Solar Energy: Mansour Alramlawi, Aouss Gabash, Erfan Mohagheghi, Pu Li
Solar Energy: Mansour Alramlawi, Aouss Gabash, Erfan Mohagheghi, Pu Li
Solar Energy: Mansour Alramlawi, Aouss Gabash, Erfan Mohagheghi, Pu Li
Solar Energy
journal homepage: www.elsevier.com/locate/solener
A R T I C L E I N F O A B S T R A C T
Keywords: In many developing countries where the energy supply capacity is less than the demand, energy companies
Photovoltaic hybrid system distribute the available power according to a specific time schedule, leading to grid scheduled blackouts (GSBs)
Battery lifetime cost for the costumers. A suitable way to solve this problem for residential loads is to install a combined photovoltaic
Economic model predictive control (PV) and battery storage system as a replacement of the grid during the blackout periods. However, it is well-
Grid blackouts
known that the lifetime of battery storage will be considerably shortened, if it is not operated appropriately. This
paper explores the potential benefits of applying economic model predictive control (EMPC) to optimize the
operation of a hybrid PV-battery system to address the GSB problem. The aim of our control strategy is to cover
the load as much as possible and meanwhile minimize the total cost of the energy consumed from the grid. In
particular, the battery is operated in such a way that its lifetime will be prolonged. A comparison between
different operation strategies for the system considering the seasonal variation in the load and the PV power
generation is made. The results show that the proposed approach leads to a significant reduction of the total
energy consumed from the grid while decreasing the curtailment of the generated power from the PV-array, and
maximizing the battery lifetime.
⁎
Corresponding author.
E-mail addresses: [email protected] (M. Alramlawi), [email protected] (A. Gabash), [email protected] (E. Mohagheghi),
[email protected] (P. Li).
https://doi.org/10.1016/j.solener.2017.12.022
Received 20 September 2017; Received in revised form 24 November 2017; Accepted 11 December 2017
0038-092X/ © 2017 Elsevier Ltd. All rights reserved.
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
Nomenclature Pav . l (t ) total available power from PV-battery system and grid to
cover connected loads at time t
αg (t ) grid status at time t Pav . pv (t ) available power from PV-array at time t
ηch battery bank charging efficiency Pch . g (t ) battery bank charging power from grid at time t
ηdis battery bank discharging efficiency Pch . pv (t ) battery bank charging power from PV-array at time t
γch parameter for battery bank charging current limit Pch (t ) total battery bank charging power at time t
λsoc (t ) effective ampere-hour weighting factor at time t Pdisp . g (t ) total dispatched power from grid at time t
τ grid-ON period in hour Pdisp . pv (t ) total dispatched power from PV-array at time t
Ac (t ) effective consumed ampere-hour from battery bank at Pdis (t ) battery bank discharging power at time t
time t Pg . max maximum allowed power to be extracted from grid
Ac′ (t ) actual consumed ampere-hour from battery bank at time t Pl . g (t ) power to load from grid at time t
Atotal total effective capacity of the lead-acid battery Pl . pv (t ) power to load from PV-array at time t
Cb . init initial cost of battery bank in $/kW h Pmax . c . stc maximum output power of PV-cell under standard test
Cb . l (t ) cost of battery bank lifetime loss at time t in $/kW h conditions
Ce . g cost of energy dispatch from the grid in $/kW h Pmax . c (t ) maximum output power of PV-cell at time t
DOD battery bank depth of discharge Pmax . m . stc maximum output power of PV-module under standard test
e charge of electron, i.e., 1.602 × 10−19 conditions
F1 total cost of energy dispatch from the grid in $ Pmax . pv (t ) maximum output power of PV-array at time t
F2 total cost of the battery bank lifetime loss in $ Preq . l (t ) required load power at time t
FF (t ) fill factor of PV-cell at time t
Qb . n battery bank nominal capacity in ampere-hour
FF0 (t ) nominal fill factor of PV-cell at time t Rs series resistance of PV-cell
FF0. stc nominal fill factor of PV-cell at standard test conditions
rs (t ) normalized series resistance of PV-cell at time t
FFstc Fill factor of PV-cell at standard test conditions s1,… ,5 binary variables for controllable switches’ status
SOC (t ) battery bank state of charge at time t
G (t ) global solar irradiance at time t
Isc . c . stc short circuit current of PV-cell under standard test condi- SOCmax maximum value of SOC
tions SOCmin minimum value of SOC
Isc . c (t ) short circuit current of PV-cell at time t SOCnorm (t ) normalized battery bank SOC at time t
Isc . m . stc short circuit current of PV-module under standard test T grid scheduled blackouts cycle in hour
conditions Ta (t ) ambient temperature in Celsius at time t
J total operating cost of PV-battery system in $
Tc (t ) PV-cell temperature in Celsius at time t
k EMPC sampling time
Ta . k (t ) ambient temperature in Kelvin at time t
kB Boltzmann’s constant, 1.38 × 10−23 Vb . n battery bank nominal voltage
Ki short circuit current temperature coefficient of PV-cell
Voc . c . stc open circuit voltage of PV-cell under standard test condi-
Kv open circuit voltage temperature coefficient of PV-cell tions
Lf (t ) battery bank lifetime losing factor at time t
Voc . c (t ) open circuit voltage of PV-cell at time t
N total simulation time in hour Voc . m . stc open circuit voltage of PV-module under standard test
n diode ideality factor conditions
Np EMPC prediction horizon length in hour
voc (t ) normalized open circuit voltage of PV-cell at time t
Nc . m number of PV-cells in PV-module Vt PV-cell thermal voltage at time t
Np . m number of parallel PV-modules in PV-array Ub(t ) sequence of Np predicted binary control variables
ub(t ) binary control variables vector at time t
Ns . m number of series PV-modules in PV-array
Uc(t ) sequence of Np predicted continues control variables
NOCT nominal operating PV-cell temperature
uc(t ) continuous control variables vector at time t
Pav . g (t ) available power from grid at time t
costs, by developing a power management system. Moreover, a master- (SOC) can considerably reduce the battery lifetime of lead-acid bat-
slave control strategy was proposed in Hijjo et al. (2016) to reduce the teries (Kaiser, 2007). A simple solution for this problem is increasing
fuel consumption of a PV-diesel-battery system for non-residential loads the battery size, and meanwhile, increasing the charging power to keep
to deal with grid scheduled blackouts (GSBs), i.e., power supply from the SOC level as high as possible. This solution could increase the in-
the grid is switched off daily for a certain period of time. vestment and operation cost of the system. Therefore, many studies
In Khoury et al. (2015), it was shown that an optimal design of a were made to extend the battery lifetime by optimizing the system
hybrid PV-battery system could make it more beneficial than a diesel- operation.
