13.76 MPa.
𝑝𝑖𝑠𝑠 Steady-state injection flow rate, 159.0 l/s.
𝑝𝑝𝑠𝑠 Steady-state production flow rate, 158.1 l/s.
𝑖
𝛼[0∶200] Hourly coefficients of the transient injection bottomhole pressure response function corresponding to changes in injection flow rate.
𝑝
𝛼[1∶200] Hourly coefficients of the transient injection bottomhole pressure response function corresponding to changes in production flow rate.
𝑖
𝛽[1∶200] Hourly coefficients of the transient production bottomhole pressure response function corresponding to changes in injection flow rate.
𝑝
𝛽[0∶200] Hourly coefficients of the transient production bottomhole pressure response function corresponding to changes in production flow rate.
𝜔𝑠𝑠 Constant universal pressure response rate after 200 timesteps.
𝛾 Slope of the linear 𝑄𝑝𝑟𝑜𝑑𝑚𝑎𝑥 constraint.
𝛿 Intercept of the linear 𝑄𝑝𝑟𝑜𝑑𝑚𝑎𝑥 constraint.
𝜂 𝑔𝑓 Surface plant specific power, 88.09 kW/(l/s)𝑔𝑒𝑜𝑓 𝑙𝑢𝑖𝑑 .
𝜂 𝑝𝑢𝑚𝑝 Injection pump power conversion factor, 1.14 kW/(MPa l/s).
𝜌𝑡 Electricity price at hour 𝑡.
𝑐 𝑝𝑙𝑎𝑛𝑡 Annuitized fixed costs associated with surface plant oversizing, $132 804/MW𝑛𝑒𝑡,𝑝𝑙𝑎𝑛𝑡 -yr.
𝑐 𝑡𝑟𝑎𝑛𝑠 Annuitized fixed costs associated with interconnection oversizing, $5629/MW-yr.
𝑐 𝑝𝑢𝑚𝑝 Annuitized fixed costs associated with injection pump oversizing, $32 130/MW-yr.

C.3.1. Variable bounds operational variables at timestep 0 represent the status of a plant in
These constraints establish hard bounds on the decision variables steady-state (i.e. inflexible) operation.
of the optimization problem. Some variables are only bounded on one
𝑥𝑝𝑟𝑜𝑑
0
= 𝑝𝑝𝑠𝑠 (C.3a)
end.
𝑥𝑖𝑛𝑗
0,𝑛
=0 ∀𝑛 ∈ [1 ∶ 6] (C.3b)

𝑦𝑐𝑎𝑝 ≥ 𝑝𝑐𝑎𝑝1 (C.2a) 𝑥𝑝𝐵𝐻𝑃

0
= 𝑝𝑝𝐵𝐻𝑃 𝑚𝑖𝑛 (C.3c)

𝑦𝑖𝑐𝑎𝑝 ≥ 𝑝𝑐𝑎𝑝2 (C.2b) 𝑥𝑖𝐵𝐻𝑃

0
=𝑝 𝑖𝐵𝐻𝑃 𝑚𝑖𝑛
(C.3d)

