*3.2. Load Model of Charging and Discharging Distributed Energy Storage* 3.2.1. Electric Vehicle (EV)

There have been 2,356,657 EVs recorded in the national regulatory platform between 2017 and 2019, of which blade electric vehicles (BEV) account for 84.6%, plug-in hybrid electric vehicles (PHEV) account for 15.3%, and fuel cell electric vehicles (FCEV) account for 0.01% [49]. The disorderly charging behavior of EVs will create harmonics, aggravating the peak–valley load difference and increasing grid loss. The factors affecting the temporal and spatial distribution characteristics of EVs' charging and discharging are as follows: different EVs types, fast- or normal-charging mode of charging facilities, EVs users' behavior, number of different types of EVs, environment temperature, and time-of-use price. The EV charging power is assumed to be uniformly distributed U(*a*, *b*). The initial state of charge (SOC) is assumed to be normal distribution N(*μ*, *σ*). The detail modelling process is as follows.

(1) Respectively obtain the charging time probability of buses, private cars, taxis, and official vehicles, charging power, charging time, and discharge probability and time within the preset time period. Obtain a daily 15-min cumulative charging load and discharge power of all types of EVs.

According to the national standard of China GB/T 20,234, shown in Table 1, fastcharging mode adopts DC, with a maximum voltage of 750/1000 *V* and the maximum currents of 80/125/200/250 A, i.e., the charging powers from 60 kW through 250 kW. The normal charging mode adopts AC, with the maximum voltage 440 *V* and the maximum current 16/32/63 A, i.e., the charging powers from 7.04 kW through 27.72 kW.


**Table 1.** Different EV charging modes according to national standard of China GB/T 20,234.

Taking the taxi operation in Beijing, China, as examples, large-class taxi drivers shift every 24 h and small-class taxi drivers shift every 12 h. The ratio of large class to small class is about 4:1. The large class taxi drivers have a long rest and can choose normal charging, satisfying U(14.08, 27.72) (referring to Table 1) at public charging piles in residential or commercial areas. The small-class taxi drivers have a short rest time, and generally charge in fast mode, satisfying U(80, 250) (referring to Table 1). They charge twice a day, respectively, 2:00–5:00 am (the 8th through 20th time period) and 12:00–14:00 pm (the 48th through 56th time period) for large class taxi, and 2:00–4:00 am and 12:00–14:00 pm for small-class taxi. The battery capacity is assumed to be uniformly distributed under U(80, 100). The SOC is assumed to be normally distributed under N(0.5, 0.1). The charging time length, *T* (number of time periods), is calculated as follows, where *C*EV is the battery capacity (kW·per time period), *η* is the charging efficiency 0.8–0.9, and *Pcharge* is the charging power (kW).

$$T = (1 - \text{SOC}) \cdot \frac{\mathbb{C}\_{\text{EV}}}{\eta \cdot P\_{\text{charge}}} \tag{1}$$

(2) According to the probability distribution simulating the type of EV and its charging behavior, randomly sample the charge power, battery capacity, SOC, charge starting time of each EV. Then calculate charging time duration and charging end time. Finally, calculate cumulative charge load curve for 96 time periods.

The battery capacity of private cars follows a uniform distribution, U(20, 30), as does slow-charging power, U(4, 6). Considering that the expectation of the probability distribution at the beginning of charging of private cars is 17.6 h, the variance is 3.4. The expectation of the initial state of charge is 0.6, while the variance is 0.2. The charging efficiency is 0.9. The mean value of the probability distribution of mileage is 3.2, and the variance is 0.88. Figure 3a depicts the load curves of private cars of different sizes during disorderly charging as the number of vehicles grows from 200,000 in 2020 to 1,000,000 in 2025.

**Figure 3.** The disorderly charging load curve of EVs: (**a**) private cars; (**b**) taxis; (**c**) buses; (**d**) official and special vehicles.

Similarly, the disorderly charging load curves of electric taxis, buses, and special vehicles are respectively shown in Figure 3b–d.

The total charge load curve of the four types of EVs shown in Figure 4 includes the cars in Figure 3a, taxis in Figure 3b, buses in Figure 3c, and official and special vehicles in Figure 3d. The maximum charging power, 3425 MW, of EVs in 2025 in Figure 4 accounts for no more than 2.8% of current provincial peak load, about 123,810 MW in 2020.

