Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: Due to the lack of inertial support from photovoltaic (PV) systems, large-scale PV integration into the existing
Frequency control grid poses a major threat to frequency stability. To address this, numerous techniques for incorporating energy
Deloaded PV storage systems have been proposed in the existing literature, but they incur significant operational costs. PV
Optimal deloading
systems with deloading mechanisms have been researched as an alternative efficient strategy in many studies as a
Techno-economic analysis
Renewable energy
key frequency response support after a disturbance. However, deloaded PV systems incur a cost burden due to
idle reserve capacity during regular operation. Consequently, an optimum deloading of PV systems is required to
enhance the grid’s frequency stability while minimizing its total operating cost. To that aim, this paper proposes
a gradient descent-based optimization method to determine the optimum deloading of PV systems. The opti
mization considers both frequency response constraints (frequency nadir and rate of change of frequency –
ROCOF) and operating cost constraints (generation, up-regulation, and value of lost load cost). The proposed
technique is employed in a modified IEEE 39 bus New England System for varying levels of PV integration under
two distinct cases. The optimization and relevant simulations are performed in Python and DIgSILENT Power
Factory platforms. In the case of a major generator outage with PV penetrations of 30%, 40%, and 50%, the
optimal deloading percentages are 6.96%, 5.931%, and 4.968%, respectively, with no load shedding. Moreover,
in the interconnection outage case, the optimal deloading percentages increase to 14.92%, 13.3%, and 12.42%,
respectively, with load shedding. Nevertheless, in each case, deloaded PV systems consistently preserve fre
quency stability in the grid while incurring minimized operating costs. Furthermore, the performance and cost of
the proposed solution are compared to two other primary frequency response support systems, viz. Battery
Energy Storage System (BESS) and Synchronous Condenser (SC). The optimally deloaded PV system outperforms
both these devices considering both technical and economic aspects.
* Corresponding author at: Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka, 1205, Bangladesh.
E-mail addresses: ishtiakmahmud6@gmail.com (I. Mahmud), nahid@eee.buet.ac.bd (N.-A. Masood), atikjawad001@gmail.com (A. Jawad).
#
All authors contributed equally to this work.
https://doi.org/10.1016/j.epsr.2023.109457
Received 21 January 2023; Received in revised form 26 April 2023; Accepted 2 May 2023
Available online 6 May 2023
0378-7796/© 2023 Elsevier B.V. All rights reserved.
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
2
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
overall operating cost can be done. Up-regulation cost and load shedding 2.1. Photovoltaic system modeling
cost, caused by excess and insufficient deloading of PV systems,
respectively, also need to be analyzed in order to get a full under A PV system consists of a collection of PV modules. The PV system’s
standing of the economic aspect of its operation. The following contains modules are connected in series and parallel, respectively, to enhance
the key contributions of this paper: the voltage and current. The voltage of a PV system dictates its output
power. Hence, there is a particular voltage at which the PV system
• A gradient descent-based optimization technique is proposed to find generates maximum power for a specific level of temperature and irra
the optimum deloading percentage in a deloaded PV system for a diance, as depicted in Fig. 1. Usually, a PV system is operated using the
high PV enriched grid. This optimization technique focuses on MPPT algorithm. In this mode, PV is operated at a voltage VM for which
maintaining frequency stability after a disturbance while minimizing the power output, PM , is the maximum for a particular value of irradi
the operating cost under normal operating conditions. The technique ance and temperature. Moreover, VM is dependent on irradiance (R), the
is applied to two case studies involving a major generator outage and value provided by the PV module’s manufacturer (VM0 ), temperature
an interconnection network outage. (T), and temperature correlation (TCF ) factor as expressed by
• A comprehensive cost analysis is performed on the deloaded PV
ln(R)
system for different PV penetration levels under both case studies. VM = VM0 ∗ ∗ [1 + TCF (T − TSTC )] (1)
ln(RSTC )
The up-regulation cost and value of the lost load are considered in
the overall operating cost to get a complete economic view of its Here, TSTC and RSTC are standard temperature and irradiance,
operating costs. respectively.
• The proposed model’s outcome is validated by comparing it to two
current primary frequency regulation devices: Battery Energy Stor
age System (BESS) and Synchronous Condenser (SC). 2.2. Deloaded PV system modeling
The rest of the paper is organized as follows: In Section 2, a brief When a PV system is operated at a different voltage, Vnew , away from
description of the system architecture and the operation of system VM , the generated power (Pnew ) will be lower than PM (Fig. 2). Thus,
components are provided. The proposed optimization methodology with additional power can be obtained at that configuration by changing the
relevant diagrams is discussed in Section 3. Section 4 describes a operating voltage closer to VM . The reserve power, Pdl , can be expressed
detailed description of the case studies and simulation scenarios. In as
Section 5, the proposed model’s performance and validity are evaluated Pdl = PM − Pnew (2)
in comparison to the BESS and SC integrated grid over a range of PV
penetration levels under the two case studies. Section 6 presents a brief And the operating voltage Vnew for generating Pnew :
overview of the potential applications of the proposed method. Lastly, Vnew = VM + Vdl (3)
Section 7 provides some insights and a summary of the outcomes of the
proposed model. This type of system where PV is operated at a lower power compared
to PM and possesses additional reserve is known as a deloaded PV system
2. System architecture [19]. To extract this reserve power, at the moment of necessity, a signal
proportional to Δf is subtracted from the output voltage. Therefore, the
This section describes the system architecture used in this paper. A operating voltage VPVnew becomes,
brief description of the working mechanisms of the grid-connected PV VPVnew = VM + Vdl − Kgain ∗ Δf (4)
system as well as the deloaded PV system modeling is presented here. All
the modellings and the time-domain dynamic simulations are done in Following a disturbance, the frequency starts to decline in a power
DIgSILENT PowerFactory software [34]. system. As a consequence, the operating voltage VPVnew also decreases.
This increases the output power of the PV system. Hence, a PV system
can provide additional power to the grid, proportional to the frequency
deviation, following a contingency. In this manner, a PV system can
3
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
mimic the governor reaction of a conventional SG. The frequency droop details of the utilized algorithm.
mechanism of the deloaded PV model is depicted in Fig. 3.