𝑦𝑝𝑐𝑎𝑝 ≥ 𝑝𝑐𝑎𝑝3 (C.2c) Eqs. (C.3a) and (C.3b) set starting production and injection flow
rates to their steady-state values, while Eqs. (C.3c) and (C.3d) do the
𝑥𝑝𝑟𝑜𝑑
𝑡 ≥0 ∀𝑡 ∈ [0 ∶ 8760] (C.2d)
same for the production and injection bottomhole pressures.
𝑝 𝑖𝑠𝑒𝑔
≥ 𝑥𝑖𝑛𝑗
𝑡,𝑛 ≥0 ∀𝑡 ∈ [0 ∶ 8760], 𝑛 ∈ [1 ∶ 6] (C.2e)
𝑖𝐵𝐻𝑃 𝑚𝑎𝑥 C.3.3. Pressure formulation
𝑝 ≥ 𝑥𝑖𝐵𝐻𝑃
𝑡
𝑖𝐵𝐻𝑃 𝑚𝑖𝑛
≥𝑝 ∀𝑡 ∈ [0 ∶ 8760] (C.2f)
The constraints describing pressure evolution in response to changes
𝑥𝑝𝐵𝐻𝑃
𝑡 ≥ 𝑝𝑝𝐵𝐻𝑃 𝑚𝑖𝑛 ∀𝑡 ∈ [0 ∶ 8760] (C.2g) in pumping are central to the accurate representation of geothermal
flexibility. These represent the pressure at each well bottomhole as
Eqs. (C.2a)–(C.2c) constrain the grid capacity of the surface plant
a superposition of linearized pressure response functions referencing
built by the model to be larger than that of an inflexible baseload plant changes in pumping rates during the previous 200 timesteps. The
operating on the same wellfield, and the injection pump power capacity motivation and justification for this approach is described in greater
to be larger than that of a baseload plant. Eq. (C.2d) forces production detail in Appendix B.1.
rate to always be positive, preventing negative production during times
of negative electricity prices. Eq. (C.2e) establishes bounds on each of 𝑥𝑖𝐵𝐻𝑃
𝑡 =𝑥𝑖𝐵𝐻𝑃
𝑡−1
+ (𝑒𝑖𝑛𝑗 𝑖𝑛𝑗 𝑖
𝑡 − 𝑒𝑡−1 ) ⋅ (𝛼0 − 𝜔 )+
𝑠𝑠
∑ ( 𝑖𝑛𝑗
the six injection pumping levels: each variable controls pumping across (𝑒𝑡−𝑎 − 𝑒𝑖𝑛𝑗 ) ⋅ (𝛼𝑎𝑖 − 𝜔𝑠𝑠 )−
𝑡−𝑎−1
a 53 l/s range. Finally, Eqs. (C.2f) and (C.2g) establish a maximum 𝑎∈[1∶200] ∀𝑡 ∈ [1 ∶ 8760] (C.4a)
)
injection BHP and minimum production and injection BHPs. The max-
(𝑥𝑝𝑟𝑜𝑑 𝑝𝑟𝑜𝑑 𝑝
𝑡−𝑎 − 𝑥𝑡−𝑎−1 ) ⋅ (𝛼𝑎 − 𝜔 )
𝑠𝑠
+
imum injection BHP represents a cap on reservoir pressurization due
to concerns over induced seismicity, while the minimum BHPs (equal (𝑒𝑖𝑛𝑗
𝑡−1
− 𝑥𝑝𝑟𝑜𝑑
𝑡−1
) ⋅ 𝜔𝑠𝑠
to the BHPs under steady-state inflexible operation) limit the pressure 𝑥𝑝𝐵𝐻𝑃 = 𝑥𝑝𝐵𝐻𝑃 − (𝑥𝑝𝑟𝑜𝑑 − 𝑥𝑝𝑟𝑜𝑑 ) ⋅ (𝛽0𝑝 − 𝜔𝑠𝑠 )+
𝑡 𝑡−1 𝑡 𝑡−1
drawdown to prevent fracture closure. ∑ ( 𝑖𝑛𝑗
(𝑒𝑡−𝑎 − 𝑒𝑖𝑛𝑗
𝑡−𝑎−1
) ⋅ (𝛽𝑎𝑖 − 𝜔𝑠𝑠 )−
𝑏∈[1∶200] ∀𝑡 ∈ [1 ∶ 8760] (C.4b)
C.3.2. Initial conditions )
(𝑥𝑝𝑟𝑜𝑑 𝑝𝑟𝑜𝑑 𝑝 𝑠𝑠
𝑡−𝑎 − 𝑥𝑡−𝑎−1 ) ⋅ (𝛽𝑎 − 𝜔 ) +
Most of the state variables in the model are formulated in a man-
ner that references previous states. The initial conditions set for all (𝑒𝑖𝑛𝑗
𝑡−1
− 𝑥𝑝𝑟𝑜𝑑
𝑡−1
) ⋅ 𝜔𝑠𝑠

