Integration of Solar Photovoltaic Plant in the Eastern Sumba Microgrid Using Unit Commitment Optimization

Abstract

Integrating renewable energy sources (RES) into island microgrids is usually done to provide a cost-effective electricity supply. The integration process is carried out by scheduling generating unit operations with a unit commitment (UC) scheme to ensure low system operating costs. This article discusses developing a UC optimization method for integrating solar photovoltaic plants in Indonesia’s Eastern Sumba microgrid power system. The scope of this study is the optimization algorithm of the UC, which consists of a priority list (PL) for the UC stage and an economic dispatch (ED) that relies on a genetic algorithm (GA) to minimize total operating costs (TOC). The results show that the PL-GA algorithm performs better than the extended priority list (EPL), and combinations of genetic algorithm and Lagrange, by applying continuous problem dispatch and improved binary GA hourly dispatch to meet ramping constraints. The application of RES incentive programs, such as carbon taxes and incentives for RES generation in calculating the TOC, shows an improvement in the financial feasibility analysis of the internal rate of return (IRR) and net present value (NPV) of actual projects in Indonesia.

1. Introduction

The integration of RES into the existing power system is one of the solutions for providing sustainable electricity to communities located on islands and in remote areas [1,2]. With the increasing intention to develop electricity for islands and remote areas, RES development has rapidly expanded, requiring specific system integration mechanisms. Due to the variability characteristics of solar photovoltaic (PV) plants, where the solar radiation can drop abruptly, there are concerns about how its integration may significantly affect the operation of existing power systems. The same is valid for wind production, but this is not considered in this paper, based on the case study’s local resources. Many countries have launched financial incentive programs to promote the adoption of RES. In 2018, Vu Ba Hau et al. determined the financial benefit for microgrids of the incentives and tax benefits regarding integrating RE generation [3]. This initiative is further supported by the 2022 Glasgow Climate Pact, which set out a commitment to implement programs to support the implementation of RES via enforcing incentives, tax deductions, and tax benefits [4,5].

In the short to medium term, one of the methods for integrating RES generation is UC [6]. The UC method manages the operation of the electric power system by determining the operating schedule and electricity production of each generating unit in the electric power system, including RES, to meet load requirements (electricity demand) with minimum cost or following other criteria [7,8]. The UC must meet several system constraints in operation, such as the power balance, the spinning reserve (SR), and the power limit. The UC method requires optimization techniques to meet load requirements with minimum costs, followed by the ED [9].

In 2018, Nemati et al. conducted a UC and ED optimization using a real-coded GA-based PL algorithm (PL-GA), using Mixed-Integer Linear Programming (MILP) for handling network topology constraints. The algorithm lead to a 40% improvement in total operation cost (TOC) compared to a random method with various convergence times [6]. In 2019, Sarjiya et al. also explored the hybrid PL-GA optimization technique. The PL-GA generated the initial population and continued with the individual evaluation process for the UC stage, and the ED stage was solved with Lagrangian [10]. The results indicate an improvement in TOC of 8% with a long convergence time. In 2021, Rendroyoko et al. used the PL-GA algorithm for the UC stage and the PL for the ED, where the initial population is obtained from the best chromosome (inherited) from the PL process and the next from a random collection. Another ED stage using GA was proposed to reach the lowest TOC [11].

Several methods have thus been tested to solve UC problems, such as Particle Swarm Optimization (PSO) and genetic algorithm (GA), PL-GA, MILP, and a combination of PL and GA algorithms [6,10,11,12,13]. Meta-heuristic algorithms (GA or PSO) usually converge to a similar optimal solution [14,15]. Hybrid optimization algorithms can provide a UC solution with fairly accurate operation costs, with some weaknesses in the GA problem related to speed and convergence to a local optimum. Based on the literature, this research relies on a PL-GA algorithm with a combination technique that improves computational accuracy and convergence speed [16].

UC schemes for integrating solar PV plants may reduce the cost of electricity. However, to make a microgrid system feasible, a techno-economic study is required to determine the feasibility of investing in the integration of solar PV plants in the first place. From an economic perspective, one of the challenges in integrating solar PV plants is the high investment capital of the battery energy storage system (BESS), which will burden the economic feasibility calculation. One way to overcome this problem is to reduce the cost of installing BESS by implementing incentives to support the implementation of RES technology.