**Figure 4.** The total charging load curve of private cars, taxis, buses, and official and special vehicles.

#### 3.2.2. Ice Storage System

Ice storage systems make use of the low electricity price at night to make ice and store it in an ice storage device. During the peak period of power consumption in the daytime, the ice melting releases its cooling capacity, which reduces the system cooling pressure during the peak load period and reduces the system's operational cost. The main equipment of an ice storage system includes a refrigerator and an ice storage tank. Its working mode can be divided into separate cooling by the electric refrigerator, separate cooling by the ice storage tank, simultaneous cooling by the electric refrigerator and ice storage tank, and combined cooling by the electric refrigerator and ice storage tank. The electric load and cooling power curves of an ice storage system are shown in Figure 5.

**Figure 5.** An ice storage system: (**a**) electric load curve of ice storage system; (**b**) cooling power of an ice storage system.

Ice storage units can effectively use surplus power for refrigeration at night, transferring part of the power load at the peak of the day to the low-cost power period at night. The refrigeration host does not turn on, or turns on less, during peak times to reduce power load and improve the load composition of the power grid, which is conducive to the stable operation of the power grid.

#### 3.2.3. Considering Demand Response

According to the power consumption characteristics of users and the operation time and energy consumption demand of different electrical equipment, power loads can be divided into uncontrollable loads, transferable loads, and interruptible loads.

Uncontrollable loads have no energy storage characteristics; the power consumption time is relatively fixed or may use power at any time according to the users' wishes, such as lighting load, or load from TVs, computers, etc. Their operation time or power fluctuation range are small and basically without ability to transfer load. Transferable loads owe to equipment with or without energy storage characteristics, but whose power consumption periods are flexible. Washing machines and timed rice cookers do not have energy storage, but they use electricity flexibly, and, as their total power consumption is certain, they have load transfer capacity. Interruptible loads, obtain from equipment such as air conditioners and water heaters, have flexible power consumption characteristics and short-term poweroff operation without great impact on users, but they may deviate from the original power consumption habits, such as by causing room and water temperature changes. For more on the operational model of integrated controllable load, we refer the reader to [50]. The response curve of a controllable load is shown in Figure 6, below.

According to the Chinese national standard of residential electricity, each family uses maximum 6 kW as the basic design capacity for ordinary residential living in a 61–100 m2 house. The power of the washing machines is generally 0.7–1 kW. The power of the timed rice cooker for a family having six persons is generally 0.8–1 kW. The power of an air conditioner with a heating function is generally 1.1–1.5 kW and there are typically 2–3 air conditioners for a large family. The power of the water heater is generally 1.2–2 kW. All household appliances always don't use electricity at the same time, to avoid tripping circuit breakers. Thus, for a family, the transferable load, including a washing machine or a timed rice cooker, is about 0.7–1 kW. The interruptible load including an air conditioner or a water heater is about 1.1–1.5 kW.

Supposing there are 93 million people in a province, and every six persons form a family, we can deduce that there are about 15.5 million households in a province. Assuming the maximum load transfer of 20,000 households is 9 MW during the 45th–48th period, 14 MW during the 66th–68th period, and 8.4 MW during the 79th–84th period, we deduce that, ideally, there are about 6975 MW, 10,850 MW, and 6510 MW load transfers at three peak load periods, respectively. For residential interruptible loads, from 10:00 to 12:00 am, shutting down two air conditioners, or only one from 15:00–19:00 pm can reduce peaking load. In Figure 6c,d, if the maximum interrupted load of 20,000 households is 22 MW in the 47th–48th period and 7.5 MW in the 74th–76th period, we deduce that, ideally, there are about 17,050 MW and 5813 MW for a province. Thus, there are about maximum 24,025 MW over 11:30–12:00 am, 10,850 MW over 16:00–16:30 pm, 5813 MW over 18:30–19:00 pm, and 6510 MW over 19:45–21:00 pm in terms of demand response. During the maximum demand response load, the transferable and interrupted load accounts for 29% and 71%, respectively.

(**c**) (**d**)

**Figure 6.** The response curve of 4000 to 20,000 household loads. (**a**) Transferable load curve before and after peak shifting and valley filling; (**b**) load transfer curve; (**c**) interrupted load curve before and after interrupting load; (**d**) interrupted load curve.

#### 3.2.4. Heat Storage System

Heat storage devices can adjust the heating load through the process of repeatedly cycling heat storage and heat release.

