1

THE ENERGY ISLAND AN INVERSE PUMP ACCUMULATION STATION

Authors
W.W. de Boer, KEMA Consulting
F.J. Verheij, KEMA Consulting
D. Zwemmer, Lievense BV
R. Das, Gebr Das
Contact person: F.J. Verheij, KEMA Consulting , P.O. Box 9035, 6800 ET Arnhem, the Netherlands.
Tel.+31263562445, Fax.+31263515456, e-mail: [email protected]

Summary
In this paper we describe an innovative concept for an energy storage system consisting of an inverse offshore pump
accumulation station (IOPAC). To store energy, water is pumped out from a reservoir, and to deliver energy, water is
flowing into the reservoir passing through water turbines. The IOPAC will be stationed on an artificial island at sea,
called Energy Island, consisting of a ring of dikes enclosing a 50 meters deep dredged reservoir. The proposed energy
storage system will have a maximum generation capacity between 2.000 and 2.500 MW dependent on the water
level. The store capacity will be 30 GWh. These design criteria does originate from the capability of the system to
handle (most of) the imbalance due to wind energy forecast errors as well as to download wind energy at night.

Introduction
Energy storage can offer several attractive features for electricity production, especially in combination with wind
energy. It can compensate unexpected wind power variations and/or store energy during high wind periods at night in
order to prevent wasting wind power or conventional units being shut down. Also it can be applied for download
conventional power during low energy prices and extract power during peak hour prices.
The storage system of interest in this paper is an inverse offshore pump accumulation station (IOPAC) that seems to
be capable for these hard tasks. This paper will describe the technical concept first. Thereafter background will be
given for the chosen preliminary design specifications of the storage system followed by the conclusions.

Technical concept
The IOPAC will be stationed on an artificial island at sea, called Energy Island, consisting of a ring of dikes
enclosing a 50 meters deep dredged reservoir. In figure below an artist impression of the concept is given.

Figure 1 Artists impression of the inverse off shore pump accumulation station
Rudolf
D A S
Rudolf
D A S
Rudolf
D A S
Rudolf
D A S
2
The size of the internal reservoir will be 10 x 6 km holding a capacity of 30 GWh. The island can be build from the
sand, which will be obtained from deepening the reservoir. In Figure 2 a cross section of the concept is given.

Figure 2 Cross section of the concept through the turbine

The intended depth of the reservoir will be 50 m below sea level. The water level of the reservoir will vary between
30 and 40 m. The pump and turbine capability are combined in one pump/turbine system for which a Francis type
seems to be most suited. The preliminary design, which will be motivated in next section, does have a maximum
pump/turbine power rate of 2.500 MW, existing of 16 modules with a maximum of ~ 160 MW. The efficiency of the
pumps and turbines near maximum load is around 90%, resulting in a total efficiency of 81%.

One great uncertainty in the concept, is seawater leakage into the reservoir. For the coast of the Netherlands, thick
layers of clay were found, on a right depth (50 m below sea level) and with a good thickness (40 m). This clay layer
should be able to resist the ground water pressure at -90 m below sea level. From aside bentonite walls should
prevent seawater leakage. Current practice shows bentonite walls till a depth of 60 m.

On the island additional power systems can be build such as wind turbines and safety critical installations (eg LNG).
The reservoir on the right of figure 1 is a tidal central, but is only attractive when the difference between high and
low tide is strong. Last but not least, it can offer coast protection, a barrier against increasing water level due to
climate change.

In this paper we will further be focused on the size of the storage system.

Optimal size storage system
The following applications for storage systems are attractive:
1. Regulating power for Program Responsible parties and/or TSO to compensate wind power forecast errors
2. Download capability during high wind periods at night
3. Download capacity during off peak for conventional power plants combined with additional generating
capacity during peak load
4. Additional generating capacity during cooling water limitations
5. Primary control.
In this paper we will show our approach for deriving an optimal sized storage system for the first 2 (most important)
applications.

Required storage size for wind power forecast errors
We have analyzed the required storage size for wind power forecast errors for 2 scenarios of nominal installed wind
power in the Netherlands:
2015: 4.600 MW: 2.200 MW off shore + 2.400 MW on shore
Rudolf
D A S
Rudolf
D A S
3
2020: 9.000 MW: 6.000 MW off shore + 3.000 MW on shore

The last scenario is a goal for the Dutch government to have installed in 2020 or at least in 2030.