This research uses the PL-GA UC algorithm to optimize the integration of solar PV plants in a microgrid power system to meet electricity requirements with minimum cost. The microgrid system in this research consists of diesel engines, a solar PV plant, an electric load, and an energy storage system. The optimization algorithm is a hybrid of the PL and GA techniques. The PL technique is used in the initial stages of UC to get the optimal scheduling for the initial population. The combination of PL and GA is used later for the ED to obtain the lowest operational costs. The PL-GA algorithm is integrated with the BESS optimization to minimize electrical load variability. Several constraints have been implemented in this algorithm, enabling GA to avoid the trap of local optima and speed up the convergence. For this purpose, a simulation model of a microgrid system was developed based on the Eastern Sumba power system, Sumba Island, Indonesia.

This paper is structured as follows: the concept of UC for microgrid systems is described in Section 2, including RES conditioning and optimization algorithms for UC. Section 3 presents the actual UC case study in the Eastern Sumba microgrid system. The UC procedure based on the PL and GA algorithms is described in Section 4. The simulation results are explained in Section 5, along with an analysis of computation time, reserve requirements, and operating costs. Discussions of simulation results are presented in Section 6. Conclusions are provided in Section 7. The detailed abbreviations and definitions used in the paper are listed in Abbreviations.

2. Problem Description

2.1. UC for Microgrid Systems

UC is a cornerstone planning technique for power systems, namely determining the schedule for the power production for each generating unit in the system to fulfill electricity requirements at the minimum cost. Optimization of the UC implementation is subject to several system constraints, such as power balance, SR, and other individual constraints, such as unit output limits, maximum ramp rates, and minimum up/down hours. The UC method in microgrids with intermittent RES production, as implemented in this research, is shown in Figure 1 [11].

In microgrid power systems, the main objective of UC is to minimize the TOC [4].

$$min \mathrm{T}\mathrm{O}\mathrm{C}=\sum _{t=1}^T\sum _{i=1}^I(C_{{SU}_i}{v}_{t,i}+f_{t,i})$$

(1)

$C_{{SU}_i}$ and $v_{t,i}$ denote the startup cost of unit i and startup status of unit i at time t, respectively, and $f_{t, i}$ denotes the fuel cost of unit i at time t. The fuel cost function $f_{t,i}$ used in this paper is using a quadratic form.

$$f_{t,i}=a_i{p}_{t,i}^2+b_i{p}_{t,i}+c_i{u}_{t,i},$$

(2)

where $a_i$, $b_i,$ and $c_i$ indicate the coefficients for fuel cost formulation, and $u_{t,i}$, and $p_{t,i}$ denote the generating unit’s status and dispatched power of unit i at time t, respectively.

The objectives of (1) and (2) are subject to power systems constraints.

(a)

Power balance

The total power output must satisfy the electricity system’s demand requirement.

$${\sum }_{i=1}^I{p}_{t,i}=D_t.$$

(3)

$D_t$ denotes the electricity system’s demand requirement at time t.

(b)

Spinning reserve (SR)

The total generation must be greater or equal to the load plus the SR needed to guarantee the stability of the power system. The total SR must be larger than 10% of the demand.

$${\sum }_{t=1}^T{\sum }_{i=1}^I(u_{t,i}{P}_{{max}_i}-p_{t,i})\ge 10\%D_t.$$

(4)

$P_{{max}_i}$ denotes the maximum power output of generating unit i.

(c)

Power generation limit

The active power generated has maximum and minimum limits:

$$L_i.V_{t,n}\le P_{t,i}\le U_i.V_{t, i},$$

(5)

$$P_{{min}_i}\le P_{it}\le P_{{maks}_i}, P\in R.$$

(6)

$P_{{min}_i}$ and $P_{{max}_i}$ denote the minimum and maximum power output of generating unit i.