14
W. Ricks et al. Applied Energy 313 (2022) 118807

Eq. (C.4a) describes the injection BHP at hour 𝑡 as a function of

the injection BHP at the previous hour and the cumulative effects Table D.9
of injection and production flow rate changes at previous timesteps, Statistics for historical and modeled future price series. All prices are given in
$/MWh.
and Eq. (C.4b) does the same for production BHP. The first line of
Price series Hours of Avg. Daily STD Avg. Weekly
each constraint references the pressure at the previous timestep, as
≤0 pricing STD
well as an ‘‘instantaneous’’ (i.e. on a timescale much less than an
2019 AZ 615 46.3 73.2
hour) change in pressure due to pumping on the same well at the
2019 N-CA 359 23.4 37.7
current timestep. The second and third lines capture the superimposed 2019 S-CA 1162 35.2 50.6
effects of all production and injection rate changes at the previous 2019 ID 300 27.2 46.8
200 timesteps. The last line references the difference between the net 2019 NV 467 39.4 62.3
pumping rate into the reservoir at the previous timestep and the steady- 2019 OR 184 12.3 20.2
2019 TX 309 46.2 80.8
state net pumping rate, which captures the lingering cumulative effects 2019 UT 359 27.7 46.8
of changes in pumping at all previous timesteps. The development and 2030 BAU baseline 851 22.0 23.4
validation of these pressure constraints are described in greater detail 2030 BAU winter peaking 0 15.9 21.9
in Appendix B.1. 2030 BAU high solar 607 27.5 31.3
2030 BAU high wind 917 23.6 30.3
2030 CO2 tax baseline 338 33.1 38.2
C.3.4. Production rate and generation 2030 CO2 tax winter peaking 245 29.3 36.4
Production flow rate is limited by production well BHP, while usable 2030 CO2 tax high solar 1393 39.3 44.7
production flow is constrained by the surface plant capacity and grid 2030 CO2 tax high wind 1863 33.6 40.1
interconnection. The physical constraints on maximum production flow
rate and the conversion between production flow and electrical power
are described in greater detail in Appendix B.2.

𝑒𝑝𝑝𝑤𝑟
𝑡 = 𝑥𝑝𝑟𝑜𝑑
𝑡 ⋅ 𝜂 𝑔𝑓 ∀𝑡 ∈ [1 ∶ 8760] (C.5a) ∑
𝑒𝑖𝑛𝑗 𝑖𝑠𝑠
(C.6g)
𝑡 ≤ 8760𝑝

𝑒𝑝𝑝𝑤𝑟 𝑝𝑒𝑎𝑘 𝑐𝑎𝑝 𝑡∈[1∶8760]