We have analyzed the size of the forecast faults based on 24 hours ahead and 6 hours ahead forecast. For the
imbalance due to the 24 hours ahead forecast we applied real wind speed measurements and forecast made with a
Hirlam method. It is the same data set as applied in [1] and [2].

For dimensioning the storage system the following specifications are required for maximum:
- Amount of power consuming / delivering [MW]
- Rate of change capability [MW/min]
- Energy capturing capability [GWh].

Based on the analyses of the yearly data for wind power realization / forecast, we have found the following values.
We have showed them as a function of MW, GWh, MW/min. The values are depicted with their confidence intervals
of 90% and 98%.

Figure 3 Summary required size of storage system as a function of the 90% en 98% confidence interval based
on forecast errors (forecast 24 hours ahead)
The MW values are easy to derive; just observing the tailor ends of the cumulative distribution of the yearly
imbalance data. Similar the rate of change values. The GWh data gives us information of the persistency of the
imbalance, but are more difficult to derive. It was shown that the integrated value of the imbalance is not symmetric,
meaning that there is more shortage then surplus. Therefore the system can be saturated soon, which disables its
functionality. Also if the integrated imbalance holds more then 8 hours, we can interfere. We did apply the following
strategy. We fed the imbalance as input power for the storage system and did operate the system in such a way that it
was kept in a capable state of action, meaning around half of its capacity by means of selling or buying power on
intra day market. So we did combine the imbalance reduction with market selling/buying actions. At peak hours we
did sell power if a surplus was noticed and we did buy power during off peak hours if a shortage was noticed. This
seems to be feasible, because a surplus or shortage of energy available in the reservoir can be noticed in advance by
simply monitoring the water level of the reservoir. Based this strategy we found GWh sizes as depicted in Figure 3.

The depicted MW- GWh hours are based on the imbalance due to 24 hours ahead forecast. An 24 hours ahead
forecast can be used for nominating wind power in the E(nergy) Program. The E-program is made by a Program
Responsible party to nominate its power to be supplied / consumed and is submitted to the TSO; time period 24 hrs
ahead (more precise 12-36 hrs ahead). Coming closer to the moment now, the Program Responsible party will most
probably refine its nomination by means of more accurate forecast and intra day trade actions. We have roughly
MW
GWh
0 500 1.000 1.500 2.000 2.500 3.000 3.500
80
60
40
20
Imbalance
w.e. 2015
c.i. 90%
9 MW/min
Imbalance w.e.
2020, c.i. 98%
38 MW/min
Imbalance w.e.
2020 ,c.i. 90%
19 MW/min
Imbalance w.e.
2015, c.i. 98%
18 MW/min
Preliminary design
IOPAC
c.i. confidence interval
w.e. wind energy
Size imbalance due to forecast errors (24 hours ahead)
MW
GWh
0 500 1.000 1.500 2.000 2.500 3.000 3.500
80
60
40
20
Imbalance
w.e. 2015
c.i. 90%
9 MW/min
Imbalance w.e.
2020, c.i. 98%
38 MW/min
Imbalance w.e.
2020 ,c.i. 90%
19 MW/min
Imbalance w.e.
2015, c.i. 98%
18 MW/min
Preliminary design
IOPAC
c.i. confidence interval
w.e. wind energy
Size imbalance due to forecast errors (24 hours ahead)
4
assessed the impact of these kind of actions. This was done based on [3]: taken the difference in forecast accuracy
between 24 hours ahead and 6 hours ahead. From figure 5 of [3], it can be derived that the forecast error (normalized
RMSE) is reduced from 14% to 8%. As a rough assumption we did multiply our imbalance values with a factor 0,57
(8/14), given the results showed in figure 4. These values are within the domain of our preliminary design of the
storage system.