3.1.1. Optimization equations
3. Proposed methodology At first, the cost function of the optimization problem is defined. To
this end, the overall operating cost is divided into three parts: power
The section describes the proposed method for determining the op generation cost, up-regulation cost due to PV deloading, and value of
timum PV deloading percentage to maintain system frequency stability lost load due to load shedding. As a result,
while minimizing the overall cost for a specific PV penetration. The
Overall operating cost = Power generation cost + Upregulation cost
optimization algorithm and the associated constraints are explained
first, followed by an in-depth description of the method. Note that, no + Load shedding cost
previous research has considered both stability and economical (6)
perspective when determining the optimum deloaded PV size. More
The up-regulation cost of a PV system can be calculated by multi
over, consideration of load shedding cost and up-regulation cost is also
plying the unused deloaded capacity of a PV system by the per unit up-
unique to this paper. Here, the PV penetration level is considered as the
regulation cost. The load shedding cost can be calculated by multiplying
ratio of the maximum possible output of the PV system to the total power
the quantity of load shed by the per unit load shedding cost. Per unit
generation of the grid [35]. Hence, for a fixed level of PV installation
values associated with each cost are presented in Table 1. Hence, the
capacity, the PV penetration level (pen) can be expressed as the ratio of
cost function can be written as:
output PV power at 0% deloading to the total power generation of the
grid. ng (
∑ )
f = ai + bi Pgi + ci P2gi + CPV ∗ PPV + Cur ∗ Pdl + CVOLL ∗ LS (7)
PV generation output at 0% deloading i=1
pen = (5)
Total power generation of the system Here, ai , bi and ci are generator constants for ith generator. In this
system, ai = a, bi = b and ci = c. ng is the number of generators. Pgi is
3.1. Optimization methodology the power generated by ith generator. CPV , Cur , and CVOLL are the per unit
PV power cost, up-regulation cost, and load-shedding cost respectively.
The description of the equations required for solving the optimiza PPV is the generated PV power, Pdl represents the deloaded power ca
tion problem is discussed in the optimization methodology, along with pacity and LS denotes the load shedding amount. PPV and Pdl can be
4
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
min max
Here, fnadir , fnadir , fnadir are frequency nadir, minimum allowable
written as follow: frequency nadir, and maximum allowable frequency nadir. Similarly,
(
d
) ROCOF, ROCOFmin , ROCOFmax are ROCOF, minimum allowable ROCOF,
PPV = PG ∗ pen ∗ 1 − (8) and maximum allowable ROCOF.
100
5
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
6
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
the lowest point in the frequency excursion graph of a system. And, updated solution points
ROCOF (rate of change of frequency) represents the initial rate of fre
quency deviation of a system following a disturbance. It can be calcu 4. Simulation settings and case analyses
lated by the following formula [44]:
df 1 Pdis ∗ f0 The proposed methodology for finding the optimal deloading of the
ROCOF = = ∗ (22) PV system is thoroughly assessed in this section. The used test system’s
dt 2 IR
specifications are explained in depth along with the simulation scenarios
Here, Pdis is the size of the disturbance (MW), f0 denotes the nominal that were taken into consideration. Thorough outcome analyses of the
frequency of the system before the disturbance (Hz) and IR stands for the optimizations are carried out for every simulation scenario. In addition,
system’s overall inertia. (MWs). the used algorithm for this specific optimization process is justified by
Step 4: Checking optimization constraints: After calculating the fnadir comparing it with an existing algorithm.
and ROCOF, it should be checked whether the obtained values follow the
required frequency stability constraints (16–17). If the obtained pa
rameters do not satisfy the constraints, an increase in load shedding is 4.1. Test system
introduced in the system. The value of the increment, ΔLS, is equal to
1MW. With this updated load-shedding amount, step 2 and 3 is per The IEEE 39-bus New England system is used for performing simu
lation in this research work [43]. For evaluating the stability aspects of
formed again until the constraints are satisfied.
Step 5: Calculation of gradient: The obtained point after step 4 is power systems, this test grid is a popular and widely researched standard
system [45]. It has widely been used for testing and validating new
considered the current solution point. To find the updated solution
point, the gradient at the current solution point is calculated using (18). models for dynamic analysis. In addition, the dynamic models used in
this test system (the governor model, the automated voltage regulator
Step 6: Calculation of updated solution point: From the obtained
gradient, the updated solution point can be calculated by subtracting the model, and the load model) are all state-of-the-art models that have been
scaled version of the gradient value from the current solution point. validated in earlier studies [46]. This test system is very robust in terms
Scaling is performed by multiplying the value of the gradient with the of inertia, making it suitable for frequency stability investigation. As a
learning rate. From the gradient obtained in step 5, the updated solution result, changing this grid for renewable integration and performing
is calculated using (19). frequency stability analysis is also evident in the existing literature [31].
Step 7: Checking iteration halting condition: If the difference between Taking all of these factors into account, this test system was selected for
two consecutive solution points satisfies (20), the optimization process is simulations in this research. This grid contains 19 load buses which
halted. The deloading percentage in the updated solution point is consume a total power of 6097.1 MW. The rated frequency of the grid is
considered the optimum deloading percentage of that deloaded PV 60 Hz with a base MVA of 100 MVA. To achieve a specific level of PV
system. Otherwise, steps 2 to 6 are performed again to obtain the new integration, several SGs of the test system (at buses 37,35, 36, and 32,
consecutively) are replaced with PV systems of the same rated
7
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
generation capacity. Note that, PV intermittency is not considered percentage and load-shedding size, respectively. Fig. 6(c) presents a
considering the short duration of simulations. For each PV penetration three-dimensional view of the obtained locus of solution points at each
level, the total power generation and load are kept constant. A step for better visualization of the process. Additionally, an arrow in the
single-line diagram of the modified test grid with 50% PV penetration is graph indicates the direction of the optimum solution path finding. The
depicted in Fig. 5. figures demonstrate the gradual convergence of solution points towards
the optimal solution with minimum cost.
