Optimization of ESS Scheduling for Cost Reduction in Commercial and Industry Customers in Korea

Abstract

Various attempts have been made to reduce carbon emissions in the energy sector as part of global net zero emissions trends. Among them, interest in the use of energy storage systems (ESSs) for energy efficiency is growing. Utilities intend to improve the efficiency of investment and operating costs by reducing the maximum peak and leveling the load. Many ESS-scheduling optimization techniques have been studied to reduce the peak demand, balance the load, or reduce the cost corresponding to these two purposes from the customer’s point of view. In this paper, a method for cost minimization that simultaneously considers both the peak demand and load balancing is proposed, and the results and analysis of a case study in Korea Electric Power Corporation (KEPCO), Korea are presented. Through these results, we show that there is a priority among the objective functions of the ESS schedule, that demand charge is more important than energy charge, and that the ESS schedule problem for customers to reduce costs is also beneficial to power system operation by the utility’s rate policy.

1. Introduction

To solve global warming and environmental problems, numerous efforts have been made to be carbon neutral globally. Because the proportion of carbon emissions in the energy field is also significant [1], various efforts have been made, such as new and renewable power generation, the prevalence of electric vehicles, and the efficient management of building energy [2,3,4,5]. These efforts to contend with carbon neutrality are reducing carbon emissions from power plants by switching to renewable power sources, such as solar and wind power, as well as improving the efficiency of energy consumed to reduce overall energy consumption and thereby mitigating carbon emissions.

Technically, an energy storage system (ESS) is sometimes used to compensate for intermittent power generation outputs, such as solar or wind power. An ESS is also used to improve the overall energy efficiency by charging the energy storage device during a time when the power load is low and discharging the power stored in the device during the peak load time [6,7,8]. From the customer’s point of view, they will be more interested in reducing energy costs, even though ESSs can be used for various purposes [9,10]. Accordingly, we intend to present a solution on how to use ESSs to reduce electricity costs most efficiently. When an ESS is utilized in terms of power load, the ESS must be charged when the load is low (mainly at night), and then, the power charged in the ESS must be discharged when the load is high (mainly during the day) to reduce the peak load. Conversely, in terms of electricity cost, the ESS must be charged during the period of low unit rate (mainly at night) and discharged during the period of high unit rate (mainly during the day) to reduce the electricity cost. Therefore, an efficient method of ESS usage is to apply scheduling for the ESS with the strategy described above.

ESS scheduling is used to solve various types of problems, as follows. From the point of view of a distribution network operator, there is an approach to find the optimal ESS schedule to reduce the peak load of the distribution line or for system stabilization through load leveling [11,12,13,14,15,16]. To avoid the risk of differences in energy storage transactions, an ESS-scheduling method based on average variance optimization that considers the price uncertainty of the day ahead and real-time energy markets was proposed [17]. There has also been an attempt to reduce the total power loss through ESS optimal scheduling by considering the large-scale integration of new and renewable resources, such as solar power (PV) and wind turbines in the active distribution network [18]. Related to a microgrid under dynamic charging, a method to provide a day-ahead scheduling strategy for microgrid ESS has been proposed to minimize the cost paid by household customers by utilizing commonly available ESSs in the microgrid [19]. From the operation of battery ESSs (BESSs) under dynamic rates, a multipurpose optimization model was designed to balance the goals of minimizing household electricity costs and CO₂ emissions [20].

Most research addresses the problem of reducing the maximum load and minimizing the electricity cost based on the unit rate separately [21,22,23,24]. AI-based approaches are used to reduce peak load in buildings, such as reinforcement learning [25], and to optimize the power flow at the household level with LSTM and heuristic algorithms [26]. Eventually, utilities intend to reduce the difference between the maximum load and the minimum load of electricity they supply by linking the customer’s monthly load pattern with the electricity bill. Therefore, the utility charges individual customers a high base rate for the highest peak load over the year and also charges a high unit rate during the peak time for the daily load used every month to reduce the load, and thus, the power supply should be dispersed over several periods. The Korea Electric Power Corporation (KEPCO) also operates this type of tariff system, and customers can choose a system depending on their circumstances, which has different rates for each period and season [27].