Figure 4 Summary required size of storage system as a function of the 99,4% en 98% confidence interval
based on forecast errors; forecast 6 hours ahead

Required size for download during high wind periods at night
Another welcome energy storage feature is to download wind power during high wind periods at night. At night, at
low load periods, the conventional power units are lowered. If the wind power generation at those periods is high, the
conventional units have to be lowered more. This can be done till a certain amount before shutting down the units.
Shutting down units (and afterward starting up again) is not economic and should be limited as far as possible. We
have analyzed, that between 2.000 MW 4.000 MW of wind power can be absorbed in the Dutch grid at off peak
hours, without additional shutting down power plants. We will further apply an average value of 3.000 MW. Thus, if
wind power exceeds 3.000 MW, wind power must be limited, or being absorbed in an energy storage system.

In figure below we do show the yearly wind power production for the 2020 scenario, based on our measurement data
set. We have given the night-production above 3.000 MW marked with red and
*
.
MW
GWh
0 500 1.000 1.500 2.000 2.500 3.000 3.500
80
60
40
20
Imbalance w.e.
2015, c.i. 90%
Imbalance w.e.
2020, c.i. 98%
Imbalance w.e.
2020, c.i. 90%
Imbalance w.e.
2015
c.i. 98%
24 6 hr ahead
24 6 hr ahead
Preliminary design
IOPAC
c.i. confidence interval
w.e. wind energy
MW
GWh
0 500 1.000 1.500 2.000 2.500 3.000 3.500
80
60
40
20
Imbalance w.e.
2015, c.i. 90%
Imbalance w.e.
2020, c.i. 98%
Imbalance w.e.
2020, c.i. 90%
Imbalance w.e.
2015
c.i. 98%
24 6 hr ahead
24 6 hr ahead
Preliminary design
IOPAC
c.i. confidence interval
w.e. wind energy
5
Figure 5 Yearly power production of a 6.000 MW off shore and 3.000 MW on shore scenario for the
Netherlands.

To dimension the storage system, 6.000 MW could be a maximum value. The energy storage of the preliminary
designed (2.000 MW and 2.500 MW) can absorb 2,37E6 MWh (integrations of power within 2.250 MW marge).
This can be sold during peak hours, taking into account an efficiency of 81% remains 1,92E6 MWh. We assume that
this can be sold for 50/MWh giving a revenue of 95 M. If we can absorb more the 2.250 MW, the revenues are
higher but will not justify the additional E-connection costs. This trade off and other benefits analysis such as CO2
reduction are still part of a current study.

Conclusions
In this paper we introduced our Energy Island, an innovative plan for a large energy storage system. Still part of a
current study, we are analyzing the benefits in respect to the costs of additional regulating reserve, download (wind)
power, CO2 reduction and coast protection. The first results show that this concept is technical and economical
feasible.

References
1. M. Gibescu (Delft University of Technology, the Netherlands), A.J. Brand (ECN, the Netherlands), Estimation of
System Balancing Requirements due to the Integration of Large-Scale Wind Energy. Proceedings Workshop Sixth
International Workshop on Large Scale Integration of Wind Power and Transmission Networks for Off shore
Windfarms, Delft, 26-28 October 2006
2. The value of wind power from a market perspective
C. Kleinschmidt, F. Verheij, H. Cleijne, N. Moldovan (KEMA, the Netherlands). The value of wind power from a
market perspective. Proceedings Workshop Sixth International Workshop on Large Scale Integration of Wind
Power and Transmission Networks for Off shore Windfarms, Delft, 26-28 October 2006
3. L. von Bremen, J. Tambke, N. Saleck, D. Heinemann (University of Oldenburg, Germany). Integration of 18 GW
Wind Energy into the Energy Market Practical. Confidence in Large-scale offshore wind farming: Wind Power
Predictability and stable Grid Integration of 25 GW German Wind Power. Proceedings Workshop Sixth
International Workshop on Large Scale Integration of Wind Power and Transmission Networks for Off shore
Windfarms, Delft, 26-28 October 2006
0 50 100 150 200 250 300 350 400
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
P real
P real daluren
Wind Power production for one year
> 3.000 MW
during off peak
(red)
2.250 MW
days
MW
0 50 100 150 200 250 300 350 400
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
P real
P real daluren
Wind Power production for one year
> 3.000 MW
during off peak
(red)
2.250 MW
days
MW