based system for residential loads in the areas suffering from GSBs. In As an example, an optimal operation strategy for a stand-alone
another work (Khoury et al., 2016), the authors proposed a real-time wind-solar-diesel-battery system was proposed in Zhao et al. (2013) and
demand side management approach for a PV-battery system to control Dufo-López et al. (2017) to minimize the fuel consumption of the diesel
the operational time of the loads. The work in Pillai et al. (2016) generator and meanwhile maximize the battery lifetime. An optimal
showed the techno-economic feasibility of using a hybrid PV-battery control method for a stand-alone wind-PV-battery system considering
system as a backup power supply to meet residential loads, considering battery lifetime was developed in Wu et al. (2015) to reduce the system
the seasonal variation of load profiles. It is worth to mention that the operation cost. In Teleke et al. (2010), a rule-based control strategy was
above reviewed studies (Mitra, 2010; Khoury et al., 2016, 2015; introduced to extend the battery lifetime in a grid connected PV-battery
Hooshmand et al., 2015; Hijjo et al., 2016; Pillai et al., 2016) did not system. In addition, the battery lifetime was considered in Yang et al.
consider the battery lifetime. (2008) and Abbes et al. (2014) to determine the optimal size of a PV-
However, it was proved that the battery lifetime has a strong in- wind-battery system.
fluence on the total operation cost of the PV hybrid systems (Jossen In this study, the dynamic behavior of the battery is considered.
et al., 2004). For instance, operating the battery at a low state of charge Therefore, the operation of the PV-battery for multiple time steps
126
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
should be taken into account to optimize the system operation. Model In this paper, the hybrid PV-battery system model is significantly im-
predictive control (MPC) (Maciejowski, 2002; Camacho and Alba, proved to include more real operational constraints of the system
2013) has been widely used for optimal operation of renewable energy components, a PV-module model, and the cost of the battery lifetime. In
hybrid systems because it has the ability to explicitly handle system addition, an economic model predictive controller is developed to
dynamics, data forecasting, and operation constraints. An economic minimize the total costs of the dispatched energy from the grid as well
model predictive controller (EMPC) differs from the standard MPC in as the cost of battery lifetime loss. This is achieved by optimizing the
that the controller provides its actions to improve the economic per- power flow in the system using a receding horizon strategy.
formance of the process, rather than tracking a setpoint (Angeli et al., From the discussions above, it can be clearly seen that many studies
2012; Rawlings et al., 2012). The potential of using EMPC in power have been carried out on the optimal operation of the on-grid and off-
system management was investigated in Hovgaard et al. (2010) to op- grid PV-hybrid systems which aim to decrease the operating costs of the
timize the operation of a controllable power generators for minimizing system, the fuel consumption of the diesel generator, and the curtail-
the costs of the generated power. In Tedesco et al. (2015) an EMPC was ment of the PV power generation. However, most of the previous stu-
used to carry out a cost effective supervision control strategy for de- dies were focused on the intermittent behavior of the renewable energy
creasing the operational costs of a grid connected community micro- sources and/or the loads as the main problem to be solved. But the
grid. problem of the discontinuous power supply from the grid is rarely
Furthermore, in Zhu et al. (2015) an hourly switched MPC strategy considered, although it is still a common issue in many developing
was developed for optimal energy dispatch for an isolated PV-diesel- countries in the world.
battery hybrid system, to minimize the fuel consumption of the diesel In this paper, we presents a new approach to optimizing the op-
generator and at the same time to prevent an excessive use of the eration of a hybrid PV-battery system considering the GSB problem.
battery bank. In Xu and Singh (2011), a hybrid control strategy com- This work is a significant extension of our study in Alramlawi et al.
bining both standby backup and MPC was proposed to enhance the (2017) by optimizing the system operation considering the cost of
performance of an electric storage system so as to decrease the blackout battery lifetime. In comparison to Khoury et al. (2015), Wu and Xia
periods in electrical distribution networks. MPC was also used in (2015), and Alramlawi et al. (2017) and the above studies, the major
Núñez-Reyes et al. (2017) to provide an optimal scheduling for the contributions of this study include:
operation of a PV-battery system connected to a grid which maximize
the economic benefit of the PV plant considering the electricity market (1) A comprehensive model for a residential PV-battery system is de-
prices. An optimal dispatch strategy of a PV-wind-battery-diesel system veloped to address the GSB problem. In this model real operational
was developed in Tazvinga et al. (2014) to minimize the system op- constraints of the system components and the costs of the battery
eration costs. The work in Zhang et al. (2015) also used MPC for lifetime are taken into account.
minimizing the overall operation costs by scheduling a residential mi- (2) A new EMPC-based operation strategy is proposed to minimize the
crogrid. total operation costs of the PV-battery system considering the bat-
A special property of a residential hybrid PV-battery system to tery lifetime and the GSB problem.
campaign scheduled blackouts is that there are switches of the power (3) A comparison between different operation scenarios is made to
flow to different components during the operation. This leads to diffi- verify the advantages of the proposed strategy and to illustrate the
culties in decision-making for design and operation of such systems. impacts of considering the costs of the battery lifetime.