𝑡 ≤𝑝 ⋅𝑦 ∀𝑡 ∈ [1 ∶ 8760] (C.5b)
Eq. (C.6a) formulates injection rate at hour 𝑡 in terms of deviations
𝑒𝑝𝑝𝑤𝑟
𝑡 − 𝑒𝑖𝑝𝑤𝑟
𝑡 ≤ 𝑦𝑖𝑐𝑎𝑝 ∀𝑡 ∈ [1 ∶ 8760] (C.5c) from the steady-state injection rate, 𝑝𝑖𝑠𝑠 . Three variables 𝑥𝑖𝑛𝑗 control
𝑡,[1∶3]
𝑥𝑝𝑟𝑜𝑑
𝑡 ≤ 𝛾 ⋅ 𝑥𝑝𝐵𝐻𝑃
𝑡 +𝛿 ∀𝑡 ∈ [1 ∶ 8760] (C.5d) reductions in injection rate, with each able to reduce total injection by
1/3 of the steady-state injection rate. 𝑥𝑖𝑛𝑗
𝑡,[4∶6]
have a similar function, but
Eq. (C.5a) sets the power generation associated with a given pro- each increase total injection by 1/3 of the steady-state injection rate.
duction flow rate to be the product of the flow rate and the surface Each of these six 𝑥𝑖𝑛𝑗
𝑡,𝑛 corresponds to a specific injection rate segment,
plant specific power. Eq. (C.5b) constrains this power to be less than i.e. 0–53 l/s, 53–106 l/s, etc. Eqs. (C.6b) and (C.6c) ensure that these
the product of the net generating capacity of the surface plant and the segments are activated in the correct order. Eq. (C.6d) prevents the
peaking factor. Eq. (C.5c) constrains the plant’s total net generation, model from simultaneously raising and lowering the injection rate,
equal to generation less parasitic load, to be less than the capacity which might generate additional revenue during times of negative
of the its grid interconnection at all timesteps. Eq. (C.5d) bounds electricity pricing. Eq. (C.6e) calculates the total injection pumping
the production flow rate from above based on a linearized nonlinear load at hour 𝑡. The first line expresses load due to pumping at the
constraint, as detailed in Appendix B.2. steady-state injection rate on the current injection BHP as the product
of these quantities and the pump power factor 𝜂 𝑝𝑢𝑚𝑝 . The next three
C.3.5. Injection rate and parasitic load lines deal with reductions in this load due to reductions in injection
Injection pumping rate and the resulting parasitic load are the most rate, which are calculated based on the minimum injection BHP and
difficult quantities to capture in a LP model, as the actual physical the current injection rate. The final three lines deal with increases in
relationships are highly nonlinear. We formulate the injection load in injection load due to increased injection rate, which are calculated
such a way that it is always greater than or equal to the actual load based on the maximum injection BHP and the current injection rate.
that would exist at the same pumping rate and wellhead pressure. The result of this formulation is that reductions in parasitic load due to
reductions in injection rate are always modeled as less than or equal
∑ ∑
𝑒𝑖𝑛𝑗
𝑡 =𝑝
𝑖𝑠𝑠
− 𝑥𝑖𝑛𝑗
𝑡,𝑛 + 𝑥𝑖𝑛𝑗
𝑡,𝑛 ∀𝑡 ∈ [1 ∶ 8760] to their true physical values, while additional injection load due to
𝑛∈[1∶3] 𝑛∈[4∶6]
increased injection rate is always modeled as greater than or equal to
its true value. Though it necessarily overestimates the parasitic load
at most pressures and injection rates, this formulation is still able
(C.6a) to capture most of the costs and benefits of increasing or decreasing
𝑥𝑖𝑛𝑗
𝑡,3
≥ 𝑥𝑖𝑛𝑗
𝑡,2
≥ 𝑥𝑖𝑛𝑗
𝑡,1
∀𝑡 ∈ [1 ∶ 8760] injection rate, respectively. When injection rate is kept at its steady-
(C.6b) state value, Eq. (C.6e) returns the exact value of the parasitic load.
𝑥𝑖𝑛𝑗 ≥ 𝑥𝑖𝑛𝑗 ≥ 𝑥𝑖𝑛𝑗 ∀𝑡 ∈ [1 ∶ 8760] Eq. (C.6f) requires that the model build enough pump capacity to
𝑡,4 𝑡,5 𝑡,6
meet injection power requirements at all points throughout the year.
(C.6c)
Finally, Eq. (C.6g) requires that the average injection rate over the
𝑥𝑖𝑛𝑗
𝑡,[1∶3]
+ 𝑥𝑖𝑛𝑗
𝑡,[4∶6]
≤ 𝑝𝑖𝑠𝑒𝑔 ∀𝑡 ∈ [1 ∶ 8760]
modeled year be less than or equal to the steady-state injection rate.
(C.6d) This requirement ensures that the flexible geothermal plant experiences
𝑒𝑖𝑝𝑤𝑟
𝑡 = (𝑥𝑖𝐵𝐻𝑃
𝑡
− 𝑝𝑖𝛥𝑃 ) ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑝𝑖𝑠𝑠 − a thermal drawdown rate no greater than that of an inflexible plant.
𝑝𝑖𝑊 𝐻𝑃 𝑚𝑖𝑛 ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑥𝑖𝑛𝑗
𝑡,1
−
(𝑝𝑖𝑊 𝐻𝑃 𝑚𝑖𝑛 + 2𝑝𝑖𝑊 𝐻𝑃 2 ) ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑥𝑖𝑛𝑗
𝑡,2
−
(𝑝 𝑖𝑊 𝐻𝑃 𝑚𝑖𝑛 + 3𝑝 𝑖𝑊 𝐻𝑃 3 − 2𝑝𝑖𝑊 𝐻𝑃 2 ) ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑥𝑖𝑛𝑗 + ∀𝑡 ∈ [1 ∶ 8760]
Appendix D. Supplementary results
𝑡,3
(𝑝𝑖𝐵𝐻𝑃 𝑚𝑎𝑥 − 𝑝𝑖𝛥𝑝 + 4𝑝𝑖𝑊 𝐻𝑃 4 − 3𝑝𝑖𝑊 𝐻𝑃 3 ) ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑥𝑖𝑛𝑗 𝑡,4
+
(𝑝𝑖𝐵𝐻𝑃 𝑚𝑎𝑥 − 𝑝𝑖𝛥𝑝 + 5𝑝𝑖𝑊 𝐻𝑃 5 − 4𝑝𝑖𝑊 𝐻𝑃 4 ) ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑥𝑖𝑛𝑗 𝑡,5
+
(𝑝𝑖𝐵𝐻𝑃 𝑚𝑎𝑥 − 𝑝𝑖𝛥𝑝 + 6𝑝𝑖𝑊 𝐻𝑃 6 − 5𝑝𝑖𝑊 𝐻𝑃 5 ) ⋅ 𝜂 𝑝𝑢𝑚𝑝 ⋅ 𝑥𝑖𝑛𝑗 𝑡,6 D.1. Price series analysis
(C.6e)
𝑒𝑖𝑝𝑤𝑟
𝑡 ≤ 𝑦𝑝𝑐𝑎𝑝 ∀𝑡 ∈ [1 ∶ 8760] Fig. D.11 shows price duration curves for all 16 historical and
(C.6f) modeled future price series used in this study. Table D.9 provides