4.2. Simulation scenarios The overall operating cost, which includes the costs of power gen
eration (SG: 619,480.5847 $/h and PV system: 113,854.14 $/h), up-
The proposed approach is applied to two outage cases in order to regulation (1410.46488 $/h), and load shedding (0 $/h), is deter
assess its implementation: (1) A major generator outage (Generator 5 – mined to be 734,745.1896 $/h. As the optimally deloaded system does
508 MW), and (2) An interconnection network outage (Generator 1 – not require load shedding to satisfy frequency constraints, the load
1000 MW). For each outage case, three different PV penetration sce shedding cost is zero in this scenario.
narios are simulated. These scenarios represent 30% to 50% PV pene
tration in the grid with a 10% increment in each step. Additional details 4.3.1.2. 40% PV penetration scenario. For this particular scenario, the
about these scenarios are presented in Table 2. For each PV penetration PV penetration is enhanced to 2457 MW. Employing the same meth
scenario, steps 1 to 7 are performed sequentially to obtain the optimal odology as the previous scenario, the optimization converged after 89
deloading percentage. For automation and optimization purposes, Py iterations with an optimal deloading percentage of 5.931% (145.689
thon is integrated with DIgSILENT PowerFactory in this work. MW). Note that there is still no required load shedding for the optimal
solution. The grid is capable of sustaining frequency stability solely with
4.3. Result analysis deloaded PV support. Additionally, Fig. 7 illustrates the entire process
through a three-dimensional view. The solution point convergence
The goal of this subsection is to provide a concise yet thorough pattern is comparable to the previous scenario. Despite the increased
outcome analysis based on the simulation findings from the case studies deloaded PV size, the overall deloading percentage has decreased with
mentioned above. The suggested method’s viability is evaluated an increased PV penetration.
throughout the process. Due to higher PV integration, the power generation cost of SGs has
declined to 488,720.1929 $/h. Despite the increase in PV generation
4.3.1. Case 1: major generator outage event cost to 151,805.52 $/h, the impact is not as significant as the reduction
The Generator 5 (508 MW) outage event is simulated in this case in SG production cost. Moreover, the up-regulation cost has marginally
study to optimize PV deloading using the proposed approach. three increased to 1602.579924 $/h due to the larger PV deloading size.
scenarios for this case—30%, 40%, and 50% PV penetration—are given Nevertheless, the total overall operating cost has reduced to
below with the associated findings and descriptions. 642,128.2928 $/h. Like the previous scenario, load shedding cost is 0 $/
h denoting the absence of load shedding.
4.3.1.1. 30% PV penetration scenario. For this level of PV penetration,
the PV capacity is set equal to 1842 MW. The gradient descent optimi 4.3.1.3. 50% PV penetration scenario. In this final scenario, PV pene
zation algorithm is initiated with an initial value of d0=0 and LS0=0 tration accounts for half of the total generation in the grid. The same
MW. Here, (11) represents the cost function, and (12)-(17) demonstrate optimization process yields an optimal deloading percentage of 4.968%,
the corresponding optimization constraints. To optimize the function, an which corresponds to a PV deloading size of 152.542 MW. The three-
adaptive learning rate with an initial value of 0.1 is employed. Following dimensional view, depicted in Fig. 8, indicates a visible trend of solu
initialization, a dynamic simulation of the generator outage event is tion points toward the optimal solution. In this scenario, the required
conducted. The frequency excursion curve from the simulation is uti deloading size marginally increases, while the corresponding deloading
lized to calculate the corresponding fnadir and ROCOF, which are then percentage decreases compared to the previous scenarios. Interestingly,
evaluated against the predefined constraints. In the event of violating no load shedding is again required to maintain system stability during
constraints, the load shedding value is incrementally increased by 1 the outage event. In terms of cost at the optimal solution point, the
MW, and the outage event is repeatedly performed with updated values overall operating cost decreases to 504,149.1332 $/h. Similarly, to the
until the constraints are satisfied. The specific sets of values that previous scenarios, the cost of SG generation significantly decreases to
conform to the constraints are chosen as the current solution point. 312,714.2663 $/h, while the cost of PV generation increases at a slower
Subsequently, the objective function (11) and gradient (18) are evalu rate to 189,756.9 $/h. The up-regulation cost also increases to
ated at the current solution point, and the solution point is updated 1677.96684 $/h, while the cost of load shedding remains at 0 $/h.
using the gradient descent technique in accordance with (19). The dif According to the findings, as the PV integration increases, the
ference between two successive solution points is then calculated to optimal deloading size required also increases. However, the overall
verify the fulfillment of the halting criteria (20). If the difference exceeds percentage of optimal deloading decreases gradually. Load shedding is
the predetermined threshold, the iterative process is repeated with the not required for any scenarios in the event of a major generator outage
updated solutions until convergence is achieved. while ensuring minimum cost. Furthermore, the overall operating cost
In this instance, the iterative process was performed 67 times to decreases with increased PV penetration.
obtain an optimum deloading percentage of 6.96% (128.224 MW) with
0 MW load shedding. Fig. 6(a)-(b) illustrate the optimization process 4.3.2. Case 2: interconnection network outage event
path to identify the point of minimum cost concerning the deloading In the previous case, no load shedding was required to maintain
Table 2
Summary of simulation scenarios.
Case Study Simulation PV penetration SG output PV generator output
scenario (%) (MW) (MW)
Case 1: Major generator outage event Case 2: Interconnection network outage event Scenario 1 30 4299 1842
(Generator 5 – 508 MW) (Generator 1 – 1000 MW) Scenario 2 40 3684 2457
Scenario 3 50 3070.5 3070.5
8
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
Fig. 6. Visual representation of finding optimal solutions using gradient descent algorithm in 30% PV penetration case for a major generator outage: (a) Cost vs
Deloading percentage, (b) Cost vs Load shedding size, and (c) Cost vs Deloading percentage vs Load shedding size.
minimum operating costs during the major generator outage. To and upregulation costs, to determine the optimal deloading.
investigate whether more severe disturbances result in a combination of
deloading and load shedding, a larger tripping event is investigated. 4.3.2.2. 40% PV penetration. The optimization process yielded an
Thus, the interconnection network outage (Generator 1 - 1000 MW) is optimal solution of 13.3% PV deloading, equivalent to 326.8 MW, and
considered for the 30%, 40%, and 50% PV penetration scenarios in this 119 MW loading, after 97 iterations for the 40% PV penetration. Inter
case. estingly, the deloading percentage followed a decreasing trend, similar
to the previous generator and interconnection outage cases. In order to
4.3.2.1. 30% PV penetration. Following the same process as the previ maintain economic frequency stability, load shedding was required. The
ous scenarios, the optimization is conducted and the solution is found solution trajectory, shown in Fig. 10, exhibited increased complexity
after 72 iterations. The result shows that 14.92% of PV deloading towards the end due to multiple cost combinations. However, the min
(equivalent to 274.83 MW) is required along with 110.6 MW load imum operating cost achieved was $741,246/h, which is lower than that
shedding, as demonstrated in Fig. 9. Due to significant power distur of the 30% PV scenario. The incorporation of more PV generation in
bance and increased grid dynamics, the resultant path of the simulation place of SGs contributed to this cost reduction.