The general energy storage system adopts the two charging and discharging cycles of "valley charging and peak discharging" and "flat charging and peak discharging", every day, to reduce the times of charging and discharging to ensure the service life of the energy storage device. Three different types of heat storage devices are selected for the project, and the heat storage power and capacity meet the normal distribution. The expected values are configured as 1 MW/4 MWh, 2 MW/8 MWh, and 4 MW/12 MWh accounting for 1/3. At the same time, the heat storage exothermic conversion rate of the heat storage device is 0.92, and its heat storage/exothermic efficiency is 0.90. The initial heat storage state and the initial time of heat storage/exothermic efficiency follow different uniform distributions. By changing the total number of heat storage devices to 100, 200, 300, 400, or 500, the disordered charge–discharge curves of heat storage devices in different scales can be obtained from Figure 7a–d, respectively.

**Figure 7.** Electrical load curve of a heat storage system: (**a**) curve of a type 1 (1 MW/4 MWh) heat storage device; (**b**) curve of a type 2 (2 MW/8 MWh) heat storage device; (**c**) curve of a type 3 (4 MW/12 MWh) heat storage device; (**d**) comprehensive heat storage load curve of a heat storage device.

#### 3.2.5. Decentralized Electrochemical Energy Storage

Under the time-of-use price proposed by the power grid company, industrial users adopt electrochemical energy storage devices because of their advantages of high energy density, fast response, and low maintenance cost [51]. The user-side energy storage uses the peak valley price difference to obtain income, namely, charging in the low electricity price period and discharging in the peak period to profit between peak and valley electricity prices. The total installed capacity of electrochemical energy storage in a province from 2021 to 2025 is about 480 MW, 1680 MW, 3840 MW, 6000 MW, and 8400 MW, respectively by year, and a decentralized system of 1 MW/2 MWh accounts for 40%, of 2 MW/5 MWh account for 40%, and of 6 MW/36 MWh accounts for 20%. In order to make use of the price differences between peaks and valleys, under two-charge and two-discharge modes, the optimal strategy is to charge the energy storage in the valley period of 0:00–8:00 am every morning and in the normal period of 12:00–14:00 pm, and discharge with a total duration of 4h during the peak period of 10:00–12:00 am and 14:00–16:00 pm every day. The 35% of full charging and discharging load curves of 200, 700, 1600, 2500, and 3500 devices are shown in Figure 8a,b, respectively.

**Figure 8.** The charging and discharging load curves of the planned capacity of electrochemically stored energy from 2021 to 2025: (**a**) 35% charging and discharging; (**b**) full charging and discharging.

In a sample case, the maximum discharging load is 8243 MW and the maximum charging load is 7462 MW in full charging and discharging mode, resulting in reduced battery life. In a realistic case, to protect battery life, a 30–40% charging and discharging mode is adopted. In the sampled 35% case, the maximum discharging load is 2942 MW and the maximum charging load is about 2671 MW.

#### 3.2.6. Distributed Storage Aggregation Provider (DSAP)

A DSAP is an independent organization that integrates all kinds of distributed energy storage resources and provides them to market buyers. As an intermediate organization in the distributed energy storage and power grid for providing power resources, it extracts, evaluates, and integrates distributed energy storage resources through professional means, integrates decentralized energy storage into system-schedulable resources, and participates in the operation of the power system.

The grid regulation takes reducing the variance of load curve and reducing the peak load as its charging and discharging objectives when dispatching and managing distributed energy storage. In order to reduce the negative impact of distributed energy storage on the power grid, combined with the time of use price mechanism, the goals of reducing the peak-to-valley difference of loads and reducing peak load can be set. A day is divided into 24 time periods-hours-and the optimization variable is the charge and discharge power of energy storage equipment in each time period.

In Figure 9a, EVs are the main charging load for the basic load curve in the period of 2:00–6:00 a.m. and 12:00–14:00 p.m., accounting for 65% and 52%, respectively, leading to an increased maximum load of 5167–5260 MW at 2:15–3:00 a.m. and of 5526–5670 MW, without a heat storage system, at 12:00–13:45 p.m., as shown in Figure 9b, corresponding to the electricity prices of 0.3052 yuan/kWh and 0.6104 yuan/kWh in Table 2, respectively. Until 2025, considering the heat storage system in summer, an extreme and unrealistic case for evaluating extreme maximum loads, the total increased maximum load of five energy storage systems between 6226–8209 MW at 2:15–3:00 a.m. and 6012–7829 MW at 12:00–13:45 p.m. accounts for 5–7% of Guangdong's unified regulation load of 123,810 MW in 2020.