(d)

Minimum up/downtime

The generating unit must meet the minimum on/off time before being allowed to change status, as follows.

$${\sum }_{\tau =t-T_{{up}_i}}^{t-1}{u}_{\tau ,i}\ge T_{{up}_i{w}_{t,i}} \forall t\ge T_{{up}_i},$$

(7)

$$T_{{up}_{0,i}}+{\sum }_{\tau =1}^{t-1}{u}_{t,i}\ge T_{{up}_i}{w}_{t,i}, \forall t<T_{{up}_i},$$

(8)

$${\sum }_{\tau =t-T_{{down}_i}}^{t-1}(1-u_{t,i})\ge T_{{down}_i}{v}_{t,i}, \forall t<T_{{down}_i},$$

(9)

$$T_{{down}_{0,i}}+{\sum }_{\tau =t}^{t-1}(1-u_{t,i})\ge T_{{down}_i}{v}_{t,i}, \forall t<T_{{down}_i}.$$

(10)

$T_{{up}_i}$ and $T_{{down}_i}$ denote the minimum up-time and downtime of generating unit i, respectively, and $T_{{up}_{0,i}}$ and $T_{{down}_{0,i}}$ indicate the initial up-time and downtime counter of generating unit i, and $v_{t,i}$ and $w_{t,i}$ denotes the startup and shutdown status of unit i at time t, respectively, and τ denotes a time index like t.

(e)

Ramp rate

The ramp rate constraint is the generating unit’s limit to increase (ramp-up) or decrease (ramp-down) the power dispatch output between time t. The constraints are formulated as follows:

$${-P}_{{rd}_i}\le p_{t,i}-p_{t-1,i}\le P_{{ru}_i}.$$

(11)

$P_{{ru}_i}$ and ${-P}_{{rd}_i}$ denote the ramp-up and ramp-down limit of generating unit i.

(f)

Must-run unit

According to the forecasted output, RES power plants are treated as must-run units and directly calculated into the power supply mechanism for the power system. Together with power load demand, RES formulates the net load as follows:

$$P_{{nett}_t}=D_t-P_{{re}_t}.$$

(12)

$P_{{re}_t}$ denotes the ramp-up and ramp-down limit of generating unit i

(g)

Distribution system power flow

In calculating the TOC, each dispatch operation of the generating unit must consider the distribution power flow. The power flow here is solved using the Newton–Raphson method [17].

$$I_i={\sum }_{k=1}^n{Y}_{ik}{V}_k i=1, 2, \dots n,$$

(13)

$$P_i=\left|V_i\right|{\sum }_{k=1}^n\left|V_k\right|\left|Y_{ik}\right|cos({\theta }_{ik}+{\delta }_k-{\delta }_i),$$

(14)

$$Q_i=-\left|V_i\right|{\sum }_{k=1}^n\left|V_k\right|\left|Y_{ik}\right| sin({\theta }_{ik}+{\delta }_k-{\delta }_i).$$

(15)

$Y_{ik}$ denotes the Y matrix, which refers to a matrix of admittance parameters, describing an electrical network viewed as a black box with ports.

2.2. Microgrid Power System, with Solar Photovoltaic and BESS

As with any power system, microgrids use multiple distributed power sources and operate the power supply while maintaining a balance between demand and supply [18]. The microgrid usually includes several distributed generation sources such as diesel engines, micro-hydro, wind turbines, and PV panels [19]. At the point of implementation, the system is developed by integrating existing generating systems, usually diesel generators in microgrids like the studied use case in Indonesia, with some PV plants. To ensure continuity of electricity supply, a microgrid system is typically also equipped with a BESS [20].

In preparation for the implementation of the UC, the RES power plant must also be considered according to its availability. Weather forecasting, historical data, and costs are critical in preparing the generation schedule for intermittent RES. With accurate weather forecasts, the UC of solar PV can be performed accurately. PV panels convert solar energy into electricity, and their power output depends on the solar irradiance G and is affected by the ambient temperature. Equation (16) presents the production of solar PV.

$$P_{PV}=\left\{\begin{array}{l}\hspace{1em}\hspace{1em}0, G < G_i\\ \hspace{1em}\mu .G, G_{in}\le G < G_{rate}\\ \hspace{1em}{P}_{rate}, G \ge G_{rate}\end{array}\right.$$

(16)

$P_{PV}$ is the power output of solar PV, $G_{in}$ and $G_{rate}$ express the minimum and rated insolation, and μ is the coefficient of the solar panel.