Recently, a grid connected residential PV-battery system was studied in
Wu and Xia (2015) for developing a daily schedule of optimal switching The rest of the paper is organized as follows. In Section 2, the model
to minimize the consumed energy costs. The proposed model in Wu and of the hybrid PV-battery system is presented. Section 3 describes the
Xia (2015) was recently modified in Alramlawi et al. (2017) to consider optimization problem and the proposed EMPC. The results of a case
GSB problem that was investigated in Khoury et al. (2015). In our initial study are presented in Section 4. The paper is concluded in Section 5.
work (Alramlawi et al., 2017), six operation modes were proposed.
Each of these operation modes gives the available energy sources dif- 2. Modelling of the hybrid PV-battery system
ferent priority levels to cover the load and to charge the battery bank.
In addition, an optimization problem was formulated to select one fixed The structure of the PV-battery system considered in this paper is
operation mode for each season in the year. However, the study in shown in Fig. 1. The system consists of a PV-array, a battery bank, a
Alramlawi et al. (2017) did not optimize the power flow between grid, a load, an AC charger to charge the battery bank from the grid, a
system components and the cost of battery lifetime was not considered. DC charger to charge the battery bank from the PV-array, a DC/AC
127
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
inverter to supply AC power to the load from the battery bank and from Zhang et al., 2017). According to Lorenzo (1994) the Rs value can be
the PV-array and 5 controllable switches (SW1–SW5). It is to note that calculated using the available data in the PV-module datasheet (Yingli
the following real operational constraints of the system components are Solar, 2017), by
considered:
FFstc ⎞ Voc . c . stc
Rs = ⎛1− ⎜ ⎟
• The changing in the battery state of charge should be within the ⎝ FF0. stc ⎠ Isc . c . stc (10)
specified upper and lower limits. Moreover, the maximum charging here FFstc and FF0. stc will be calculated by Eqs. (5) and (6) and Eqs. (8)
power should be limited and compatible with the battery bank size. and (9) using Pmax . c . stc, Isc . c . stc, Voc . c . stc and Ta . k . stc values of the PV-cell.
Here, it should be noted that the battery bank cannot charge and Note that the available data normally in the datasheet are for the PV-
discharge at the same time. In addition, the battery bank cannot module and not for the PV-cells which are connected in series inside the
charge from different chargers simultaneously. PV-module. Therefore,
• The load cannot be covered from different AC sources if they are not
Pmax . m . stc
synchronized. In addition, the system cannot dispatch more than the Pmax . c . stc =
available power from the PV-array or allowed from the grid. Nc . m (11)
Furthermore, the PV-battery system output power should not exceed Voc . m . stc
the load demand. Voc . c . stc =
Nc . m (12)
The mathematical model of the major components of the system and Isc . c . stc = Isc . m . stc. (13)
its operation constraints are modeled below in detail.
Based on the maximum output of the PV-cell, the output of the
whole PV-array can be calculated as follows
2.1. PV-array model
Pav . pv (t ) = Ns . m Np . m Nc . m Pmax . c (t ) (14)
The mathematical model of a single diode PV-cell with a series re-
sistance (Lorenzo, 1994; Park and Choi, 2015) is used in this paper to here Ns . m and Np . m are the number of PV-modules connected in series
calculate the output power from each cell according to the maximum and parallel, respectively.
power point algorithm. The maximum power generated by the PV-cell
is expressed as (Lorenzo, 1994): 2.2. Battery bank
Pmax . c (t ) = Voc . c (t ) Isc . c (t ) FF (t ) (1) The battery bank is used to guarantee a continuous power supply to
where the values of both of Voc . c (t ) and Isc . c (t ) depend on the solar ir- the load during the blackout periods or at low solar irradiance time. The
radiance and the temperature of the PV-cell. The instantaneous value of performance of the battery bank is represented by the state of the
Voc . c (t ) and Isc . c (t ) can be calculated, respectively, as follows: charge SOC (t ) value at each time instant. The SOC (t ) value increases
due to charging the battery bank from the grid or from the PV-array,
Voc . c (t ) = Voc . c . stc + Kv (Tc (t )−25) (2)
and decreases due to discharging for covering the load demand. In this
G (t ) study, the charging and discharging processes is controlled by the
Isc . c (t ) = (Isc . c . stc + Ki (Tc (t )−25)) proposed EMPC through controllable switches, see Fig. 1. The change of
1000 (3)
SOC of the battery bank is expressed by
where
Pdis (t ) s5 (t )
(NOCT −20) SOC (t + Δt ) = SOC (t ) + ηch Pch (t )Δt − Δt
Tc (t ) = Ta (t ) + G (t ). ηdis (15)
800 (4)
where Pch (t ) is given by
The FF (t ) is the ratio of the maximum power that can be produced