In this paper, we attempt to solve the ESS schedule by reducing the peak load year-round from the customer’s point of view and simultaneously minimizing the cost of the daily load used every month. Therefore, a method for cost minimization that simultaneously considers both the peak demand and load balancing is suggested, and the results and analysis of a case study in KEPCO, Korea are presented. Through this result, we show that there is a priority among the objective functions of the ESS schedule, that demand charge is more important than energy charge, and that the ESS-schedule problem for customers to reduce costs is also beneficial to power system operation by the utility’s rate policy.

The contributions of this paper are as follows.

• To optimize the ESS schedule, it is necessary to consider both the peak load year-round and daily load used every month, rather than simply reducing the peak load or balancing the load.
• In addition, a cost-effective methodology that considers both reducing the peak load and balancing the load is presented.
• Although the ESS-schedule problem is aimed at reducing costs from the customer’s point of view, it shows that it is also beneficial to power system operation by the utility’s rate policy.
• We demonstrate that priority exists among the objective functions of the ESS schedule and that demand charges are more important than energy charges.

The remainder of this paper is organized as follows. The list of nomenclature is introduced as shown in

Table 1. Nomenclature.

Notation	Description
$Peak Load\left(kw\right)$	The maximum load without the ESS
$Peak Loa{d}_{ESS}\left(kw\right)$	The peak load reduced by using the ESS
$Loa{d}_t$	The load before using the ESS at time t
$Loa{d}_{ES{S}_t}$	The load after using the ESS at time t
$ES{S}_{SOC}$	The state of charge of the ESS
$ES{S}_{o{p}_t}$	The operation of the ESS at time t
$O{P}_{ESS}$	The operation in the ESS
$ES{S}_{SO{C}_t}$	The state of charge of the ESS at time t
$ES{S}_{charg{e}_t}$	The capacity that the ESS does charge at time t
$PC{S}_{cp}$	The maximum capacity of power conversion system (PCS)
$ES{S}_{discharg{e}_t}$	The capacity that the ESS does discharge at time t

. Section 2 explains how to define the problem, and Section 3 describes the ESS-scheduling framework to solve the defined problem. Section 4 analyzes the results by applying the framework presented in Section 3, focusing on the case of Korea’s TOU electricity tariff system. Section 5 concludes with the contributions of this paper and future tasks.

2. Problem Definition

A customer who owns an ESS hopes to reduce electricity bills by charging electricity in the ESS when electricity rates are low and using the electricity stored in the ESS when electricity rates are high. Meanwhile, the utility hopes that the customer’s ESS operation can help stabilize the system by reducing the peak load or contributing to load leveling, thereby preventing the overinvestment of the facility and operating it efficiently. Therefore, the utility establishes a tariff system policy such that the customer’s ESS operation helps the system stabilization, thereby reducing the electricity bills from the customer’s perspective and leveraging it for efficient system operation from the utility perspective.

Similarly, KEPCO operates various tariff systems, such as the basic rate peak linkage system and the demand management optional tariff (i.e., critical peak pricing, CPP). The characteristic of these pricing plans is that the customer’s maximum peak year-round is linked to the basic rate, and the actual usage rate is applied by the hourly unit price. Subsequently, it is possible to decrease the peak load and lead the customer’s usage pattern simultaneously such that the load is equalized over different periods. For example, regarding the General Service (A) II tariff system shown in

Table 2. General Service (A) II price table (from KEPCO in Korea).

Classification		Demand Charge (won/kW)	Energy Charge (won/kWh)
			Time Period	Summer (1 June–31 August)	Spring/Fall (1 March–31 May/ 1 September–31 October)	Winter (1 November–28 February)
High- Voltage A	Option I	7170	Off-peak load	57.7	57.7	66.4
			Mid-load	108.9	65.1	96.8
			Peak load	131.4	76.4	111.6
	Option II	8230	Off-peak load	52.4	52.4	61.1
			Mid-load	103.6	59.8	91.5
			Peak load	126.1	71.1	106.3
High- Voltage B	Option I	7170	Off-peak load	57.1	57.1	66.1
			Mid-load	105.7	63	93.4
			Peak load	122.1	68.4	107.6
	Option II	8230	Off-peak load	51.8	51.8	60.8
			Mid-load	100.4	57.7	88.1
			Peak load	116.8	63.1	102.3

, we can observe that it is a differential rate system that has distinguished standard rates depending on the periods and seasons. Therefore, to minimize the electricity bills with respect to customers, there is a need for peak shaving for a basic rate reduction and arbitrage based on the difference in unit price over time.