The BESS has a critical position in the microgrid, especially with RES production, by providing load leveling and contingency operational benefits. BESS acts as a power source when the power output from the power plant and the RES cannot meet the load requirements. The power expresses the energy stored (ES) in the battery charged P₊, and the power released P₋ over a single time step in an hour.

$${ES}_{t+1}={ES}_t+\left(P_{t+}-P_{t-}\right)\ast {∆}_t$$

(17)

t denotes time in hours. The P denotes power in MW. The time step and ${∆}_t$ used in this paper is one hour. The ${ES}_{t+1}$ is the energy stored in a single step ahead, $P_{t+}$ is the charged power, $P_{t-}$ is the discharged power, and the storage battery has a charging capacity limit and should not be over-charged. Here, $B_m$ is the battery capacity.

$$ES \le B_m.$$

(18)

2.3. Demand Response

The next effort to mitigate system operating costs is to shift peak loads to reduce the number of operating machines. In this research, demand response (DR) proposes a peak load shifting mechanism from one peak load period to another. Here, the input parameters of the GA for DR operation are entered where the allocation of transfer of some of the load is the result of GA optimization, which has a maximum limit of shifting power and time constraints. The shifted load is formulated as follows:

$$P_{sft}=P_{str} \le P_{drmax}.$$

(19)

$P_{sft}$, $P_{str}$, and $P_{drmax}$ denote load shifted from time t to time τ, and the maximum shifted power limit.

With all PV, BESS, and DR, the final net load demand power $D_{nett}$ can be formulated as follows:

$$D_{nett}=D_t-P_{pvt}+P_{+t}-P_{-t}-P_{sft}+P_{{st}_{\tau }}.$$

(20)

2.4. Financial Feasibility of Microgrid System Integration

In evaluating financial feasibility to obtain benefits and advantages in implementing microgrid integration, the metrics used in this study to assess financial feasibility are NPV, IRR, and Simple payback (SPB) period. The NPV is calculated by subtracting the present value of the cash outflows from the current value of cash inflows over the life of the investment, as follows [21]:

$$NPV=C_0+\frac{C_1}{1+i}$$

(21)

$C_1$ is the payoff, and $C_0$ is the required investment, and i is the discount rate.

The IRR is the discount rate, which makes NPV = 0. The formula used to calculate the IRR for an investment project lasting for t years is as follows:

$$NPV=C_0+\frac{C_1}{1+IRR}+\frac{C_2}{{(1+IRR)}^2}+\cdots +\frac{C_t}{{(1+IRR)}^t}$$

(22)

The SPB period estimates how long a project will take to generate sufficient cash flow to pay back its initial costs.

3. Case Study: Eastern Sumba Microgrid System, Indonesia

This study aims to develop and implement a PL-GA optimization algorithm in the UC scheme for integrating a PV solar power plant in the Eastern Sumba system, Sumba Island, East Nusa Tenggara province, Indonesia [22]. Sumba Island is an iconic island for RES initiatives, as the electricity supply still consists of several distributed microgrid systems [15]. This microgrid system was chosen as a case study because it is a fast-growing electric power system with a high potential for developing solar photovoltaic plants. This microgrid system consists of 17 diesel engines (200 kW–1 MW) and one solar PV plant with a 1 MWp capacity with a single line diagram (SLD), as shown in Figure 2.

The details of the parameters of each generating unit of the SLD of the Eastern Sumba microgrid system are shown in

Table 1. Parameter of generator units of Eastern Sumba system.