by the PV-cell to the product of the open-circuit voltage and short- Pch (t ) = Pch . g (t ) s1 (t ) + Pch . pv (t ) s2 (t ). (16)
circuit current of the PV-cell. The formula from Lorenzo (1994) is used
to calculate FF (t ) : To prevent overcharging or deep discharging of the battery bank,
the SOC should be limited according to the capability of the battery
Pmax . c (t ) bank (Gabash and Li, 2012), i.e.,
FF (t ) = = FF0 (t )(1−rs (t ))
Voc . c (t ) Isc . c (t ) (5)
SOCmin ⩽ SOC (t ) ⩽ SOCmax (17)
where
while, SOCmin is related to the depth of charge of the battery bank by
v (t )−ln (voc (t ) + 0.72)
FF0 (t ) = oc SOCmin = (1−DOD) SOCmax (18)
voc (t ) + 1 (6)
where rs (t ) and voc (t ) are the normalized series resistance and open where
circuit voltage for the PV-cell, respectively, which are calculated by SOCmax = Vb . n Qb . n. (19)
Isc . c (t ) To avoid battery overheating, the maximum charging current
rs (t ) = Rs
Voc . c (t ) (7) should be limited (Koutroulis and Kalaitzakis, 2004). According to
Alramlawi et al. (2017), the charging power should be constrained by
Voc . c (t )
voc (t ) = Pch ⩽ γch Vb . n Qb . n
Vt (t ) (8) (20)
where where γch is a parameter for the limit of the charging current, which is
typically between 0.1 and 0.2 to enable an optimum charge. The
nkB Ta . k (t )
Vt (t ) = charging and discharging process of the battery bank cannot take place
e (9)
simultaneously. In addition, the battery bank cannot be charged by
The Rs value depends on the material of the PV-cell and is affected different chargers (i.e., grid charger or PV-array charger) at the same
by the PV-cell operating conditions. However, this effect could be in- time (Wu and Xia, 2015). To satisfy these conditions the following
significant and negligible (De Soto et al., 2006; Rahman et al., 2014; constraint should be held
128
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
s1 (t ) + s2 (t ) + s5 (t ) ⩽ 1. (21) from GSBs, i.e., the available power from the grid has a predefined
periodic scheme. Hence, the Grid-ON and Grid-OFF periods are speci-
Based on Eq. (21), only one of the controllable switches (SW1, SW2
fied from the power distribution company and known for the user, as
and SW5) can be ON, i.e., the battery bank will not be charged from the
shown in Fig. 3. The available power from the grid can be described as
grid or PV-array and discharged simultaneously. In addition, the bat-
tery bank will not be charged from the grid and PV-array at the same Pav . g (t ) = αg (t ) Pg . max (28)
time.
Due to the synchronization issue of AC power, the DC / AC inverter where αg (t ) represents the status of the grid, i.e., when αg (t ) = 1 the
and the grid cannot supply power to the load simultaneously. Thus, grid power is available and when αg (t ) = 0 , the gird power is un-
available. The behavior of αg (t ) leads to a periodic rectangular pulse
⎧ s3 (t ) + s4 (t ) ⩽ 1 wave, as shown in Fig. 3. Here, the rectangular pulse wave cycle T is the
⎩ s5 (t ) + s4 (t ) ⩽ 1.
⎨ (22) length of the GSB cycle and τ is the Grid-ON period.
129
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
Įg
Table 2
IJ Type, size and capability of the system components.
1
Component Size/Capability
ON OFF ON
PV-module (250 Wp) × 12 = 3 kWp
DC/AC Inverter 3 kW
0 t
AC charger 2.88 kW
DC charger 2.88 kW
T Battery bank (12 V–200 Ah) × 4 = 9.6 kW h
Fig. 3. The illustration of the rectangular pulse wave representing the grid status.
Table 3
The battery bank parameters.
Parameter DOD (–) ηch (%) ηdis (%) Vb . n (V) Qb . n (Ah) γch (–)
Table 1
Pav . l (t ) = Pl . g (t ) s4 (t ) + Pl . pv (t ) s3 (t ) + Pdis (t ) s5 (t ). (42)
Options of the GA tool in MATLAB.
Pdisp . pv (t ) = Pch . pv (t ) s2 (t ) + Pl . pv (t ) s3 (t ) (43)
Option Value
Pdisp . g (t ) = Pch . g (t ) s1 (t ) + Pl . g (t ) s4 (t ). (44)
Population size 1500
Elite count 100 The proposed EMPC is illustrated schematically in Fig. 4. At each
Crossover function Intermediate time step the EMPC reads the current value of the SOC (t ) and the
Crossover fraction 0.8 forecasted load, grid power and PV-array power. Then, it determines
Mutation function Gaussian
the optimal values for control variables u c (t ) and ub (t ) over the pre-
Selection function Stochastic uniform
Creation function Uniform diction horizon Np by solving the above defined optimization problem
Scaling function Rank based using the EMPC optimizer, see Fig. 4. The control variables computed
Function tolerance 1e−3 for the first time step will be applied to the system. After that, the
Constraint tolerance 1e−4
predictive horizon moves to the next time step and the whole procedure
Use parallel true
is repeated. Note that in this paper we assumed that we have a perfect
forecasting of the values of the solar power generation, the demand and
the available power from the grid in the predictive horizon.
The optimization problem formulated in Eqs. (29)–(39) leads to a
mixed-integer nonlinear programming (MINLP ) problem consisting of
130
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
Table 4
The PV-module parameters.