points is also more scattered. Unlike the previous cases, this scenario
requires load shedding to preserve the frequency response of the system 4.3.2.3. 50% PV penetration. Increasing the PV capacity to 50% of the
while ensuring economic viability. The overall operating cost (similar to total generation required a longer convergence time of 105 iterations for
Case 1) for this scenario is $846,391/h, which is the highest among all optimization. The optimal solution involved a combination of 12.42%
cases due to the significant disturbance size. For the first time, the load PV deloading, corresponding to 381.4 MW, and 129 MW of load shed
shedding cost (55,500 $/h) is also incorporated into it. Nonetheless, this ding for economic frequency stability. Fig. 11 depicts the trajectory of
specific combination of deloaded PV and load shedding results in the the solution points to the optimum point in three dimensions. Like the
minimum operating cost compared to all other possible solution points. previous scenarios, the overall deloading percentage decreased while
This outcome justifies the robustness of our optimization approach, the size of load shedding slightly increased. The operating cost also
which effectively considers all relevant factors, including load shedding decreased to $622,541/h, the lowest among all three scenarios for
9
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
Fig. 7. Visual representation of finding optimal solutions using gradient descent algorithm in 40% PV penetration case for a major generator outage (Cost vs
Deloading percentage vs Load shedding size).
Fig. 8. Visual representation of finding optimal solutions using gradient descent algorithm in 50% PV penetration case for a major generator outage (Cost vs
Deloading percentage vs Load shedding size).
10
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
For validation purposes, the selected test system is the IEEE 39 bus,
which was previously used in Section 4. Simulations are performed for
all PV penetration scenarios for both cases (Case 1 and Case 2) discussed
in Section 4. For each scenario, the outage event simulation is performed
for the optimally deloaded PV as well as BESS and SC. Note, both the
BESS and SC capacities are set to match the optimal deloading size. A
Fig. 10. Visual representation of finding optimal solutions using gradient UFLS setting is also used (5% load shed) which is activated if the system
descent algorithm in 40% PV penetration case for interconnection network frequency falls below 59 Hz. The overall operating cost of these sce
outage (Cost vs Deloading percentage vs Load shedding size). narios is also compared against each other to get an understanding of
their feasibility.
11
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
Table 3
Optimal deloading percentage and deloaded power capacity for each PV penetration scenario.
Case Study PV penetration level Optimal deloading Optimal Load-shedding Deloaded power Number of
(%) (%) (MW) (MW) iterations
Table 4
Summary of the overall operating cost in optimally deloaded case for each PV penetration scenario.
Case Study PV penetration Overall operating Generation cost of Generation cost Power generation Up-regulation Load shedding
level (%) cost ($/h) SG ($/h) of PV ($/h) cost ($/h) cost ($/h) cost ($/h)
Table 5
Comparative analysis between gradient-descent optimization and genetic algorithm-based optimization.
Case Study PV penetration Gradient descent optimization Genetic algorithm-based optimization (10 chromosomes)
level (%) Optimal solution No. of Simulation Time Optimal solution No. of Simulation Time
for deloading (%) simulations (in seconds) for deloading (%) simulations (in seconds)
required required
within 59 Hz without activating UFLS. Moreover, the ROCOF values are integrated grid are 58.85121 Hz, 58.8742 Hz, and 58.86606 Hz,
0.13156 Hz/s, 0.14182 Hz/s, and 0.14404 Hz/s, consecutively. Though respectively, indicating a slightly weaker performance than the deloa
all ROCOF magnitudes are well within the acceptable limit, the deloa ded PV. Meanwhile, the frequency nadir points for the SC integrated grid
ded PV scenario experiences the minimum ROCOF following the outage. (58.4324 Hz, 58.372 Hz, and 58.0779 Hz, respectively) are the lowest
In a similar manner, the simulated frequency excursion curves for the among all three mechanisms, suggesting a weak ability to maintain
40% PV penetration are shown in Fig. 12(b). The calculated frequency frequency response. These results provide further validation for the use
nadir points are 59.0434 Hz, 58.981 Hz, and 58.9712 Hz, respectively of the optimized deloaded PV system in ensuring grid stability.
for the same three mechanisms consecutively. In addition, the ROCOF
magnitudes are 0.14314 Hz/s, 0.15298 Hz/s, and 0.1622 Hz/s, in 5.2.1.2. Operating cost comparison. A comprehensive analysis of oper
sequence. Interestingly, the optimized deloaded PV support again de ating costs is performed using the per-unit data provided in Table 1 for
livers better performance than the other two supporting mechanisms, the optimized deloaded PV, BESS, and SC integrated grid. The results of
halting frequency decline without triggering UFLS and doing so with the the analysis reveal that for the 30% PV penetration, the overall oper
lowest ROCOF values. Finally, for the 50% PV penetration, the simula ating cost of the deloaded PV integrated grid is 734,745.1896 $/h,
tions are again conducted and the associated frequency excursion curves excluding any load shedding cost. Conversely, for both the BESS and SC
are illustrated in Fig. 12(c). The outcomes display a similar pattern with connected grid, the total operating cost is considerably higher at
deloaded PV maintaining the frequency nadir at 59.0451 Hz without 19,222,903.16 $/h and 892,193.8464 $/h, respectively. It is noteworthy
UFLS activation. BESS and SC, on the other hand, trigger UFLS activa that the power generation costs for all three instances are the same.
tion, resulting in frequency nadirs of 58.99662 Hz and 58.92054 Hz, in Furthermore, neither the BESS nor the SC integrated grid has any PV
that order. When it comes to ROCOF numbers, the deloaded PV inte reserve allocated for frequency response support (0$/h up-regulation
grated grid has the lowest ROCOF (0.16 Hz/s), while the BESS and SC cost). However, the significantly higher operating cost of these two
integrated grids have slightly greater ROCOFs of 0.16232 Hz/s and mechanisms explains the considerable difference in cost with the
0.17264 Hz/s, consecutively. deloaded PV integrated grid. Additionally, UFLS is activated in both of
To better understand the performance differences between the three them, and 5% of the total load is disconnected, resulting in additional
frequency response support mechanisms, additional simulations were load shedding costs. These reasons justify the lower overall operating
conducted without the UFLS scheme. The resulting frequency response cost of the deloaded PV system in comparison to the other two
curves for the 30%, 40%, and 50% PV integrated scenarios are presented mechanisms.