2.1. Definition of Objective Function

To minimize electricity bills for customers who use a seasonal and time-based differential tariff plan in which pricing standards differ according to periods and seasons, it is necessary to simultaneously achieve two goals: reducing the basic rate and reducing the rate through arbitrage using the difference in unit price over time. Therefore, the objective function’s role in ESS scheduling is to reduce the basic rate using peak shaving and to maximize the amount of arbitrage based on the difference in the unit price for each period. To determine the priority of the objective function, let us consider the tariff structure described above. The tariff structure calculates the total amount by applying the unit price for each season/period to both the basic rate and actual usage. In the case of the basic rate, this month’s basic rate is determined according to the peak load in the previous year. If the peak load exceeds the maximum over this year, the new basic rate increases over the next year. In addition, as we can observe from the tariff system, the basic rate is set relatively more expensive than the actual usage rate. Thus, it can be expected that the higher the peak load, the more expensive the electricity bills that must be paid over a longer period.

On the contrary, in the case of actual usage, the unit price increases with usage; however, this effect is limited only to that month. That is, we find that the utility places more importance on the reduction in the peak load than the load leveling. Therefore, peak shaving should be applied to the objective function, and then, arbitrage is used within the level that does not affect the peak load such that customers’ electricity bills can be reduced the most.

$$Bill=Peak Load\left(kw\right)\times Demand Charge+Consumption\times Energy Charge$$

(1)

$$\mathrm{Objective} \mathrm{Function} 1: Minimize\left(Bill\right)$$

(2)

The electricity bill is calculated by the sum of the amount multiplied by the peak load in the previous year by the demand charge and the amount multiplied by the actual usage by the energy charge, as shown in Equation (1). The objective function is to minimize the calculated electricity bill, as shown in (2).

Then, in the first term constituting the electricity bill,

$$\Delta Peak Load\left(kw\right)=Peak Load\left(kw\right)-Peak Loa{d}_{ESS}\left(kw\right)$$

(3)

In the second term of the bill,

$$\Delta \mathrm{Consumption}={\displaystyle \sum }_{t=0}^{23}\left(Loa{d}_t-Loa{d}_{ES{S}_t}\right)$$

(4)

Therefore,

$$\begin{array}{c}\Delta Bill=\Delta Peak Load\left(kw\right)\times Demand Charge\\ +\Delta Consumption\times Energy Charge\end{array}$$

(5)

In Equation (3), $Peak Load\left(kw\right)$ is the maximum load without the ESS, and $Peak Loa{d}_{ESS}\left(kw\right)$ is the peak load reduced using the ESS for peak shaving. Hence, $\Delta Peak Load\left(kw\right)$ is the difference between the maximum load before and after using the ESS. In Equation (4), $Loa{d}_t$ means the load before using the ESS, and $Loa{d}_{ES{S}_t}$ implies the load after using the ESS. Hence, $\Delta Consumption$ is the difference between the load used before and after using the ESS. In Equation (5), $Demand Charge$ is a fixed value, but $Energy Charge$ changes according to seasons and periods. Because $\Delta Bill$ can be represented as Equation (6), the objective function is minimizing $\Delta Bill$, as shown in (7).

$$\begin{array}{c}\Delta Bill=\left(Peak Load\left(kw\right)-Peak Loa{d}_{ESS}\left(kw\right)\right)\times Demand Charge\\ +{\displaystyle {\displaystyle \sum }_{t=0}^{23}}\left(Loa{d}_t-Loa{d}_{ES{S}_t}\right)\times Energy Charge\end{array}$$

(6)

$$\mathrm{Objective} \mathrm{Function} 2: Minimize\left(\Delta Bill\right)$$

(7)

2.2. Definition of Constraints

Here, we explain some constraints applied when scheduling the charging and discharging of the ESS to achieve the objective function obtained in previous studies. First, the ESS can be operated daily by discharging the amount charged from 0 to 24 every day. In other words, the valid $ES{S}_{SOC}$ must be set to $ES{S}_{SOC}=0$ at 0 o’clock every day. Here, SOC is an abbreviation of the state of charge, which means the state of charge of the ESS. Generally, because full charging and discharging shortens the lifespan of the ESS, it is essential to maintain the minimum amount of charge.