Parameter Pmax . m . stc (W) Voc . m . stc (V) Ioc . m . stc (A) NOCT °C Kv (%/°C) Ki (%/°C) Nc . m (–) n (–)
1200
Winter Spring Summer Fall
1000
800
G (Wh/m2)
600 (a)
400
200
0
0 24 48 72 96 120 144 168 192
35
Winter Spring Summer Fall
30
25
T a (C)
20
(b)
15
10
5
0
0 24 48 72 96 120 144 168 192
200
P-pv-max (w)
150
(c)
100
50
0
0 24 48 72 96 120 144 168 192
Time (Hour)
Fig. 5. PV-array model illustration. (a) Global solar irradiance at ground level. (b) Ambient temperature. (c) The maximum generated power by the PV-module.
four continuous and five binary control variables which are time-de- available power to cover the load in both Grid-ON and Grid-OFF per-
pendent during the prediction horizon (i.e., 9 × Np in total), as shown iods. Therefore, during the blackout periods, the load will be optimally
Eqs. (40) and (41). In addition, the existence of the nonlinear cost covered by the battery and/or PV-array. The sampling time for the
function and nonlinear constraints leads to a complex MINLP problem. EMPC is set to be 1 h.
Genetic algorithm (GA) is an effective stochastic optimization al-
gorithm developed by Holland (1975) and Goldberg (1989) inspired by
the natural selection process which mimics the biological evolution. GA 4. A case study
is different from traditional optimization algorithms in that they work
with many candidate solutions known as a population rather than just To illustrate the potential of the proposed EMPC to optimize the
one solution. At each step, GA modifies a population of candidate so- operation of the PV-battery system shown in Fig. 1, a case study is
lutions by performing the selection, crossover and mutation processes adapted from Alramlawi et al. (2017) with data components in Table 2.
to produce new children for the next generation (Zilouchian and The parameters of the battery bank are listed in Table 3 while the PV-
Jamshidi, 2000). Over successive generations, the population will be module specifications in Table 4 (Yingli Solar, 2017). The number of
closer to an optimal solution. GA has no restrictions regarding the parallel and series connected modules are taken to be 3 and 4, re-
model of the system or the type of the decision variables (Stojanovski spectively. The price of the lead-acid batteries is set to 213 $/kW h
and Stankovski, 2012). Because of these features, in our study we em- (Khiareddine et al., 2015) which gives Cb . init = 2044.8 $.
ploy GA to solve the MINLP problem described in Eqs. (29)–(39), by To show the effects of seasonal variations in load and PV power
using the GA tool in MATLAB (2016). Meanwhile, parallel computing is profiles, two days from each season of a year are selected. The solar
used to speed up the calculations. The parameters used for the GA tool irradiance and temperature data are taken from Solar energy services
in Matlab are given in Table 1. for professionals (2017) and shown in Fig. 5(a) and (b), respectively.
It is to note that the prediction horizon (Np ) of the economic model According to Alramlawi et al. (2017), we consider τ = 8 and T = 16 to
predictive control is defined to be equal to the length of the grid describe the GSB cycle, see Fig. 3. The maximum allowed power to be
scheduled blackout cycle, i.e., the total length of Grid-On and Grid-Off extracted from the grid is 5 kW and the cost of the consumed power
periods, see Fig. 3. That means the EMPC will optimally utilize the total from the grid is Ce . g = 0.15 $/kW h.
The output of each PV-module is generated using the PV-array
131
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
3 (a)
2
0
0 24 48 72 96 120 144 168 192
2
Winter Spring Summer Fall
Preq-l & Pav-l(kW)
1.5
1 (b)
0.5
0
0 24 48 72 96 120 144 168 192
3
W in te r S p r in g Sum m er F a ll
Pav-pv & Pdisp-pv (kW)
2.5
1.5 (c)
1
0.5
0
0 24 48 72 96 120 144 168 192
2.5
W in te r S p r in g Sum m er F a ll
2
Pch(kW)
1.5
(d)
1
0.5
0
0 24 48 72 96 120 144 168 192
2.5
W in te r S p r in g Sum m er F a ll
2
Pdis(kW)
1.5
(e)
1
0.5
0
0 24 48 72 96 120 144 168 192
12
11 W in te r S p r in g Sum m er F a ll
10
9
SOC (kWh)
8
7
6 (f)
5
4
3
2
1
0
0 24 48 72 96 120 144 168 192
Time (Hour)
Fig. 6. System operation in arbitrary two days for each season for Scenario A. (a) Available power (dashed-blue) and power dispatch (solid-red) from grid. (b) Required load power
(dashed-blue) and available power to cover load (solid-red). (c) Available power (dashed-blue) and power dispatch (solid-red) from PV-array. (d) Charging power of battery bank (solid-
blue) and the charging power limit (dashed-red). (e) Discharging power of battery bank. (f) Battery bank SOC (solid-blue) and SOC upper and lower limits (dashed-red). (For inter-
pretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
132
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
• Scenario A: The backup operation stratgy (Khoury et al., 2015), i.e., dispatched energy from the grid in comparison to Scenario A, as
seen in Table 7. The reason behind this is that the EMPC forces the
the system is used only to cover the load during the blackout per-
system to dispatch power from the PV-array even if the grid is
iods, but without using EMPC.
• Scenario B: The EMPC-based operation strategy but without con-
available during the day hours, as shown in Fig. 7a,b,c.
On the other hand, reducing the dispatched energy from the grid
sidering the cost of the battery lifetime, i.e., F2 in Eq. (29) is not
decreases the total charging power as well. Therefore, the SOC level
included.
• Scenario C: The EMPC-based operation strategy with considering
of the battery bank is lowered, as seen in Fig. 7f.