in Figs. 13(a)-(c), respectively. As anticipated, the frequency nadir The same analyses are repeated for the 40% and 50% PV penetration
points for the deloaded PV remain unchanged across all scenarios. scenarios, and the same trend is observed as in the previous scenario. For
However, it is evident that the BESS and SC exhibit significant perfor the 40% PV penetration, the overall operating cost of the same-sized
mance differences. Specifically, the frequency nadir points for the BESS deloaded PV system, BESS, and SC connected grid is 642,128.2928
12
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
Fig. 12. Frequency response curve for a major synchronous generator outage event (with UFLS scheme activated) for PV penetration level of (a) 30%, (b) 40%,
(c) 50%.
$/h, 21,627,589.72 $/h, and 800,111.3788 $/h, respectively. The 5.2.2.1. Frequency response performance comparison. In the case of 30%
deloaded PV integrated grid remains economically feasible by a signif PV integration in the grid, Fig. 14(a) illustrates the simulated frequency
icant margin when compared to the other two mechanisms. Similarly, excursion plots for the deloaded PV, BESS, and SC integrated grid. The
for the 50% PV penetration scenario, the cost of the deloaded PV inte frequency nadir values calculated for these mechanisms are 58.97 Hz,
grated grid is the lowest (504,149.1332 $/h), while the cost for BESS 58.82 Hz, and 58.802 Hz, consecutively. In all cases, the frequency nadir
(22,469,565.09 $/h) and SC (662,341.9318 $/h) integration in the grid falls below 59 Hz, triggering UFLS for the predetermined outage of 110
remains very high. Although BESS provides good frequency support, the MW. It should be noted that even under these conditions, the deloaded
cost of a BESS-integrated grid is much higher than that of the other two PV system exhibits the highest frequency nadir value. Additionally, the
mechanisms in each scenario. Furthermore, SC integration increases the ROCOF values are 0.2274 Hz/s, 0.2392 Hz/s, and 0.2438 Hz/s,
overall cost, but with very poor frequency support. These results once consecutively. As in the previous case, the ROCOF values are within the
again confirm the economic feasibility of utilizing the optimized allowable range, but the deloaded PV support once again generates the
deloaded PV support mechanism. lowest ROCOF. These results demonstrate that even during intercon
nection outages, the optimized PV support performs better than the
5.2.2. Case 2: interconnection network outage event other two mechanisms in preserving frequency stability.
A second case study is conducted to validate the performance of the After conducting simulations for the 40% and 50% PV penetration
optimized deloaded PV system in the event of an interconnection scenarios, the corresponding frequency excursion plots are presented in
network outage. As discussed in Section 4, the deloading sizes for the Figs. 14(b)-(c). It can be seen that the deloaded PV system still delivers
30%, 40%, and 50% PV penetration scenarios were determined to be the superior frequency response as the PV penetration increases in the
180.8 MW, 326.8 MW, and 381.4 MW, respectively. To minimize costs grid. Specifically, for the 40% PV scenario, the frequency nadir values
during the outage event, load shedding was also necessary, with corre are 58.986 Hz, 58.812 Hz, and 58.808 Hz, consecutively for the deloa
sponding amounts of 110 MW, 119 MW, and 129 MW for each scenario. ded PV, BESS, and SC integrated grid. For the 50% scenario, the results
For comparison purposes, the same-sized BESS and SC systems were follow a similar trend, with values of 58.973 Hz, 58.809 Hz, and 58.802
integrated into the grid with the same UFLS settings. As the optimization Hz, consecutively. In both scenarios, the ROCOF values demonstrate the
findings already include load shedding values for optimum operation, it same trend as seen in the graphs. It is noteworthy that SC support re
should be noted that only UFLS scheme integrated grids are taken into mains the weakest mechanism, while BESS performs somewhat better
consideration. The performance and operating cost of the three mech than SC. Nonetheless, the optimized deloaded PV system still out
anisms during the outage event are discussed below. performs both of them in providing frequency support during
13
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
Fig. 13. Frequency response curve for a major synchronous generator tripping event (without UFLS schemes activated) for PV penetration level of (a) 30%, (b) 40%,
(c) 50%.
interconnection outages. performance of the three cases. Although the frequency nadir of the
BESS integrated grid was somewhat closer to the deloaded PV integrated
5.2.2.2. Operating cost comparison. Similar to the previous case study, grid, the SC integrated grid shows far inferior performance. Even in the
this interconnection outage event’s total operating cost analysis is car interconnection network outage case where load shedding was required,
ried out. Note, load shedding cost is added for each of the cost calcu the deloaded PV system provided enhanced frequency stability
lations. The findings indicate that, even with the substantial up- compared to the other mechanisms. Table 6 and Table 7 provide an
regulation and load shedding costs, the cost of operation of the deloa overview of the frequency nadir points and the ROCOF magnitudes
ded PV integrated grid for 30% PV integration (846,391.1 $/h) is the found in the deloaded PV, BESS, and SC integrated grids for both cases.
lowest. The operating costs of BESS and SC integrated grid, however, are Regarding the operating cost aspect, increased PV penetration in the
much higher, coming in at 28,409,373.59 $/h, and 956,604.03 $/h grid leads to a reduction in the overall operating cost in each case.
respectively. Because of the high operating costs of BESS and SC, there is Furthermore, the deloaded PV integration incurs the lowest cost in every
a substantial cost discrepancy. The same trend is observed when PV scenario, whether the load shedding cost is involved or not. Due to the
penetration increases to 40% and 50%. The deloaded PV integrated grid low operating cost of PV systems, deloading operation remains a cost-
again has the lowest cost (741,245.8 $/h), while the integration of BESS effective option compared to the utilization of BESS or SC. For better
(24,871,962.33) and SC (829,589.79) results in a drastic increase in cost comparison, Table 8 provides an in-depth cost breakdown of each sce
for 40% PV penetration. Similarly, for 50% PV penetration, the oper nario in the two cases.