Power Plant	Pmax (MW)	Pmin (MW)	a ($/h)	b ($/MWh)	c ($/MWh2)	MUT (h)	MDT (h)	HSC ($)	CSC ($)	CS (h)	IS (h)	SFC (lt/kWh)
D1	0.56	0.2	116.2	84.82	70.2	1	0.5	125	220	0	−1	0.298
D2	0.5	0.2	91.5	72.5	68.59	1	0.5	150	200	0	−1	0.298
D3	1	0.4	37.5	47.5	42.5	1	0.5	150	200	0	−2	0.303
D4	0.4	0.2	289	95.4	65.88	1	0.5	150	250	0	0	0.288
D5	0.4	0.2	248	98.5	76.22	1	0.5	150	300	0	−5	0.535
D6	0.63	0.4	76.6	92.2	65.19	5	1	85	185	0	1	0.298
D7	0.63	0.44	63.5	69.5	58.68	1	0.5	125	220	0	−1	0.298
D8	0.63	0.44	67.6	72.8	54.92	1	0.5	150	200	0	−1	0.298
D9	0.63	0.44	62.6	68.2	59.84	1	0.5	150	200	0	−2	0.303
D10	0.63	0.44	62.5	74.4	56.09	1	0.5	150	250	0	0	0.288
D11	0.63	0.44	61.25	70.8	58.44	1	0.5	150	300	0	−5	0.535
D12	0.63	0.44	62.8	70.4	58.3	4	2	50	105	0	−1	0.303
D13	0.63	0.44	58.4	77.8	55.84	4	2	51.5	100	0	−2	0.275
D14	0.63	0.44	63.6	81.2	51.15	4	2	50.5	102	0	−2	0.275
D15	0.63	0.44	56.5	75.6	57.66	4	2	52	102	0	−2	0.275
D16	0.63	0.44	62	76	55.05	4	2	50	102	0	−2	0.275
D17	0.63	0.44	60	76.8	55.98	4	2	50	105	0	−2	0.275

. The Eastern Sumba microgrid system is a 20 kV medium-voltage power system with three substations. First, there is the Kambajawa power plant, which is the center for diesel power generation; second, the 20 kV Waingapu substation, which has distribution installations for customer loads and backup generating units; and third, the Humbapraing substation, which has solar PV. A 20 kV distribution line connects each 20 kV substation. The fuel cost coefficients, cost characteristics, startup costs, minimum uptime (MUT), minimum downtime (MDT), and other parameters are given in Table 1.

This study develops a power system simulation model using data from the Eastern Sumba system, an extract of which is proposed in Figure 3. The following is the monthly load profile data for the year 2020 of the microgrid system. The April load profile data is selected for this paper’s simulation model. The load profile data for the system displays the typical characteristics of electricity demand in developing regions in eastern Indonesia, with a peak load of 8.30 MW occurring around 07.00–08:00 p.m. and a large gap with the minimum load consumption. The lowest average load of the Eastern Sumba system is 4.32 MW, which occurs at around 2.00 p.m.

The forecasted data of a solar PV plant’s average hourly output power profile is used in the simulation model, as shown in Figure 4.

The operational control aims to fully serve the electricity load during the day by diesel generators, solar PV, and batteries. The diesel engine unit is operated to serve the load throughout the night, and the battery helps during transition times and peak loads.

By referring to the above operational planning and using the UC method, the electricity load requirements are expected to be served at a minimum cost while maintaining system stability.

4. The PL-GA Algorithm

The priority list (PL) method is simple and fast but produces suboptimal solutions [23,24]. The significance of the PL relies upon committing generation units based on the order of increased operating cost, such that the lowest cost units are first selected until the load is satisfied. The unique feature of the PL method has attracted continuous development and improvement by combining it with other methods to improve its performance. GA is a general-purpose search method inspired by the principles of genetics and evolutionary mechanisms in nature [25]. The basic principle is the process of a population of solutions to a problem (genotypes) in the form of encoded information individuals that evolve.

This research proposes a combination of PL and GA-based algorithms for the UC problem. To solve the UC problem, the PL method is carried out early to arrange the generator units’ operation scheduling and obtain the initial population. The ED stage follows to obtain the lowest TOC. At the initial population stage of the GA, one individual uses the dispatch value from the PL to accelerate the convergence.