From Table 7, it can be clearly seen that the cost of the dispatched
the cost of the battery lifetime, i.e., F2 in Eq. (29) is included, which
energy from the grid F1 is decreased and the total cost of the battery
is the proposed strategy.
bank lifetime loss F2 is increased, in comparison to Scenario A. As a
result, the total operation costs of the system J is decreased except in
The computation is carried out on a desktop with Intel Core I7 CPU
winter where the cost of the lifetime loss of the battery bank is
3.4 GHz and 8 GB RAM. The initial values, denoted by (0), chosen for all
(0) (0) (0) (0) (0) highly increased due to the lack of generated power from the PV-
computations are as follows: Pch . g = Pch . pv = Pdis = Pl . pv = s1,… ,5 = 0 and
array, as seen in Fig. 7c (dashed-blue).
SOC (0) = 0.75 × SOCmax .
The comparison results are shown in Tables 5–7 and Figs. 6–8. The • Scenario C: In this scenario, with the same input data as in Scenario
B, but the EMPC is used to optimize the PV-battery system operation
results of each operation scenario are discussed below in detail.
considering the cost of the battery lifetime. This mean that both the
Table 7
Cost of the dispatched energy for each scenario.
Seasons F1("$") F2("$") J ("$") F1("$") F2("$") J ("$") F1("$") F2("$") J ("$")
Winter 3.91 5.02 8.93 3.2 6.42 9.62 3.26 5.1 8.36
Spring 3.3 3.51 6.81 1.84 4.46 6.3 1.96 3.66 5.62
Summer 4.99 5.65 10.64 3.48 5.92 9.4 3.39 5.56 8.95
Fall 3.25 3.51 6.86 1.69 4.42 6.11 1.95 3.53 5.48
133
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
3 (a)
2
0
0 24 48 72 96 120 144 168 192
2
Winter Spring Summer Fall
Preq-l & Pav-l(kW)
1.5
1 (b)
0.5
0
0 24 48 72 96 120 144 168 192
3
Pav-pv & Pdisp-pv (kW)
1.5 (c)
1
0.5
0
0 24 48 72 96 120 144 168 192
2.5
Winter Spring Summer Fall
2
Pch(kW)
1.5
(d)
1
0.5
0
0 24 48 72 96 120 144 168 192
2.5
Winter Spring Summer Fall
2
Pdis(kW)
1.5
(e)
1
0.5
0
0 24 48 72 96 120 144 168 192
12
11 Winter Spring Summer Fall
10
9
SOC (kWh)
8
7
6 (f)
5
4
3
2
1
0
0 24 48 72 96 120 144 168 192
Time (Hour)
Fig. 7. System operation in arbitrary two days for each season for Scenario B. (a) Available power (dashed-blue) and power dispatch (solid-red) from grid. (b) Required load power
(dashed-blue) and available power to cover load (solid-red). (c) Available power (dashed-blue) and power dispatch (solid-red) from PV-array. (d) Charging power of battery bank (solid-
blue) and the charging power limit (dashed-red). (e) Discharging power of battery bank. (f) Battery bank SOC (solid-blue) and SOC upper and lower limits (dashed-red). (For inter-
pretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
134
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
3 (a)
2
0
0 24 48 72 96 120 144 168 192
2
Winter Spring Summer Fall
Preq-l & Pav-l(kW)
1.5
1 (b)
0.5
0
0 24 48 72 96 120 144 168 192
3
Pav-pv & Pdisp-pv (kW)
1.5 (c)
1
0.5
0
0 24 48 72 96 120 144 168 192
2.5
Winter Spring Summer Fall
2
Pch(kW)
1.5
(d)
1
0.5
0
0 24 48 72 96 120 144 168 192
2.5
Winter Spring Summer Fall
2
Pdis(kW)
1.5
(e)
1
0.5
0
0 24 48 72 96 120 144 168 192
12
11 Winter Spring Summer Fall
10
9
SOC (kWh)
8
7
6 (f)
5
4
3
2
1
0
0 24 48 72 96 120 144 168 192
Time (Hour)
Fig. 8. System operation in arbitrary two days for each season for Scenario C. (a) Available power (dashed-blue) and power dispatch (solid-red) from grid. (b) Required load power
(dashed-blue) and available power to cover load (solid-red). (c) Available power (dashed-blue) and power dispatch (solid-red) from PV-array. (d) Charging power of battery bank (solid-
blue) and the charging power limit (dashed-red). (e) Discharging power of battery bank. (f) Battery bank SOC (solid-blue) and SOC upper and lower limits (dashed-red). (For inter-
pretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
135
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
increase the dispatched power from the PV-array as seen in Table 6. Power & Energy Student Summit (PESS) 2013, IEEE Student Branch Bielefeld 11–16.
Meanwhile, the dispatched energy from the grid is slightly in- Goldberg, D.E., 1989. Genetic Algorithms in Search, Optimization, and Machine Learning.
Addison-Wesley, Reading.
creased. Hijjo, M., Felgner, F., Frey, G., 2016. Energy management scheme for buildings subject to
As a result, the cost of the dispatched energy from the grid is slightly planned grid outages. J. Eng. Res. Technol. 3, 58–65.
increased in comparison to Scenario A, the cost of battery lifetime Holland, J.H., 1975. Adaptation in Natural and Artificial Systems. Univ. Michigan Press.
Hooshmand, A., Asghari, B., Sharma, R., 2015. A power management system for planned
loss is highly decreased in comparison to Scenario B and the total & unplanned grid electricity outages. In: Innovative Smart Grid Technologies Latin
operation cost of the system is decreased in all seasons comparing to America (ISGT LATAM), 2015 IEEE PES. IEEE, pp. 382–386.
Scenario A and B, as shown in Table 7. Hovgaard, T.G., Edlund, K., Jørgensen, J.B., 2010. The potential of economic MPC for
power management. In: 49th IEEE Conference on Decision and Control (CDC), 2010.
IEEE, pp. 7533–7538.