15
W. Ricks et al. Applied Energy 313 (2022) 118807

Fig. D.11. Plotted price duration curves for 16 historical and modeled future price series.

Fig. D.13. Same as Fig. D.12, showing correlation with daily average standard
Fig. D.12. Number of hours with electricity prices less than or equal to $0/MWh, vs.
deviation of prices.
relative geothermal energy value improvement from in-reservoir energy storage, for the
base case model. A linear fit is shown in red, along with its r-value.

statistics for each price series, including the number of hours for which
prices are ≤ 0/MWh, and average daily and weekly standard deviations.
Figs. D.12, D.13, and D.14 plot various statistics against the relative
energy value improvement from in-reservoir energy storage for each
price series. As shown in Fig. D.12, there is a fairly strong correlation
between the number of hours with zero or negative prices and the
level of energy value improvement from IRES. The correlation between
the standard deviations and the level of energy value improvement is
weaker, with the daily standard deviation being the stronger indicator.

D.2. Energy storage efficiency

Fig. D.15 illustrates the yearly average round-trip efficiency for

optimized operation on each price series under each sensitivity case.
Values range from 61% to 91%, with an average across all runs of 81%. Fig. D.14. Same as Fig. D.12, showing correlation with weekly average standard
deviation of prices.

16
W. Ricks et al. Applied Energy 313 (2022) 118807

Fig. D.15. Round-trip efficiency for in-reservoir energy storage across all price series and sensitivity cases.

References [15] Denholm P, O’Connell M, Brinkman G, Jorgenson J. Overgeneration from