The UC problem can be simplified using the net load concept in a power system with intermittent RES. The net load is the difference between load consumption and the power generated by RES generators, which can be generated by thermal or diesel-generating units [26,27]. The operating schedule for each generating unit is prepared to meet the net load by considering the forecasted solar power and the predicted hourly electrical load profile. Figure 5 presents the flowchart of the proposed hybrid PL-GA algorithm. The main steps can be explained as follows:

(1)

The improved PL algorithm selects generating units ready to operate with sufficient power capacity and the best heat-rate (HR) value to obtain minimum operating costs at the UC stage. The priority order list is calculated based on each unit parameter related to higher capacity and better HR values. Then, the fitness of each chromosome is evaluated in the population within the constraints. The setting of upper and lower limits based on ramp up/down (11) and minimum up/downtime is checked for each chromosome to meet load requirements.

(2)

The next stage is the calculation of the ED using GA. The scheduling plan acquired by the PL method replaces one of GA’s initial populations, and the others are taken from the random generation of individuals. The chromosomal evaluation checks every chromosome in the population for conformance to the required constraints. SR and minimum up/downtime are checked for each chromosome, and a penalty is imposed for each violation. The number of penalty values is used for initiating fitness or ranking conditions as an initial reference for parent selection.

(3)

The optimization process continues with the selection process: parent selection, crossover, mutation, fitness checks, and elitism. Parent selection is done using a roulette-wheel mechanism on the fitness value condition. The next step is to do a 2-point crossover to obtain children with better characteristics and continue with the mutation step with a specific probability rate [10]. Next, proceed with the elitism stage, which is the stage of getting the best individuals. This can be done by removing the two worst individuals to be replaced by two new, better offspring [11]. Using this elitism process, GA can easily find the best combination of chromosomes from the population with the lowest TOC and still comply with all required constraints.

BESS is one solution to RES variability, especially in microgrids operated with a solar PV installation. The main objective here is to minimize operating costs by adjusting the charging and discharging mechanism to reduce the variability of the electric power load. A GA technique is used with the process steps to determine the best combination of charge and discharge, as shown in Figure 6.

In this paper, the authors also inject some individuals with pre-determined strategies besides random initial populations. Those strategies are:

• No charge, no discharge: This works as the base strategy when the battery is not operating.
• Random no discharge works as a strategy where the charge schedule is randomized, but the discharge strategy is left to the backward fixing part in the GA iteration.
• Full charge, no discharge: the battery charges as much as possible, but the discharge strategy is left to the backward fixing part in the GA iteration.
• Custom expert judgment is used as an initial custom population based on the judgment of the experience of an expert.
• Max charge and max discharge: the initial population has maximum charge during daytime and full discharge during nighttime.
• Full random charge and discharge schedule is the default for other individuals.

The microgrid system operates the BESS to minimize generation operating costs with the main criterion of minimizing the variability of electrical power loads. From the point of view of reducing load variability, it is expected that fuel consumption will be reduced by minimizing the operation of the diesel engine. Minimizing load variability means minimizing the TOC.

The BESS operation schedule is also optimized using a GA, together with UC and ED, to solve the problem of the day-ahead UC schedule. The solar PV output is always at maximum power output mode using Maximum Power Point Tracking (MPPT) and is always absorbed by the grid in a must-run unit scheme.

This simulation uses an Octave programming language on JupyterLab as the user interface (UI). The computer is powered with two DDR3 Kingston KHX1600C9D3/4GX @ 4096 MB (Kingston Technology, Fountain Valley, CA, USA) and an eight-core Intel^® Core (TM) i7-3770 CPU @ 3.40 GHz (Intel Corporation, Santa Clara, CA, USA).

5. Simulation Results

This study compares the PL-GA algorithm with the PL algorithm and the GA and Lagrange algorithm combination. The PL algorithm is a classical priority list algorithm, using GA for the ED [10].

A microgrid system simulation model has been prepared to test the proposed UC algorithm. The simulated microgrid consists of dispatchable diesel engines and a solar PV plant, including the BESS system. The simulation begins with calculating the net load to simplify the formulation of the UC problem [26]. Due to the intermittent supply of energy, it is necessary to determine a schedule that meets the net load in advance. Each scenario considers hourly random load and solar power based on the predicted values. The net load is shown in Figure 7, which results from the optimization calculation by the steps to minimize load variability. The BESS mechanism makes the system load variability profile smaller and improves the plant’s operating effectiveness.