From the above discussion, it can be concluded that the total dis- IEEE guide for design, operation, and integration of distributed resource island systems
patched energy from the grid can be considerably reduced, and the with electric power systems, 2011. IEEE Std 1547.4-2011, pp. 1–54. http://dx.doi.
org/10.1109/IEEESTD.2011.5960751.
dispatched energy from the PV highly increased by using the EMPC-
Jenkins, D., Fletcher, J., Kane, D., 2008. Lifetime prediction and sizing of lead–acid
based operation strategy for the PV-battery system which increases the batteries for microgeneration storage applications. IET Renew. Power Gener. 2 (3),
economic and environmental benefits of installation of the PV-array. 191–200.
Nevertheless, considering the battery lifetime with the EMPC is neces- Jossen, A., Garche, J., Sauer, D.U., 2004. Operation conditions of batteries in PV appli-
cations. Sol. Energy 76 (6), 759–769.
sary to decrease the cost of battery lifetime loss while reducing the cost Kaiser, R., 2007. Optimized battery-management system to improve storage lifetime in
of consumed energy by the load. renewable energy systems. J. Power Sources 168 (1), 58–65.
Khiareddine, A., Salah, C.B., Mimouni, M.F., 2015. Power management of a photovoltaic/
battery pumping system in agricultural experiment station. Sol. Energy 112,
5. Conclusion 319–338.
Khoury, J., Mbayed, R., Salloum, G., Monmasson, E., 2015. Optimal sizing of a residential
In this paper, we investigate the optimal operation of a PV-battery pv-battery backup for an intermittent primary energy source under realistic con-
straints. Energy Build. 105, 206–216.
system to provide a reliable and cost effective electrical energy source Khoury, J., Mbayed, R., Salloum, G., Monmasson, E., 2016. Design and implementation of
for the costumers suffering from grid scheduled blackouts (GSBs). A a real time demand side management under intermittent primary energy source
new comprehensive model for a residential PV-battery system is de- conditions with a pv-battery backup system. Energy Build. 133, 122–130.
Koutroulis, E., Kalaitzakis, K., 2004. Novel battery charging regulation system for pho-
veloped to address the GSB problem. The model considers real opera- tovoltaic applications. IEE Proc.-Electr. Power Appl. 151 (2), 191–197.
tional constraints of the system as well as the costs of the battery life- Lorenzo, E., 1994. Solar Electricity: Engineering of Photovoltaic Systems. Earthscan/
time. Controllable switches are added to control the on/off power flows James & James.
Maciejowski, J.M., 2002. Predictive Control: With Constraints. Pearson Education.
through the system. The inclusion of the controllable switches leads to a
MATLAB, 2016. Global Optimization Toolbox Version 3.4. The MathWorks Inc., Natick,
complex MINLP problem which is solved by using the genetic algo- Massachusetts.
rithm. A new operation strategy based on economic model predictive Mitra, J., 2010. Reliability-based sizing of backup storage. IEEE Trans. Power Syst. 25 (2),
control (EMPC) is developed to minimize the total cost of the energy 1198–1199.
Muñoz-García, M.A., Marin, O., Alonso-García, M., Chenlo, F., 2012. Characterization of
dispatched from the grid and the cost of battery lifetime loss under GSB. thin film pv modules under standard test conditions: Results of indoor and outdoor
To verify the effectiveness of the approach, we compare it with a tra- measurements and the effects of sunlight exposure. Sol. Energy 86 (10), 3049–3056.
ditional backup operation strategy and with an EMPC-based operation Murphy, P.M., Twaha, S., Murphy, I.S., 2014. Analysis of the cost of reliable electricity: a
new method for analyzing grid connected solar, diesel and hybrid distributed elec-
strategy neglecting the cost of battery lifetime. The results show that tricity systems considering an unreliable electric grid, with examples in Uganda.
the proposed approach leads to a significant reduction of the cost of the Energy 66, 523–534.
total energy consumed from the grid while decreasing the curtailment Najjar, M.B., Alameddine, A., Horkos, P.G., 2016. Supervisory control for sectored dis-
tributed generation during load shedding in lebanon’s power grid. In: IEEE
of the generated power from the PV-array and maximizing the battery International Conference on Renewable Energy Research and Applications (ICRERA),
lifetime. 2016. IEEE, pp. 73–78.
Nayar, C.V., Ashari, M., Keerthipala, W., 2000. A grid-interactive photovoltaic unin-
terruptible power supply system using battery storage and a back up diesel generator.
Acknowledgment
IEEE Trans. Energy Convers. 15 (3), 348–353.
Núñez-Reyes, A., Rodríguez, D.M., Alba, C.B., Carlini, M.Á.R., 2017. Optimal scheduling
The financial support from the German Academic Exchange Service of grid-connected pv plants with energy storage for integration in the electricity
market. Sol. Energy 144, 502–516.
(DAAD) for this work is gratefully acknowledged by the authors.
Park, J.-Y., Choi, S.-J., 2015. A novel datasheet-based parameter extraction method for a
single-diode photovoltaic array model. Sol. Energy 122, 1235–1244.
References Pillai, G., Hodgson, J., Insaurralde, C., Pinitjitsamut, M., Deepa, S., 2016. The techno-
economic feasibility of providing solar photovoltaic backup power, In: IEEE
International Symposium on Technology and Society (ISTAS), vol. 20, pp. 22.
Abbes, D., Martinez, A., Champenois, G., 2014. Life cycle cost, embodied energy and loss Rahman, S.A., Varma, R.K., Vanderheide, T., 2014. Generalised model of a photovoltaic
of power supply probability for the optimal design of hybrid power systems. Math. panel. IET Renew. Power Gener. 8 (3), 217–229.