In this paper, we assume a valid amount of $ES{S}_{SOC}$ by considering that it may differ from the most used value of approximately 10%. Through the first constraint, Equation (8), we can optimize the ESS scheduling in the direction of minimizing the load pattern and the rate per day.

$${\displaystyle \sum }_{t=0}^{23}ES{S}_{o{p}_t}=0$$

(8)

$$O{P}_{ESS}=\left\{charge, discharge, stop\right\}$$

(9)

The operation of the ESS can be selected from one of the three operations: charge, discharge, and stop, as shown in (9). Through the second constraint, Equation (10), when the ESS is charged and discharged daily, the valid $ES{S}_{SOC}$ must have a value of 0 or more, and it cannot exceed the maximum capacity of the ESS as well.

$$0\le ES{S}_{SO{C}_t}\le ES{S}_{op} t=0, 1, \dots , 23$$

(10)

The third constraint, Equation (11), indicates that the maximum of summing up the real-time $ES{S}_{SOC}$ and the $Loa{d}_t$ values must be smaller than the annual maximum power demand. Here, $Loa{d}_t$ represents hourly power demand data, and $Annual Peak$ is the annual maximum power demand.

$$\max \left(Loa{d}_t+ES{S}_{o{p}_t}\right)\le Annual Peak t=0, 1, \dots , 23$$

(11)

The fourth constraints, Equations (12) and (13), indicate that the maximum capacity that the ESS can charge per unit time cannot exceed the capacity of power conversion system (PCS), and the maximum capacity that the ESS can discharge per unit time cannot exceed the capacity of the PCS.

$$ES{S}_{charg{e}_t}\le PC{S}_{cp}t=0, 1, \dots , 23$$

(12)

$$ES{S}_{discharg{e}_t}\le PC{S}_{cp}t=0, 1, \dots , 23$$

(13)

3. Methodology for ESS-Schedule Optimization

As aforementioned, the ESS can be used to reduce the total cost by lowering the maximum load to level the load, or by applying different unit rates for each period. This chapter presents the ‘framework for the optimal solution of the ESS schedule’ that satisfies both of these factors. The ‘framework for the optimal solution of the ESS schedule’ provides a solution that enables arbitrage while lowering the maximum demand based on an algorithm that reduces the maximum demand and an arbitrage algorithm that uses the difference in rate per unit time. The framework for deriving an ESS optimal scheduling solution is shown in Figure 1. Electric demand data, ESS capacity, and electricity tariffs are required as inputs. The solver for the optimal solution performs a peak demand reduction algorithm and an energy cost reduction algorithm based on the input data. The solver outputs the ESS schedule for one day, the daily electricity rate based on the TOU tariff, and the newly updated peak power demand.

Here, the ESS-schedule optimization algorithm for cost minimization optimizes the ESS schedule in such a way that the base rate (demand charge) is reduced through annual peak load reduction, and arbitrage profits (energy charge) are taken through monthly load leveling. The ‘cost-based optimization of a daily ESS schedule’ algorithm for deriving an ESS optimal scheduling solution is shown in Algorithm 1.

Algorithm 1: The cost-based optimization of a daily ESS schedule

## Input: electric demand data, ESS capacity, electricity tariff ## Output: $Daily Price\left(D_n\right)$, $Day\text{-}base ESS Schedule$, updated $Peak\left(goal\right)$ ## Find the maximum load for the previous 11 months, ## and set the result of subtracting the ESS capacity as the target maximum load 1: initialize $Peak\left(goal\right)$ 2: set $Peak\left(goal\right)=\max \left(Peak\left(i\right)\right)-ESS\left(cap\right),$ 3:  where $0<i<n$ and $n=$ peak-price calculation period ## Calculation of daily billing during the billing period (one month): ## Minimize peak load → Load balancing based on unit price 4: for each day in billing period do 5: if $Peak\left(D_i\right)>Peak\left(goal\right)$ then 6:  update $Peak\left(goal\right)=Peak\left(D_i\right)-ESS\left(cap\right)$ 7:  call Peak Demand Reduction Algorithm (PRA) # Peak-based pricing 8: else 9:  call Cost Reduction Algorithm (CRA)     # Cost-based pricing 10: return set of ($Daily Price\left(D_n\right)$, $Day\text{-}base ESS Schedule$)

The ‘peak demand reduction algorithm (PDRA)’ to reduce the demand charge through annual peak load reduction is shown in Algorithm 2. The PDRA considers the TOU price table, $Load\left(D_n\right), Peak\left(goal\right)$, and $ESS\left(capacity\right)$ as inputs, and outputs $expected price\left(D_i\right)$, $ESS\left(remain\right)$, and updated $Peak\left(goal\right)$ as result values.