Figure 7 shows the flatter load curve resulting from a small sigma variation. This condition indicates that the battery operation control of charging and discharging is running adequately. The BESS charging and discharging mechanism is shown in Figure 8.

Figure 8a shows the solar PV and battery (in MW), the respective limits for the maximum charge rate and discharge rate, and Figure 8b shows the stored energy profile (MWh), the maximum and minimum limits of stored energy. When solar irradiance is available, the PV plant generates output power, simultaneously increasing the battery’s stored energy, and contributing to the power supply of the microgrid system to fulfill system load requirements.

Once the mechanism for regulating the charging and discharging of the energy storage system is determined, the next stage is implementing the PL-GA. The implementation of this algorithm is intended to manage the operating schedule of the generating units to meet the net load electricity requirement on the Eastern Sumba microgrid system, as shown in Figure 9. The success of regulating the operation of these generating units indicates that the PL-GA algorithm sufficiently restricts the system-generating units’ scheduling to meet load requirements with minimum cost.

In Figure 9, the negative number at the 2nd hour is the shifting demand response from the 19th hour. At the same time, the negative number at hours 12, 13, and 15 is the amount of battery charging at the possible PV output conditions.

In this research, several parameters have been tested manually, and specific parameters have been selected to get the best results. In this simulation, the generator data are taken from

Table 2. Comparison of TOC.

Method	Total Fuel Cost ($K)	Average TOC (1000$)	Elapsed Time (s)
TOC	29.99	31.57	-
GA and Lagrange	28.25	29.67	630
PL	29.76	31.19	4.6
PL-GA	29.31	30.74	39.6

, with the SR set at 25% of the load, PV penetration at 10%, a population size of 100, a maximum iteration value of 500, a crossover probability value of 1.0, and a mutation probability of 0.8. The process converges in around 150 iterations with this set of parameters, as shown in Figure 10.

Figure 10 shows that the PL-GA method performed intrestingly while providing close to optimal solutions. Comparative analysis was carried out between the simulation results with other methods and the TOC realization (as a baseline) [28]. Here, the PL-GA algorithm is compared to the simulation model using the PL algorithm and GA and Lagrange algorithm, as well as with the realization of the TOC of the Eastern Sumba microgrid system, as shown in Table 2.

Compared with the GA and Lagrange methods, the computational results of the PL-GA algorithm are slightly higher, but the runtime is much shorter. Conversely, the proposed PL-GA algorithm gives better results than the PL algorithm. Indeed, the PL method is fast, but the solution sometimes could be more optimal.

The PL-GA algorithm is a relevant technique for implementing the UC method for microgrid power systems with the RE generator by providing accurate results in a brief time. The calculations then serve as a reference for the feasibility study of the energy storage battery investment and DR implementation.

The results of efficiency calculations due to the installation of energy storage systems, as shown in

Table 3. TOC and CCR for the Eastern Sumba Microgrid System.

No.	Configuration of Microgrid	TOC	Remarks
1.	Diesel + Solar PV	0.183	Baseline
2.	Diesel + Solar PV + BESS	0.155
3.	Diesel + Solar PV + BESS + DR	0.154
4.	Capital cost recovery (CCR)	0.066	Simulation IRR = 12.09%

, are used as the basis for the financial feasibility analysis.

The calculation of operational costs by considering the installation of energy storage batteries and the DR mechanism is proposed in

Table 4. Investment Feasibility Analysis of Microgrid Integration with Carbon Tax and VAT Incentives.

Conditional Scenario	Feasibility Analysis			Remarks
	IRR (%)	NPV ($)	SPB (Year)
Diesel + Solar PV	12.04	99,171.00	6.63
Diesel + Solar PV + BESS	14.72	179,728.05	6.06	VAT incentive (9.95%)
Diesel + Solar PV + BESS + DR	12.30	108,592.03	6.57	Carbon tax 0.2 c$/kg-CO2
Capital cost recovery (CCR)	14.99	189,149.20	6.01	VAT incentive (9.95%) and carbon tax 0.2 c$/kg-CO2

. The results show a decrease of the TOC from 0.183 $/kWh to 0.154 $/kWh or savings of 0.029 $/kWh. However, these results must be improved to meet battery investment’s cost recovery needs. Therefore, funding support from the utility company or the government is required to provide incentives. The incentive support can be achieved by referring to the Indonesian Government Energy Law No. 30 of 2007 and the result of the Glasgow Climate Pact in the form of providing carbon tax incentives, which is stated in the Harmonization of Tax Regulations of the Republic of Indonesia and the regulations which support the use of RES power plants [4,5]. The results of calculations on the financial feasibility analysis of the implementation of carbon tax incentives and support for the use of RES in VAT reductions are also shown in Table 4.