solar energy in california: A field guide to the duck chart. Technical Report
[1] Sepulveda N, Jenkins J, de Sisternes F, Lester R. The role of firm low- NREL/TP-6A20-65023, Golden, CO: National Renewable Energy Laboratory
carbon electricity resources in deep decarbonization of power generation. Joule (NREL); 2015.
2018;2(11):2403–20. http://dx.doi.org/10.1016/j.joule.2018.08.006. [16] Mills AD, Levin T, Wiser R, Seel J, Botterud A. Impacts of variable renewable
energy on wholesale markets and generating assets in the united states: A review
[2] Baik E, Chawla KP, Jenkins JD, Kolster C, Patankar NS, Olson A, Ben-
of expectations and evidence. Renew Sustain Energy Rev 2020;120:109670.
son SM, Long JC. What is different about different net-zero carbon electric-
http://dx.doi.org/10.1016/j.rser.2019.109670, URL: https://www.sciencedirect.
ity systems? Energy Clim Change 2021;2:100046. http://dx.doi.org/10.1016/
com/science/article/pii/S1364032119308755.
j.egycc.2021.100046, URL: https://www.sciencedirect.com/science/article/pii/
S2666278721000234. [17] Lund PD, Lindgren J, Mikkola J, Salpakari J. Review of energy system flexibility
measures to enable high levels of variable renewable electricity. Renew Sus-
[3] McDonald RI, Fargione J, Kiesecker J, Miller WM, Powell J. Energy sprawl or
tain Energy Rev 2015;45:785–807. http://dx.doi.org/10.1016/j.rser.2015.01.057,
energy efficiency: Climate policy impacts on natural habitat for the United States
URL: https://www.sciencedirect.com/science/article/pii/S1364032115000672.
of america. PLOS ONE 2009;4(8):1–11. http://dx.doi.org/10.1371/journal.pone.
[18] Denholm P, Hand M. Grid flexibility and storage required to achieve very high
0006802.
penetration of variable renewable electricity. Energy Policy 2011;39(3):1817–30.
[4] Annual Energy Outlook 2021. Technical Report, Washington, D.C.: U. S. Energy
http://dx.doi.org/10.1016/j.enpol.2011.01.019, URL: https://www.sciencedirect.
Information Agency (EIA); 2021.
com/science/article/pii/S0301421511000292.
[5] Initial Results from the 2020 U.S. Geothermal Power Production and District
[19] Kondziella H, Bruckner T. Flexibility requirements of renewable energy based
Heating Market Report. Technical Report, Golden, CO: National Renewable
electricity systems – a review of research results and methodologies. Renew Sus-
Energy Laboratory (NREL); 2021.
tain Energy Rev 2016;53:10–22. http://dx.doi.org/10.1016/j.rser.2015.07.199,
[6] Williams C, Reed M, Mariner R. A review of methods applied by the U.S.
URL: https://www.sciencedirect.com/science/article/pii/S1364032115008643.
geological survey in the assessment of identified geothermal resources. Technical
[20] Mai T, Hand MM, Baldwin SF, Wiser RH, Brinkman GL, Denholm P, Ar-
Report 2008–1296, Menlo Park, CA: U.S. Geological Survey (USGS); 2008.
ent DJ, Porro G, Sandor D, Hostick DJ, Milligan M, DeMeo EA, Bazilian M.
[7] Williams C, Reed M, Mariner R, DeAngelo J, Galanis S. Assessment of moderate- Renewable electricity futures for the United States. IEEE Trans Sustain Energy
and high-temperature geothermal resources of the United States. Technical 2014;5(2):372–8. http://dx.doi.org/10.1109/TSTE.2013.2290472.
Report 2008–3082, Menlo Park, CA: U.S. Geological Survey (USGS); 2008.
[21] Palmintier BS, Webster MD. Impact of operational flexibility on electricity
[8] The future of geothermal energy. Technical Report INL/EXT-06-11746, Idaho generation planning with renewable and carbon targets. IEEE Trans Sustain
Falls, ID: Idaho National Laboratory; 2006. Energy 2016;7(2):672–84. http://dx.doi.org/10.1109/TSTE.2015.2498640.
[9] Augustine C. Update to enhanced geothermal system resource potential estimate. [22] Jones L. Renewable energy integration: practical management of variability,
In: Presented at the 40th GRC annual meeting. Sacramento, CA, 2016; 2016. uncertainty, and flexibility in power grids. 2nd ed.. Elsevier; 2017.
[10] GeoVision. Technical Report, U.S. Department of Energy (DOE); 2019. [23] Bloom A, Townsend A, Palchak D, Novacheck J, King J, Barrows C, Ibanez E,
[11] Matek B. Flexible opportunities with geothermal technology: Barriers and O’Connell M, Jordan G, Roberts B, Draxl C, Gruchalla K. Eastern renewable
opportunities. Electr J 2015;28:45–51. generation integration study. Technical Report NREL/TP-6A20-64472, Golden,
[12] Jafari M, Korpås M, Botterud A. Power system decarbonization: Impacts of CO: National Renewable Energy Laboratory; 2016.
energy storage duration and interannual renewables variability. Renew Energy [24] Jenkins J, Zhou Z, Ponciroli R, Vilim R, Ganda F, de Sisternes F, Bot-
2020;156:1171–85. http://dx.doi.org/10.1016/j.renene.2020.04.144, URL: https: terud A. The benefits of nuclear flexibility in power system operations with
//www.sciencedirect.com/science/article/pii/S0960148120306820. renewable energy. Appl Energy 2018;222:872–84. http://dx.doi.org/10.1016/j.
[13] Larson E, Greig C, Jenkins J, Mayfield E, Pascale A, Zhang C, Drossman J, apenergy.2018.03.002, URL: https://www.sciencedirect.com/science/article/pii/
Williams R, Pacala S, Socolow R, Baik E, Birdsey R, Duke R, Jones R, S0306261918303180.
Haley B, Leslie E, Paustian K, Swan A. Net-zero america: Potential pathways, [25] Linvill C, Candelaria J, Elder C. The value of geothermal energy genera-
infrastructure, and impacts. 2020. tion attributes: Aspen report to ormat technologies. Technical Report, Aspen
[14] Williams JH, Jones RA, Haley B, Kwok G, Hargreaves J, Farbes J, Environmental Group; 2013.
Torn MS. Carbon-neutral pathways for the United States. AGU Adv [26] Millstein D, Dobson P, Jeong S. The potential to improve the value of
2021;2(1):e2020AV000284. http://dx.doi.org/10.1029/2020AV000284, U.S. geothermal electricity generation through flexible operations. J Energy
e2020AV0002842020AV000284, arXiv:https://agupubs.onlinelibrary.wiley. Resour Technol 2020;143(1):010905. http://dx.doi.org/10.1115/1.4048981,
com/doi/pdf/10.1029/2020AV000284, URL https://agupubs.onlinelibrary.wiley. arXiv:https://asmedigitalcollection.asme.org/energyresources/article-pdf/143/1/
com/doi/abs/10.1029/2020AV000284. 010905/6657806/jert_143_1_010905.pdf.