Algorithm 2: The peak demand reduction algorithm (PDRA)

## Input: TOU Price Table, $Load\left(D_n\right)$, $Peak\left(goal\right)$, $ESS\left(capacity\right)$ ## Output: $expected price\left(D_i\right)$, $ESS\left(remain\right)$, updated $Peak\left(goal\right)$ ## Setting variables 1: set $tou=TOU\left(0\right), TOU\left(1\right), \dots , TOU\left(23\right)$  # Read from TOU Price Table 2: set $load=load\left(0\right), load\left(1\right), \dots , load\left(23\right)$ # Read from $Load\left(D_n\right)$ 3: set $peak=peak\left(goal\right)$ 4: set $ESS\left(cp\right)=ESS\left(capacity\right)$ 5: set $ch=optimizer.\mathrm{Variable}\left(24\right)$ 6: set $dc=optimizer.\mathrm{Variable}\left(24\right)$ 7: set $peak=optimizer.\mathrm{Variable}\left(1\right)$ ## Setting constraints 8: $constraint=\{$ 9: $\mathrm{sum}\left(ESS\left(ch,t\right)\right)=\mathrm{sum}\left(ESS\left(dc,t\right)\right), t=1,2,\dots ,24$ (15) 10: $\mathrm{sum}\left(ESS\left(op,t\right)\right)\le ESS\left(cp\right), t=1,2,\dots ,24$     (16) 11: $ESS\left(c\right)\le PCS\left(cap\right)$                  (17) 12: $ESS\left(d\right)\le PCS\left(cap\right)$                  (18) 13: $\max \left(Peak\left(d\right)\right)>Peak\left(goal\right)$              (19) 14: $peak\ge optimizer.\max \left(Load\left(t\right)+ch+dc\right)$       (20) 15: $\}$ ## Setting objective function and solve the problem 16: $object=optimizer.\mathrm{Minimize}\left(peak\left(goal\right)\right)$  # Minimize peak load 17: $prob=optimizer.\mathrm{Problem}\left(object, constraint\right)$ # Problem definition 18: $prob.\mathrm{solve}()$                  # Solving the problem ## Save optimal ESS schedule and update peak(goal) 19: $ch\left(d\right)=ch.value$                # Save charge schedule of ESS 20: $dc\left(d\right)=dc.value$                # Save discharge schedule of ESS 21: update $peak\left(goal\right)$ if new peak load         # Update peak load

Considering important modules, we first define the variables $ch$ and $dc$ for charging and discharging schedules for one day, respectively, and define the peak load as the variable peak. Second, under constraints (15) to (18), the amount of charge and discharge during one day of the ESS is the same, and the maximum capacity of the ESS cannot be exceeded during operation. The charging and discharging per unit time cannot exceed the PCS capacity. In constraints (19) to (20), the maximum load $\max \left(Peak\left(d\right)\right)$ of the day is set to be greater than the previous $Peak\left(goal\right)$ and the peak of the day. Third, we set the objective function to minimize the $Peak\left(goal\right)$ and set constraints to derive a solution through optimization. Finally, the optimal charge/discharge schedule result of the ESS for one day is output as $ch\left(d\right)$ and $dc\left(d\right)$, and if there is a new peak load, $Peak\left(goal\right)$ is updated.

The ‘energy cost reduction algorithm (ECRA)’ for taking arbitrage profits through monthly load leveling is shown in Algorithm 3. The ECRA considers the TOU price table, $Load\left(D_n\right)$, $Peak\left(goal\right)$, and $ESS\left(capacity\right)$ as inputs, and outputs $expected price\left(D_i\right)$, $ESS\left(remain\right)$, and updated $Peak\left(goal\right)$ as result values.