6. Discussion

The results of operational cost optimization calculations using the PL-GA algorithm have been obtained for the East Sumba microgrid system (Supplementary Materials). To solve the UC problem, the PL method is carried out early to arrange the generator units’ operation scheduling and get the initial population. This is followed by the ED stage to obtain the lowest TOC. At the initial population stage of the GA, one individual uses the dispatch value from the PL to accelerate the convergence. The PL-GA optimization algorithm for binary dispatch has been successfully developed to provide optimal results with fast convergence speed with various solutions for the application of UC in microgrid power systems with Solar PV generation.

The UC problem can be simplified using the net load concept in a power system with intermittent RES and the demand response (DR) implementation. The DR scheme here is implemented to reduce system netload variability to reduce operating costs. The net load is the difference between load consumption (with DR mechanism) and the power generated by RES generators, which are generated by diesel-generating units. The operating schedule for each generating unit is prepared to meet the net load by considering the predicted solar power and the predicted hourly electrical net load profile.

The contribution of the battery energy storage system (BESS) in regulating the battery operation control of the charging and discharging mechanism, which produces the flatter netload curve resulting from a small sigma variation, has also been discussed.

Finally, the multi-state optimization technique consists of Netload, DR mechanism, BESS optimization, and PL-GA algorithm, which can optimize the UC scheme on integrating Solar PV paired with BESS with the existing power system. Simulations on East Sumba microgrid systems show that the PL-GA algorithm produces a TOC calculation relatively similar to the previous GA and Lagrange algorithm, with a much better convergence time.

In the financial analysis, several levels of incentives impact the financial parameter index of the microgrid investment to get the optimal configuration. BESS technology’s project financial parameters and capital costs can vary widely depending on the status of the local economy, policies, and government support. The greater the value of the incentives, the more profitable the project, with higher NPV and IRR and shorter SPB periods. Table 3 shows that providing carbon tax incentives and tax relief support can help significantly reduce the results of electricity supply operating costs that have resulted from implementing the PL-GA optimization algorithm on the electric power microgrid system.

Implementing the PL-GA optimization algorithm in the UC scheme of the electric power system can have a positive effect with a significant reduction in operating costs. By implementing incentives and tax discount support policies, the investment in solar PV with BESS can be financially feasible.

Thus, utilities may implement the PL-GA optimization algorithm combination on a microgrid system with a photovoltaic generator, supported by a policy of providing incentives and tax relief support. This can be one of the primary references for the program implementation and operation of new RES plants in all island microgrid systems and remote areas, notably in Indonesia.

7. Conclusions

In this paper, we propose a UC solution using an optimization algorithm from the improved priority list genetic algorithm (PL-GA) for optimizing the integration of solar PV generators into microgrid power systems so that electricity requirements are fulfilled at minimum cost, thus integrating an economic dispatch stage. The priority list (PL) is a fast optimization technique and GA can solve optimization problems simultaneously and has a global search technique to overcome variability and uncertainty. Research in this field is urgently required to support the microgrid RES integration program for islands and rural areas, and the existing optimization methods still have limitations and require improvement.

The PL algorithm and GA are applied to a case study of an island in Indonesia, Eastern Sumba, to obtain relevant optimization results for implementing the UC. The PL-GA algorithm performs well, finding a low TOC with a short computation time compared to the extended PL and the hybrid algorithm that consists of a genetic algorithm and Lagrange relaxation optimization methods. Considering these results, combining PL-GA optimization algorithms suits a microgrid system with RES. This algorithm can be practically implemented in microgrid intelligent control systems.