ating costs of the deloaded PV, BESS, and SC integrated grid are
622,541.3 $/h, 20,878,862.04 $/h, and 686,680.377 $/h, respectively. 6. Discussion on the applicability of the proposed method
These results highlight the economic benefits of integrating deloaded PV
into the grid and the economic challenges associated with integrating The proposed method outlined in this research considers both fre
BESS and SC systems. quency response metrics and operating cost constraints to determine the
The results of the simulations conducted in both major generator optimal deloading of PV systems. Before the implementation of the
outage and interconnection network outage cases demonstrate the su proposed approach, simulations need to be carried out during the
periority of the proposed deloaded model over the existing support planning stage of the grid, while taking into consideration the grid pa
mechanisms of BESS and SC. In the major generator outage case, only rameters. Once optimization is successfully achieved, the optimal
the optimally deloaded PV system was able to maintain the system deloading of PV systems can be implemented in the grid to ensure both
frequency above 59 Hz without triggering the UFLS scheme. The stable and economic operation. Since this method is generic, it can be
simulation without UFLS activation clearly distinguishes the applied to any small or large grid as needed, and constraint limits can be
14
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
Fig. 14. Frequency response curve for interconnection network outage event (with UFLS scheme activated) for PV penetration level of (a) 30%, (b) 40%, (c) 50%.
Table 6
Frequency nadir points for Deloaded PV, BESS, and SC integrated grid (with and without UFLS scheme activation) in conducted case studies.
Case Study PV penetration Deloaded PV BESS without SC without Deloaded PV with BESS with SC with UFLS
level (%) without UFLS (Hz) UFLS (Hz) UFLS (Hz) UFLS (Hz) UFLS (Hz) (Hz)
Case 1: Major Generator Outage 30% 59.0305 58.85121 58.4324 59.0305 58.9951 58.96756
Event (508 MW) 40% 59.0434 58.87426 58.372 59.0434 58.981 58.9712
50% 59.0451 58.86606 58.0779 59.0451 58.99662 58.92054
Case 2: Interconnection Network 30% – – – 58.97 58.82 58.802
Outage Event (1000 MW) 40% – – – 58.986 58.812 58.808
50% – – – 58.973 58.809 58.802
Table 7
ROCOF magnitudes for Deloaded PV, BESS, and SC integrated grid (with and without UFLS scheme activation) in conducted case studies.
Case Study PV penetration Deloaded PV BESS without SC without Deloaded PV with BESS with SC with UFLS
level (%) without UFLS (Hz/s) UFLS (Hz/s) UFLS (Hz/s) UFLS (Hz/s) UFLS (Hz/s) (Hz/s)
Case 1: Major Generator Outage 30% 0.13156 0.14182 0.14404 0.13156 0.14182 0.14404
Event (508 MW) 40% 0.14314 0.15298 0.1622 0.14314 0.15298 0.1622
50% 0.16 0.16232 0.17264 0.16 0.16232 0.17264
Case 2: Interconnection Network 30% – – – 0.2274 0.2392 0.2438
Outage Event (1000 MW) 40% – – – 0.2581 0.2653 0.2715
50% – – – 0.2738 0.2874 0.2998
modified as per the grid planner’s requirements. If any modifications are operation are also not taken into account. However, these limitations
made to the grid, calculations must be performed again, and the updated can be overcome by adapting the constraints to consider extreme situ
modified deloading can be employed for efficient grid operation. ations and running simulations accordingly. Additionally, this method
Therefore, this method is adaptable for practical implementation in any can be utilized in the grid considering any major disturbance event as
renewable integrated grid. necessary. As observed from the results of this research, the proposed
Note, PV intermittency is not considered in this work for the short method works efficiently in every case with the utilized algorithm, and
duration of the simulations. The load-side fluctuations during grid the deloaded PV support provides the expected outcomes. Therefore, the
15
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
operating
Overall
0.892
0.800
0.662
0.956
0.829
0.687
7. Conclusions
cost
This research presents a novel optimization methodology for deter
mining the optimal deloading percentage of deloaded PV systems in a
shedding
0.154
0.154
0.154
0.056
0.060
0.065
renewable integrated grid. Unlike previous research, this work considers
Load
cost
both economic factors (including power generation, load shedding, and
up-regulation costs) and frequency stability when determining the
Synchronous condenser (M$/h)
0.00533
0.00606
0.00635
IEEE 39 bus test system, and two different case studies of outage events
0.0919
0.0683
0.113
are considered. The results show that in scenarios with 30%, 40%, and
50% PV penetration under the major generator outage case, the optimal
PV deloading is 6.96%, 5.931%, and 4.968%, respectively (with no load
generation
19.223
21.628
22.469
28.409
24.872
20.878
cost
SC of the same size, it avoids the activation of UFLS in the first case while
preserving frequency stability. Although load shedding was required in
the second case, deloaded PV still provided better frequency stability in
shedding
the grid. Moreover, the overall operating cost of the deloaded PV inte
0.154
0.154
0.154
0.056
0.060
0.065
Load
18.336
20.834
21.814
27.566
24.134
20.261
same time. Note that, the variability of PV output with respect to time
generation
0.733
0.642
0.502
0.788
0.678
0.554
deloading.
0.735
0.642
0.504
0.846
0.741
0.623
cost
cost
0.00168
0.0016
0.0031
0.0036
0.0041
& editing.
cost
Up-
Deloaded PV (M$/h)
0.733
0.642
0.502
0.788
0.678
0.554
Data availability
penetration
level (%)
30%
40%
50%
30%
40%
50%
Case 2: Interconnection
Acknowledgement
Event (1000 MW)
Generator Outage
Event (508 MW)
Network Outage
Case 1: Major
16
I. Mahmud et al. Electric Power Systems Research 221 (2023) 109457
References [24] P. Kushwaha, V. Prakash, R. Bhakar, U.R. Yaragatti, Synthetic inertia and
frequency support assessment from renewable plants in low carbon grids, Electr.
Power Syst. Res. 209 (2022), 107977, https://doi.org/10.1016/j.