Comput. Simul. 98, 46–62. Rawlings, J.B., Angeli, D., Bates, C.N., 2012. Fundamentals of economic model predictive
Alramlawi, M., Gabash, A., Li, P., 2017. Optimal operation strategy of a hybrid pv-battery control. In: IEEE 51st Annual Conference on Decision and Control (CDC), 2012. IEEE,
system under grid scheduled blackouts. In: IEEE 17th International Conference on pp. 3851–3861.
Environment and Electrical Engineering (EEEIC), 2017. IEEE, pp. 1–5. Reddy, T., 2011. Linden’s Handbook of Batteries. McGraw-Hill, New York.
Angeli, D., Amrit, R., Rawlings, J.B., 2012. On average performance and stability of Sarkar, A., Singh, J., 2010. Financing energy efficiency in developing countries-lessons
economic model predictive control. IEEE Trans. Automatic Control 57 (7), learned and remaining challenges. Energy Policy 38 (10), 5560–5571.
1615–1626. Sauer, D.U., Wenzl, H., 2008. Comparison of different approaches for lifetime prediction
Camacho, E.F., Alba, C.B., 2013. Model Predictive Control. Springer Science & Business of electrochemical systems-using lead-acid batteries as example. J. Power sources 176
Media. (2), 534–546.
De Soto, W., Klein, S., Beckman, W., 2006. Improvement and validation of a model for Seyedi, H., Sanaye-Pasand, M., 2009. New centralised adaptive load-shedding algorithms
photovoltaic array performance. Sol. Energy 80 (1), 78–88. to mitigate power system blackouts. IET Gener. Transm. Distrib. 3 (1), 99–114.
Dufo-López, R., Fernández-Jiménez, L.A., Ramírez-Rosado, I.J., Artal-Sevil, J.S., Solar energy services for professionals, 2017. < http://www.soda-pro.com > (accessed
Domínguez-Navarro, J.A., Bernal-Agustín, J.L., 2017. Daily operation optimisation of 2017-09-10).
hybrid stand-alone system by model predictive control considering ageing model. Stojanovski, G., Stankovski, M., 2012. Model Predictive Controller Employing Genetic
Energy Convers. Manage. 134, 167–177. Algorithm Optimization of Thermal Processes With Non-convex Constraints. INTECH
Gabash, A., Li, P., 2012. Active-reactive optimal power flow in distribution networks with Open Access Publisher.
embedded generation and battery storage. IEEE Trans. Power Syst. 27 (4), Tazvinga, H., Zhu, B., Xia, X., 2014. Energy dispatch strategy for a photo-
2026–2035. voltaic–wind–diesel–battery hybrid power system. Sol. Energy 108, 412–420.
Gabash, A., Li, P., 2013. Flexible optimal operation of battery storage systems for energy Tedesco, F., Mariam, L., Basu, M., Casavola, A., Conlon, M.F., 2015. Economic model
supply networks. IEEE Trans. Power Syst. 28 (3), 2788–2797. predictive control-based strategies for cost-effective supervision of community mi-
Gabash, A., Alkal, M., Li, P., 2013. Impact of allowed reverse active power flow on crogrids considering battery lifetime. IEEE J. Emerg. Sel. Top. Power Electron. 3 (4),
planning PVs and BSSs in distribution networks considering demand and EVs growth. 1067–1077.
136
M. Alramlawi et al. Solar Energy 161 (2018) 125–137
Teleke, S., Baran, M.E., Bhattacharya, S., Huang, A.Q., 2010. Rule-based control of bat- Yingli Solar, 2017. YGE 60 cell series 2, July 2017. < www.yinglisolar.com > .
tery energy storage for dispatching intermittent renewable sources. IEEE Trans. Zhang, Y., Zhang, T., Wang, R., Liu, Y., Guo, B., 2015. Optimal operation of a smart
Sustain. Energy 1 (3), 117–124. residential microgrid based on model predictive control by considering uncertainties
Wu, Z., Xia, X., 2015. Optimal switching renewable energy system for demand side and storage impacts. Sol. Energy 122, 1052–1065.
mantimal switching renewable energy system for demand side management. Zhang, Y., Gao, S., Gu, T., 2017. Prediction of iv characteristics for a pv panel by com-
Sol.Energy 114, 278–288. bining single diode model and explicit analytical model. Sol. Energy 144, 349–355.
Wu, K., Zhou, H., An, S., Huang, T., 2015. Optimal coordinate operation control for Zhao, B., Zhang, X., Chen, J., Wang, C., Guo, L., 2013. Operation optimization of stan-
wind–photovoltaic–battery storage power-generation units. Energy Convers. Manage. dalone microgrids considering lifetime characteristics of battery energy storage
90, 466–475. system. IEEE Trans. Sustain. Energy 4 (4), 934–943.
Xu, Y., Singh, C., 2011. Distribution systems reliability and economic improvement with Zhu, B., Tazvinga, H., Xia, X., 2015. Switched model predictive control for energy dis-
different electric energy storage control strategies. In: IEEE Power and Energy Society patching of a photovoltaic-diesel-battery hybrid power system. IEEE Trans. Control
General Meeting, 2011. IEEE, pp. 1–8. Syst. Technol. 23 (3), 1229–1236.
Yang, H., Zhou, W., Lu, L., Fang, Z., 2008. Optimal sizing method for stand-alone hybrid Zilouchian, A., Jamshidi, M., 2000. Intelligent Control Systems Using Soft Computing
solar–wind system with LPSP technology by using genetic algorithm. Sol. Energy 82 Methodologies. CRC Press, Inc.
(4), 354–367.
137