17
W. Ricks et al. Applied Energy 313 (2022) 118807

[27] Goyal K. Reservoir Response to Curtailment at the Geysers. In: Proceedings of [42] McClure M, Kang C, Hewson C, Medam S. ResFrac technical writeup. Technical
the 27th workshop on geothermal reservoir engineering. Stanford, CA; 2002. Writeup, Technical Report, Palo Alto, CA: ResFrac Corporation; 2021, Accessed
[28] Green S, McLennan J, Panja P, Kitz K, Allis R, Moore J. Geothermal battery July 27, 2021: https://www.resfrac.com/wp-content/uploads/2021/06/ResFrac-
energy storage. Renew Energy 2021;164:777–90. http://dx.doi.org/10.1016/ Technical-Writeup-February-13-2021.pdf.
j.renene.2020.09.083, URL: https://www.sciencedirect.com/science/article/pii/ [43] Hu D, Zheng Y, Wu Y, Li S, Dai Y. Off-design performance comparison
S0960148120315081. of an organic rankine cycle under different control strategies. Appl En-
[29] McTigue J, Zhu G, Turchi C. Hybridizing a geothermal plant with solar ergy 2015;156:268–79. http://dx.doi.org/10.1016/j.apenergy.2015.07.029, URL:
and thermal energy storage to enhance power generation. Technical Report https://www.sciencedirect.com/science/article/pii/S0306261915008582.
NREL/TP-5500-70862, Golden, CO: National Renewable Energy Laboratory; [44] Caiso oasis. 2021, Accessed: 2021-04-8, http://oasis.caiso.com/mrioasis/logon.
2018. do.
[30] Fleming M, Adams B, Randolph J, Ogland-Hand J, Kuehn T, Buscheck T, [45] Ercot market prices. 2021, Accessed: 2021-03-24, http://www.ercot.com/
Bielicki J, Saar M. High efficiency and large-scale subsurface energy storage with mktinfo/prices.
CO2 . In: Proceedings of the 43rd workshop on geothermal reservoir engineering. [46] Jenkins J, Sepulveda N. Enhanced decision support for a changing electricity
Stanford, CA; 2018. landscape: The GenX configurable electricity resource capacity expansion model.
[31] Buscheck TA, Bielicki JM, Edmunds TA, Hao Y, Sun Y, Randolph JB, Saar MO. Working Paper, Cambridge, MA: MIT Energy Initiative; 2017.
Multifluid geo-energy systems: Using geologic CO2 storage for geothermal energy [47] Lew D, Brinkman G, Ibanez E, Florita A, Heany M, Hodge B-M, Hummon M,
production and grid-scale energy storage in sedimentary basins. Geosphere Stark G, King J, Lefton S, Kumar N, Agan D, Jordan G, Venkataraman S. The
2016;12(3):678–96. http://dx.doi.org/10.1130/GES01207.1, arXiv:https://pubs. western wind and solar integration study phase 2. Technical Report NREL/TP-
geoscienceworld.org/gsa/geosphere/article-pdf/12/3/678/3334744/678.pdf. 5500-55588, Golden, CO: National Renewable Energy Laboratory (NREL);
[32] Ogland-Hand J, Bielicki J, Adams B, Buscheck T, Saar M. Using sedimen- 2013.
tary basin geothermal resources to provide long-duration energy storage. In: [48] Majer EL, Baria R, Stark M, Oates S, Bommer J, Smith B, Asanuma H.
Proceedings of the world geothermal congress 2020. Reykjavik, Iceland; 2021. Induced seismicity associated with enhanced geothermal systems. Geother-
[33] Brown DW, Duchane DV, Heiken G, Hriscu VT. Mining the earth’s heat: Hot dry mics 2007;36(3):185–222. http://dx.doi.org/10.1016/j.geothermics.2007.03.003,
rock geothermal energy. Springer; 2012. URL: https://www.sciencedirect.com/science/article/pii/S0375650507000387.
[34] de Sisternes FJ, Jenkins JD, Botterud A. The value of energy storage in [49] Sepulveda N, Jenkins J, Edington A, Mallapragada DS, Lester R. The design space
decarbonizing the electricity sector. Appl Energy 2016;175:368–79. http://dx. for long-duration energy storage in decarbonized power systems. Nature Energy