**Figure 9.** The load curve of distribution energy storage systems in 2025: (**a**) charging electric load curves of five types of energy storage systems; (**b**) DSAP total charging power load curve including EVs and ice storage, considering demand response and a decentralized shared system on a day using heat storage systems (in winter) and without the use of heat storage systems (in summer).


**Table 2.** The electricity price of large industrial power in Guangzhou, China, in its distribution grid.

#### **4. Impact on Power Grid Capacity, Load Characteristics, and Safety Margins**

*4.1. Distributed Energy Storage System on Load Side*

4.1.1. Impact on Peak Shifting, Energy Efficiency, and Economic Benefit for Consumers

The construction of power grid needs to meet the demand of peak load, resulting in low asset utilization and operational efficiency. Especially in summer, more than 40% of the grid load is air-conditioning load, while its annual power consumption is less than 10%. In Figure 9b, without considering heat storage systems until 2025, if the basic load is 123,810 MW in 2020, according to energy storage planning, using distributed energy storage for peak shifting of 1359–2353 MW at 10:00–12:00 a.m. and of 1024–1338 MW at 14:00–16:00 p.m., valley filling of 2314–5260 MW at 1:00–5:00 a.m. and a flat period fulfilling 5526–5670 MW at 12:00–14:00 p.m. can transfer no more than 1–2% of peak load, which increases to 4–5% load in valley and flat periods.

If the power consumption on a peak-load day is 2,000,000 MWh, the total charging and discharging power consumption of four energy storage systems without heat storage systems, depicted in Figure 9b, are 43,224 MWh and 4691 MWh, accounting for 2.2% and 0.23% of the daily consumption, respectively. The net increased daily power consumption is 38,533 MWh, accounting for 2% of the daily consumption. The net increased daily power consumption includes rigid power demand of EVs and energy loss of electrochemical battery system, may cause reduced energy efficiency.

According to rough estimation, the total power fee in the peak load day is 20.73 million yuan at the charging period, and a maximum 4.30 million yuan at the discharging period, as decided by the proportions at the user side, the power grid, and the supplier side. If the electrochemical battery capacity of the user side accounts for 7% of all distributed electrochemical energy storage in 2025, the total saving cost for users on a peak-load day is about 0.3 million yuan.

#### 4.1.2. Impact on Safety Margin

The disorderly charging behavior of large volumes of EVs increases in risk until 2025, although the maximum charging power, 3425 MW, of the EVs simulated in Figure 4 accounts for no more than 2.8% of current provincial peak load, and 52–65% of energy storage load. For five types of energy storage, although the whole energy storage system flattens the single peak at 10:00–12:00 a.m. and 14:00–16:00 p.m., it forms multiple subpeaks at 1:00–5:00 a.m. and 12:00–14:00 p.m. Although the risk level, at peak periods, of exceeding the limit decreases, the number of multiple sub-peak risks increases.

The orderly charging behavior of electrochemical energy storage invested in by the power grid, in case of emergency, can avoid overloading some equipment and of having low voltage of some nodes. In this case, the energy storage equipment put into operation at such critical moments can play the role of emergency support.

#### 4.1.3. Impact on Planning and Construction Considering EVs

The power demand of EV batteries, accounting for 65% of charging power in the valley period and influenced by their work mode, affects the distribution grid planning, substation capacity, and equipment selection. The site selection of a substation is selected in combination with the distribution of EV charging load. The fixed capacity of the substation refers to the determination of the main transformer capacity of the substation, and the appropriate transformer capacity load ratio shall be considered. The simultaneous rate of conventional load peaks and EV charging load peak have an important impact on the determination of transformer capacity load ratio. With the growth of EVs, the uncertainty of charging load leads to the maximum load prediction deviation, requiring great changes of total substation capacity and layout.

If a large number of EVs are charged at widely distributed charging piles, the voltage waveform of a 380 V public bus will be seriously distorted because the piles are distributed in a 400 V low-voltage distribution system. The short-term fast charging of charging equipment may cause too fast a load change, produce impulse voltage, and endanger the safety of the power grid. In order to ensure power quality, on the one hand, corresponding active-filter and reactive power-compensation devices can be equipped. On the other hand, when planning to build a large-capacity charging station, it may not share the same section of bus with loads sensitive to power quality. The uncertain characteristics of EV charging time add uncertain factors to the power flow calculation of distribution system. In order to meet safe transmission requirements under various operation modes, a certain margin can be reserved for the line capacity. The weak links can be strengthened to ensure that the line current does not exceed the limit. When considering the impact of EV development on distribution network planning, the analysis of planning operational cost is more important. It is necessary to optimize the economic model in the planning process, so that the actual planning scheme can take into account adaptability and economy.