Algorithm 3: The energy cost reduction algorithm (ECRA)

## Input: TOU Price Table, $Load\left(D_n\right)$, $Peak\left(goal\right)$, $ESS\left(capacity\right)$ ## Output: $expected price\left(D_i\right)$, $ESS\left(remain\right)$, updated $Peak\left(goal\right)$ ## Setting variables 1: set $tou=TOU\left(0\right), TOU\left(1\right), \dots , TOU\left(23\right)$  # Read from TOU Price Table 2: set $load=load\left(0\right), load\left(1\right), \dots , load\left(23\right)$ # Read from $Load\left(D_n\right)$ 3: set $peak=peak\left(goal\right)$ 4: set $ESS\left(cp\right)=ESS\left(capacity\right)$ 5: set $ch=optimizer.\mathrm{Variable}\left(24\right)$ 6: set $dc=optimizer.\mathrm{Variable}\left(24\right)$ 7: set $peak=optimizer.\mathrm{Variable}\left(1\right)$ ## Setting constraints 8: $constraint=\{$ 9: $\mathrm{sum}\left(ESS\left(ch,t\right)\right)=\mathrm{sum}\left(ESS\left(dc,t\right)\right), t=1,2,\dots ,24$ (21) 10: $\mathrm{sum}\left(ESS\left(op,t\right)\right)\le ESS\left(cp\right), t=1,2,\dots ,24$     (22) 11: $ESS\left(c\right)\le PCS\left(cap\right)$                 (23) 12: $ESS\left(d\right)\le PCS\left(cap\right)$                 (24) 13: $\}$ ## Setting objective function: Minimize energy charge 14: $cost=\left\{load\left(t\right)+ch+dc\right\} d tou$        # Calculation of energy cost 15: $object=optimizer.\mathrm{Minimize}\left(cost\right)$      # Set objective function 16: $prob=optimizer.\mathrm{Problem}\left(object, constraint\right)$ # Problem definition 17: $prob.\mathrm{solve}()$               # Solve the solution ## Save the result of optimal ESS schedule 18: $ch\left(d\right)=ch.value$            # Save charge schedule of ESS 19: $dc\left(d\right)=dc.value$            # Save discharge schedule of ESS

Considering important modules, we first define the variables $ch$ and $dc$ for charging and discharging schedules for one day, respectively, and define the peak load as the variable peak. Second, under constraints (21) to (24), the amount of charge and discharge during one day of the ESS is the same, and the maximum capacity of the ESS cannot be exceeded during operation. The charging and discharging per unit time cannot exceed the PCS capacity. Third, to minimize the cost as an objective function, the total cost is calculated by multiplying the load and the charge/discharge amount of the ESS by the rate per unit time (TOU). Subsequently, after setting cost minimization as an objective function, constraint conditions are set, and a solution is derived through optimization. Finally, the results of the optimal charge/discharge schedule of the ESS for one day are obtained as $ch\left(d\right)$ and $dc\left(d\right)$.

4. Case Study with Commercial and Industrial Tariff in KEPCO, Korea

In Section 2, the problem is defined with objective functions and constraints, and in Section 3, a framework for solving the defined problem is presented. In Section 4, the presented methodology is applied based on the actual load in the branch of KEPCO with the ESS and the applicable tariff (see Table 2 in Section 2), and the results are analyzed.

The structure diagram and the scenario of a test system are shown in Figure 2. The system is composed of PV, an ESS, an electric vehicle (EV), a PCS, a building, and a power line from the grid. Electric power is supplied from the grid, PV, ESS, and consumed by a building’s load demand, ESS, and EV. The PCS is the interface between building (as load demand) and PV, the ESS, and EV. The PCS, which is combined with the inverter and converter, is a device for converting AC and DC. The scenarios are conducted with the following methods: (1) without an ESS, (2) only peak shaving with an ESS [10,25], (3) only load balancing with an ESS [20], and (4) peak shaving and load balancing with an ESS.

The input data require electric power load data, ESS information, and seasonal/hourly rate tariffs in one hour unit, as shown in

Table 3. The information regarding input data.