doi.org/10.1016/j.apenergy.2016.05.014, URL: https://www.sciencedirect.com/ 2021.
science/article/pii/S0306261916305967. [50] Adams B, Ogland-Hand J, Bielicki JM, Schädle P, Saar M. ChemRxiv. 2021,
[35] Arbabzadeh M, Sioshansi R, Johnson J, Keoleian G. The role of energy storage http://dx.doi.org/10.26434/chemrxiv.13514440.v1, preprint.
in deep decarbonization of electricity production. Nature Energy 2019. [51] Mines G. GETEM User manual. Idaho National Laboratories INL/EXT-16-
[36] Mallapragada DS, Sepulveda NA, Jenkins JD. Long-run system value of battery 38751, Idaho Falls, ID: Idaho National Laboratories; 2016, Accessed May
energy storage in future grids with increasing wind and solar generation. Appl 5, 2021: https://workingincaes.inl.gov/SiteAssets/CAES%20Files/FORGE/inl_ext-
Energy 2020;275:115390. http://dx.doi.org/10.1016/j.apenergy.2020.115390, 16-38751%20GETEM%20User%20Manual%20Final.pdf.
URL: https://www.sciencedirect.com/science/article/pii/S0306261920309028. [52] Akar S, Augustine C, Kurup P, Mann M. Global value chain and manufacturing
[37] Latimer T, Meier P. Use of the experience curve to understand economics for at- analysis on geothermal power plant turbines. In: Presented at the 41st geothermal
scale EGS projects. In: Proceedings of the 42nd workshop on geothermal reservoir resource council annual meeting. Salt Lake City, UT; 2017.
engineering. Stanford, CA; 2017. [53] NREL (National Renewable Energy Laboratory). 2020 annual technology base-
[38] Gringarten A, Witherspoon P, Ohnishi Y. Theory of heat extraction from fractured line. Technical Report, Golden, CO: National Renewable Energy Laboratory;
hot dry rock. J Geophys Res 1975;80(8):1120–4. 2020.
[39] Olson J, Augustine C, Eustes A, Fleckenstein W. Design considerations for apply- [54] Barker B, Pingol A. Geysers reservoir performance - an update. In: Proceedings
ing multi-zonal isolation techniques in horizontal wells in a geothermal setting. of the twenty-second workshop on geothermal reservoir engineering. Stanford,
In: Proceedings of the 40th workshop on geothermal reservoir engineering. CA; 1997.
Stanford, CA; 2015. [55] Freeze R, Cherry J. Groundwater. Englewood Cliffs, N.J. U.S.A.: Prentice-Hall;
[40] Li T, Shiozawa S, McClure MW. Thermal breakthrough calculations to 1979.
optimize design of a multiple-stage enhanced geothermal system. Geothermics [56] Reilly T, Franke O, Bennett G. The principle of superposition and its application
2016;64:455–65. http://dx.doi.org/10.1016/j.geothermics.2016.06.015, URL: in ground-water hydraulics. Technical Report OFR 84-459, Albuquerque, NM:
https://www.sciencedirect.com/science/article/pii/S0375650516300712. U.S. Geological Survey; 1984.
[41] McClure M, Kang C. A three-dimensional reservoir, wellbore, and hydraulic [57] Adams BM, Schädle P. Gengeo: coupled reservoir-electricity-cost geothermal
fracturing simulator that is compositional and thermal, tracks proppant and simulator. 2020, http://dx.doi.org/10.5281/zenodo.4383139.
water solute transport, includes non-Darcy and non-Newtonian flow, and handles [58] Gorman W, Mills A, Wiser R. Improving estimates of transmission capital costs for
fracture closure. In: Presented at the SPE reservoir simulation conference. utility-scale wind and solar projects to inform renewable energy policy. Energy
Montgomery, TX; 2017. Policy 2019;135:110994. http://dx.doi.org/10.1016/j.enpol.2019.110994, URL:
https://www.sciencedirect.com/science/article/pii/S0301421519305816.

18