[1] S. Mokeke, L.Z. Thamae, The impact of intermittent renewable energy generators
epsr.2022.107977.
on Lesotho national electricity grid, Electr. Power Syst. Res. 196 (2021), 107196,
[25] H. Xin, Y. Liu, Z. Wang, D. Gan, T. Yang, A new frequency regulation strategy for
https://doi.org/10.1016/j.epsr.2021.107196.
photovoltaic systems without energy storage, IEEE Trans. Sustain. Energy 4 (4)
[2] S. Wang, Q. Dong, J. Zhang, J. Sun, X. Gu, C. Chen, Robustness assessment of
(2013) 985–993, https://doi.org/10.1109/TSTE.2013.2261567.
power network with renewable energy, Electr. Power Syst. Res. 217 (2023),
[26] C. Rahmann, A. Castillo, Fast frequency response capability of photovoltaic power
109138, https://doi.org/10.1016/j.epsr.2023.109138.
plants: the necessity of new grid requirements and definitions, Energies 7 (10)
[3] CCES. "Renewable Energy". Center for climate and energy solutions. https://www.
(2014) 6306–6322, https://doi.org/10.3390/en7106306.
c2es.org/content/renewable-energy/. (accessed Aug. 2022).
[27] M. Tavakkoli, J. Adabi, S. Zabihi, R. Godina, E. Pouresmaeil, Reserve allocation of
[4] Installed Solar Energy Capacity, OWD, 2022 [Online]. Available: https://ourwo
photovoltaic systems to improve frequency stability in hybrid power systems,
rldindata.org/grapher/installed-solar-pv-capacity.
Energies 11 (10) (2018) 2583, https://doi.org/10.3390/en11102583.
[5] M.Z. Zakariya, J. Teh, A systematic review on cascading failures models in
[28] P. Verma, T. Kaur, Power reserve control strategy of PV system for active power
renewable power systems with dynamics perspective and protections modeling,
reserve under dynamic shading patterns, Array (2022), 100250, https://doi.org/
Electr. Power Syst. Res. 214 (2023), 108928, https://doi.org/10.1016/j.
10.1016/j.array.2022.100250.
epsr.2022.108928.
[29] C. Zhong, H. Li, Y. Zhou, Y. Lv, J. Chen, Y. Li, Virtual synchronous generator of PV
[6] Ö. Çelik, M. Büyük, A. Tan, Mitigation of power oscillations for energy harvesting
generation without energy storage for frequency support in autonomous microgrid,
capability improvement of grid-connected renewable energy systems, Electr.
Int. J. Electr. Power Energy Syst. 134 (2022), 107343, https://doi.org/10.1016/j.
Power Syst. Res. 213 (2022), 108756, https://doi.org/10.1016/j.
ijepes.2021.107343.
epsr.2022.108756.
[30] H. Sharma, C. Bhende, P. Sekhar, Enhancement in frequency response capability of
[7] C. Shuai, Y. Deyou, G. Weichun, L. Chuang, C. Guowei, K. Lei, Global sensitivity
grid connected photovoltaic system, in: 2022 IEEE 2nd International Conference on
analysis of voltage stability in the power system with correlated renewable energy,
Sustainable Energy and Future Electric Transportation (SeFeT), IEEE, 2022,
Electr. Power Syst. Res. 192 (2021), 106916, https://doi.org/10.1016/j.
pp. 1–6, https://doi.org/10.1109/SeFeT55524.2022.9909329.
epsr.2020.106916.
[31] A. Jawad, N. Al-Masood, A systematic approach to estimate the frequency support
[8] M.Y. Yousef, M.A. Mosa, A.A. Ali, S.M.E. Masry, A.M.A. Ghany, Frequency response
from large-scale PV plants in a renewable integrated grid, Energy Rep. 8 (2022)
enhancement of an AC micro-grid has renewable energy resources based generators
940–954, https://doi.org/10.1016/j.egyr.2021.12.017.
using inertia controller, Electr. Power Syst. Res. 196 (2021), 107194, https://doi.
[32] K. Bates, J. Bradshaw, Costing systems and the spare capacity conundrum:
org/10.1016/j.epsr.2021.107194.
avoiding the death spiral, SSRN Electr. J. (2011), https://doi.org/10.2139/
[9] K.S. El-Bidairi, H.D. Nguyen, T.S. Mahmoud, S.D.G. Jayasinghe, J.M. Guerrero,
ssrn.2028699.
Optimal sizing of battery energy storage systems for dynamic frequency control in
[33] W. Kuckshinrichs, T. Schröder, Value of lost load: an efficient economic indicator
an islanded microgrid: a case study of Flinders Island, Australia, Energy 195
for power supply security? A literature review, Front. Energy Res. 3 (2015),
(2020), 117059, https://doi.org/10.1016/j.energy.2020.117059.
https://doi.org/10.3389/fenrg.2015.00055.
[10] X. He, S. Ge, H. Liu, Z. Xu, Y. Mi, C. Wang, Frequency regulation of multi-microgrid
[34] DIgSILENT, DigSILENT GmbH, DIgSILENT PowerFactory V15.1- User Manual,
with shared energy storage based on deep reinforcement learning, Electr. Power
DIgSILENT, 2013.
Syst. Res. 214 (2023), 108962, https://doi.org/10.1016/j.epsr.2022.108962.
[35] O. Gandhi, D.S. Kumar, C.D. Rodríguez-Gallegos, D. Srinivasan, Review of power
[11] R. Ansaripour, H. Barati, A. Ghasemi, A chance-constrained optimization
system impacts at high PV penetration Part I: factors limiting PV penetration, Solar
framework for transmission congestion management and frequency regulation in
Energy 210 (2020) 181–201, https://doi.org/10.1016/j.solener.2020.06.097.
the presence of wind farms and energy storage systems, Electr. Power Syst. Res.
[36] B. Singh, R. Mahanty, S. Singh, Security Constraints Optimal Power Flow and
213 (2022), 108712, https://doi.org/10.1016/j.epsr.2022.108712, 2022/12/01/.
Generator Rescheduling Using Interior Point Optimization Technique, Power
[12] K.N. Akpinar, B. Gundogdu, O. Ozgonenel, C. Gezegin, An intelligent power
Systems Division, CPRI, 2012, pp. 1–7.
management controller for grid-connected battery energy storage systems for
[37] "IRENA POWER GENERATION COSTS IN 2020," ed: IRENA, 2020.
frequency response service: a battery cycle life approach, Electr. Power Syst. Res.