Electric Power Load (Per Hour)	Site	KEPCO Branch Office
	Period	1 September 2019–31 March 2020
ESS Model	Maximum ESS Capacity	51 kWh
	Maximum Output (per hour)	30 kW
	Efficiency of Charge (%)	90
	Efficiency of Discharge (%)	90
PCS	Maximum Output (per hour)	30 kW

. Power load data from October 1, 2019, to March 31, 2020 were acquired from the branch office of KEPCO, Korea. The ESS capacity is the maximum capacity that the ESS can store power, and the PCS capacity is the maximum permissible capacity during the charging and discharging of the ESS. KEPCO provides several types of seasonal/time rate tariffs, and a plan with demand charge and energy charge in

Table 4. A selected plan from General Service (A) II price table (from KEPCO in Korea).

Classification		Demand Charge (won/kW)	Energy Charge (won/kWh)
			Time Period	Summer (1 June–31 August)	Spring/Fall (1 March–31 May/ 1 September–31 October)	Winter (1 November–28 February)
High- Voltage A	Option I	7170	Off-peak load	57.7	57.7	66.4
			Mid-load	108.9	65.1	96.8
			Peak load	131.4	76.4	111.6

was selected for the optimal ESS-schedule simulation. Additionally, season and time-period classification is also defined in

Table 5. Season and time-period classification (from KEPCO in Korea).

Time Period	Summer (June, July, and August)/ Spring (March, April, and May) and Fall (September and October)	Winter (November, December, January, and February)
Off-peak	23:00~09:00	23:00~09:00
Mid-peak	09:00~10:00	09:00~10:00
	12:00~13:00	12:00~17:00
	17:00~23:00	20:00~22:00
On-peak	10:00~12:00 13:00~17:00	10:00~12:00 17:00~20:00 22:00~23:00

.

Comparative analysis with other methods used in the literature is shown in Figure 3 and Figure 4 for the impact of a battery on load demand. Peak demand occurred on the 19th, and the most load demand was used on the 5th during September. The load profile without an ESS in Figure 3a shows that peak demand occurred in the daytime, and it would affect the demand charge for one year. When the ESS schedule for peak shaving was applied, peak demand was reduced to 177.8 kW from 194.0 kW, shown in Figure 3b, and would reduce the demand charge for one year. When ESS scheduling for load balancing was applied, peak demand was not reduced; instead, the used energy shifted from the on-peak time zone to the mid-peak time zone, shown in Figure 3c, with the saving energy charge. When ESS scheduling for peak shaving and load balancing was applied, peak demand was reduced to 177.8 kW from around 194.0 kW, shown in Figure 3d, because peak demand occurred on this day, and it would reduce the demand charge for one year.

One the other hand, the load profile without ESS in Figure 4a shows that the most load demand was used for this day, and it affected the energy charge just as the used amount did. When ESS scheduling for peak shaving was applied, peak demand was reduced to 171.9 kW from around 180.0 kW, shown in Figure 4b, but it would not reduce the demand charge. When ESS scheduling for load balancing was applied, the used energy shifted from the on-peak time zone to the mid-peak time zone in Figure 4c with the saving energy charge. When ESS scheduling for peak shaving and load balancing was applied, peak demand was slightly reduced, and the used energy shifted from the on-peak time zone to the mid-peak time zone, shown in Figure 4d, because the most load demand was used and a slight peak demand occurred on this day, and it reduced the energy charge just as the used amount did.

Comparative cost analysis with other methods used in the literature is shown in

Table 6. Comparative cost analysis with other methods in the literature.

Date		19 September 2019	5 September 2019	September 2019
(1) Without ESS	Demand Charge (won)	46,366	46,366	1,390,980
	Energy Charge (won)	165,900	176,665	3,634,574
	Total Cost (won)	212,266	223,031	5,025,554
(2) With ESS (Peak Shaving)	Demand Charge (won)	42,494	42,494	1,274,826
	Energy Charge (won)	165,788	176,747	3,638,954
	Total Cost (won)	208,282	219,241	4,913,780
(3) With ESS (Load Balancing)	Demand Charge (won)	46,366	46,366	1,390,979
	Energy Charge (won)	165,760	176,526	3,630,390
	Total Cost (won)	212,126	222,892	5,021,369
(4) With ESS (Peak Shaving and Load Balancing)	Demand Charge (won)	42,494	42,494	1,274,826
	Energy Charge (won)	165,788	176,526	3,632,173
	Total Cost (won)	208,282	219,020	4,906,999
Gain (won)	(2) Method	3984	3790	111,774
	(3) Method	140	139	4185
	(4) Method	3984	4011	118,555
Gain (%)	(2) Method	1.9	1.7	2.2
	(3) Method	0.1	0.1	0.1
	(4) Method	1.9	1.8	2.4

. When the peak shaving method was applied, the daily gain was 1.9% and 1.7%. When the load balancing method was applied, the daily gain was both 0.1%. When the peak shaving and load balancing method was applied, the daily gain was 1.9% and 1.7%. Monthly gains were 2.2% with peak shaving, 0.1% with load balancing, and 2.4% with peak shaving and load balancing.