#### *4.2. Distributed Energy Storage System on Power Resource Side*

#### 4.2.1. Impact on Stabilizing Output Power of Renewable Energy

In order to reduce the impact of clean energy output fluctuation on the power grid, installing appropriate distributed energy storage can increase the controllability of clean energy output power and stabilize power fluctuation to ensure the safe, stable, and economic operation of the clean energy grid with high permeability.

Based on power grids' operational requirements, the objectives of stabilizing the output fluctuations of energy storage devices fall into three categories. The first category has the lowest requirements. When the power system's operational conditions and regulation capacity are certain, it is expected that the grid connection of clean energy will not affect dynamic stability or cause frequency regulation pressure, that is, the output can meet the fluctuation constraints and frequency regulation requirement. The second type has an additional and more stringent requirement that the power generated by clean energy be able to track the power generation plan in real time. The third category has the highest requirements; the expected output must not only quickly respond to changes in power grid frequency and have the abilities of system frequency modulation and peak shaving, but also have a certain schedulability and the ability to coordinate large-scale clean energy generation.

#### 4.2.2. Improving Clean Energy Consumption

We next consider the impact on the clean energy consumption of the power grid under various application modes, such as the joint configuration of a distributed energy storage system and flexible interconnection devices, and on participation in the demand response.


negative effect in valley periods. Wind and solar grid connections reduce peak load, greatly reduce the overall time series curves, and increases the capacity margin. The net load curve has multiple valleys, and the net load in valley periods can even be negative. In order to avoid the phenomenon of abandoning wind and PV, the main network and other distributed energy sources need to have downward regulation capacity and reserve capacity. In some periods, the wind power output and load trends are inconsistent, resulting in large fluctuation times and amplitude of the net load curve, which requires the system to have a flexible climbing ability.

In order to improve the consumption level of clean energy and avoid the phenomenon of abandoning wind and PV when the net load is negative, distributed energy storage can be charged during valley periods and discharged during peak load periods. Through the time sequence transfer of the net load curve, the energy utilization rate of uncontrollable distributed generation can be increased, to avoid the phenomena of abandoning wind and PV power and high-price power purchasing under the low acceptance capacity of the system.

#### **5. Conclusions**

There is a developing trend of establishing a hierarchical and partitioned-energy internet sharing operation platform to gather wind power, photovoltaic, energy storage, flexible load, heat and cold energy systems, and energy suppliers to realize consumer energy services. The energy interconnection and sharing platform of Dongguan's local dispatching was established in 2017 and put into operation in 2019. Based on cloud–edge computing technology, it has realized the pioneering construction of distributed cloud energy storage with access to a local power grid management platform. Although the current market policy has not been liberalized, the platform performs only power monitoring and operation and the maintenance of power load equipment and does not have the functions of heat energy collection, carbon emission monitoring, or market transaction. However, with the progress of technology and the attention of the national industry, it will further realize such a platform with ideal functionality.

Building a local dispatching platform of the provincial demand side's response platform has great significance for consolidating new power system infrastructure for carbon peaking and carbon neutralization and for improving the efficiency of comprehensive energy management. Our suggestions are as follows:


**Author Contributions:** Conceptualization, J.L. and Y.X.; methodology, J.L., Y.X. and D.Z.; software, Y.X., D.Z. and J.L.; validation, D.Z., Y.X. and J.L.; formal analysis, Y.X., J.L. and D.Z.; investigation, J.L. and D.Z.; resources, Y.X.; data curation, Y.X. and J.L.; writing—original draft preparation, J.L. and Y.X.; writing—review and editing, J.L.; supervision, J.L. and D.Z.; project administration, J.L. and Y.X.; funding acquisition, J.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by China Southern Power Grid Company Limited Science and Technology Project Foundation (030000KK52190010 (GDKJXM20198116)).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** Thanks to the researchers in Sichuan Energy Internet Research Institute, Tsinghua University, who helped provide the source code and some analysis contents. We would like to express our gratitude to them for their help and guidance.

**Conflicts of Interest:** The authors declare no conflict of interest.