[38] C. Lueken, G.E. Cohen, J. Apt, Costs of solar and wind power variability for
216 (2023), 109040, https://doi.org/10.1016/j.epsr.2022.109040.
reducing CO2 emissions, Environ. Sci. Technol. 46 (17) (2012) 9761–9767,
[13] C. Jamroen, The effect of SoC management on economic performance for battery
https://doi.org/10.1021/es204392a.
energy storage system in providing voltage regulation in distribution networks,
[39] M.R. Tur, Calculation of value of lost load with a new approach based on time and
Electr. Power Syst. Res. 211 (2022), 108340, https://doi.org/10.1016/j.
its effect on energy planning in power systems, Int. J. Renew. Energy Res. 10 (1)
epsr.2022.108340.
(2020) 416–424, https://doi.org/10.20508/ijrer.v10i1.10237.g7888.
[14] S. Barakat, A. Emam, M.M. Samy, Investigating grid-connected green power
[40] W. Cole, A.W. Frazier, and C. Augustine, "Cost projections for utility-scale battery
systems’ energy storage solutions in the event of frequent blackouts, Energy Rep. 8
storage: 2021 update," United States, 2021. [Online]. Available: https://www.osti.
(2022) 5177–5191, https://doi.org/10.1016/j.egyr.2022.03.201.
gov/biblio/1786976[Online]. Available: https://www.osti.gov/servlets/pur
[15] A.F. Güven, N. Yörükeren, M.M. Samy, Design optimization of a stand-alone green
l/1786976.
energy system of university campus based on Jaya-Harmony Search and Ant
[41] C. Li, M. Liu, Y. Guo, H. Ma, H. Wang, X. Yuan, Cost analysis of synchronous
Colony Optimization algorithms approaches, Energy 253 (2022), 124089, https://
condenser transformed from thermal unit based on LCC theory, Processes 10 (9)
doi.org/10.1016/j.energy.2022.124089.
(2022) 1887, https://doi.org/10.3390/pr10091887.
[16] M.M. Samy, A. Emam, E. Tag-Eldin, S. Barakat, Exploring energy storage methods
[42] A.-.L. Cauchy, M´ethode g´en´erale pour la r´esolution des syst`emes d’´equations
for grid-connected clean power plants in case of repetitive outages, J. Energy Stor.
simultan´ees, Comptes rendus de l’Académie des Sciences, Paris (25) (1847)
54 (2022), 105307, https://doi.org/10.1016/j.est.2022.105307.
536–538.
[17] A. Fatih Güven, M. Mahmoud Samy, Performance analysis of autonomous green
[43] A. Cabrera-Tobar, E. Bullich-Massagué, M. Aragüés-Peñalba, O. Gomis-Bellmunt,
energy system based on multi and hybrid metaheuristic optimization approaches,
Review of advanced grid requirements for the integration of large scale
Energy Convers. Manage. 269 (2022), 116058, https://doi.org/10.1016/j.
photovoltaic power plants in the transmission system, Renew. Sustain. Energy Rev.
enconman.2022.116058.
62 (2016) 971–987, https://doi.org/10.1016/j.rser.2016.05.044.
[18] C. Mokhtara, B. Negrou, N. Settou, B. Settou, M.M. Samy, Design optimization of
[44] T. Athay, R. Podmore, S. Virmani, A practical method for the direct analysis of
off-grid Hybrid Renewable Energy Systems considering the effects of building
transient stability, IEEE Trans. Power Appar. Syst. (2) (1979) 573–584, https://doi.
energy performance and climate change: case study of Algeria, Energy 219 (2021),
org/10.1109/TPAS.1979.319407.
119605, https://doi.org/10.1016/j.energy.2020.119605.
[45] P. Demetriou, M. Asprou, J. Quiros-Tortos, E. Kyriakides, Dynamic IEEE test
[19] P. Zarina, S. Mishra, P. Sekhar, Deriving inertial response from a non-inertial PV
systems for transient analysis, IEEE Syst. J. 11 (4) (2017) 2108–2117, https://doi.
system for frequency regulation, in: 2012 IEEE International Conference on Power
org/10.1109/JSYST.2015.2444893, doi: 10.1109/JSYST.2015.2444893.
Electronics, Drives and Energy Systems (PEDES), IEEE, 2012, pp. 1–5, https://doi.
[46] Q. Wang, F. Li, Y. Tang, Y. Xu, Integrating model-driven and data-driven methods
org/10.1109/PEDES.2012.6484409.
for power system frequency stability assessment and control, IEEE Trans. Power
[20] P. Zarina, S. Mishra, P. Sekhar, Exploring frequency control capability of a PV
Syst. 34 (6) (2019) 4557–4568, https://doi.org/10.1109/TPWRS.2019.2919522.
system in a hybrid PV-rotating machine-without storage system, Int. J. Electr.
[47] A.M.A. Youssef, Z.J. Zhai, R.M. Reffat, Genetic algorithm based optimization for
Power Energy Syst. 60 (2014) 258–267, https://doi.org/10.1016/j.
photovoltaics integrated building envelope, Energy Build. 127 (2016) 627–636,
ijepes.2014.02.033.
https://doi.org/10.1016/j.enbuild.2016.06.018.
[21] P. Zarina, S. Mishra, Cost benefit of using deloaded PV instead of battery, in: 2016
[48] V. Arayamparambil Vinaya Mohanan, I.M. Mareels, R.J. Evans, R.R. Kolluri,
IEEE International Conference on Power Electronics, Drives and Energy Systems
Stabilising influence of a synchronous condenser in low inertia networks, IET
(PEDES), IEEE, 2016, pp. 1–4, https://doi.org/10.1109/PEDES.2016.7914468.
Gener. Trans. Distrib. 14 (17) (2020) 3582–3593, https://doi.org/10.1049/iet-
[22] P. Verma, T. Kaur, R. Kaur, Cost Assessment of Deloaded Photovoltaic Systems, in:
gtd.2020.0178.
IOP Conference Series: Materials Science and Engineering 1033, IOP Publishing,
[49] C. Zhang, Y.-.L. Wei, P.-.F. Cao, M.-.C. Lin, Energy storage system: current studies
2021, 012001, https://doi.org/10.1088/1757-899X/1033/1/012001.
on batteries and power condition system, Renew. Sustain. Energy Rev. 82 (2018)
[23] S.G. Malla, C.N. Bhende, Enhanced operation of stand-alone “Photovoltaic-Diesel
3091–3106, https://doi.org/10.1016/j.rser.2017.10.030.
Generator-Battery” system, Electr. Power Syst. Res. 107 (2014) 250–257, https://
doi.org/10.1016/j.epsr.2013.10.009.
17