Like this, in the case of December, in which the gap of TOU was large and peak load did not occur, we can observe that the ESS discharged when the unit price was high and charge when the unit price was low to reduce the total expense from ESS scheduling on 5 December in Figure 5. In the case of March, in which the gap of TOU was small and peak load did not occur, we can observe that the ESS discharged when the unit price was high and charged when the unit price was low to reduce the total expense from ESS scheduling on 5 March in Figure 6.

The monthly demand and energy and billing from September 2019 to March 2020 as a result of optimal ESS scheduling are summarized in

Table 7. Total cost with ESS and without ESS from September 2019 to March 2020.

Date		September 2019	October 2019	November 2019	December 2019	January 2020	February 2020	March 2020
Without ESS	Peak (kW)	194.0	194.0	194.0	194.0	194.0	194.0	194.0
	Monthly Peak (kW)	194.0	95.0	84.9	99.8	96.1	97.4	89.4
	Demand Charge (won)	1,390,980	1,390,980	1,390,980	1,390,980	1,390,980	1,390,980	1,390,980
	Energy Charge (won)	3,634,574	2,476,439	3,221,496	3,650,208	3,678,317	3,593,833	2,575,988
	Total Cost (won)	5,025,554	3,867,419	4,612,476	5,041,188	5,069,297	4,984,813	3,966,968
With ESS	Peak (kW)	177.8	177.8	177.8	177.8	177.8	177.8	177.8
	Monthly Peak (kW)	177.8	86.7	68.7	89.3	88.9	88.5	80.0
	Demand Charge (won)	1,274,826	1,274,826	1,274,826	1,274,826	1,274,826	1,274,826	1,274,826
	Energy Charge (won)	3,632,074	2,472,116	3,197,500	3,625,413	3,653,521	3,570,637	2,571,665
	Total Cost (won)	4,906,900	3,746,942	4,472,326	4,900,239	4,928,347	4,845,463	3,846,491
ΔBilling (won)	Peak (kW)	16.2	16.2	16.2	16.2	16.2	16.2	16.2
	Monthly Peak (kW)	16.2	8.3	16.2	10.5	7.2	8.9	9.4
	Demand Charge (won)	116,154	116,154	116,154	116,154	116,154	116,154	116,154
	Energy Charge (won)	2500	4323	23,996	24,795	24,796	23,196	4323
	Total Cost (won)	118,654	120,477	140,150	140,949	140,950	139,350	120,477
	Total Cost (%)	2.4	3.1	3.0	2.8	2.8	2.8	3.0

Note: The value of energy storage devices is not considered.

. As shown in Table 7, the September peak affected the demand charge of the remaining electricity, and the energy charge from load balancing only affected the month. The value of energy storage devices was not considered because customers could make the decision to adopt an ESS by comparing the gain with the investment if the gain with an ESS is calculated. The average cost reduction was approximately 3%, and most of them occurred in the demand charge. Although the same ESS was used to reduce the peak load and load gaps, it can be observed that the impact on the total cost differed depending on the utility rate policy.

5. Conclusions

In this paper, to optimize the ESS schedule, a cost-effective methodology was suggested to consider both reducing the peak load and balancing the load. We also presented a case study result to show that there is a priority among the objective functions of the ESS schedule and that demand charge is more important than energy charge. Finally, we showed that the ESS-schedule problem for customers to reduce costs is also beneficial to power system operation based on the utility’s rate policy.

In future research, the electric power demand prediction problem for individual customers is important, especially in terms of accuracy. In addition, cost-effective approaches with photovoltaics and ESS are needed, where the prediction of solar photovoltaic power output related to the weather is important. In addition, research on estimating the optimal installation capacity of the ESS in advance is necessary to calculate the appropriate investment scale.