CN116094000A - A hybrid energy storage system capacity and power configuration method and system - Google Patents

Abstract

A hybrid energy storage system capacity and power configuration method and system, belonging to the technical field of hybrid energy storage system control. It is characterized in that it includes the following steps: with the goal of minimizing the comprehensive cost of the hybrid energy storage system, a two-layer optimization model is constructed, the upper layer optimization model is used to solve the capacity and power configuration problems of the hybrid energy storage system, and the lower layer optimization model is used to solve the photovoltaic energy storage system. ‑The operation problem of the hybrid energy storage system; use the deep deterministic strategy gradient algorithm to solve the lower-level optimization model to obtain the operating cost, based on the operating cost, use the particle swarm algorithm to solve the upper-level optimization model to obtain the optimal power of the supercapacitor and battery and capacity configuration. Compared with the optimal operation method based on typical days, the present invention fully considers the uncertainty of photovoltaic output, so that the mixed energy storage capacity and power configuration results are more in line with actual needs.

Description

A hybrid energy storage system capacity and power configuration method and system

Technical Field

A hybrid energy storage system capacity and power configuration method and system, belonging to the hybrid energy storage system control technology field.

Background Art

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

As the price of petrochemical energy continues to rise, photovoltaic power generation has become the most promising new power generation method to replace coal-fired power in the future. Photovoltaic power generation is low-carbon, environmentally friendly and easy to maintain, but affected by the movement of the sun and the atmospheric environment, photovoltaic power generation is intermittent and volatile, which seriously affects the power quality and power supply reliability of the power grid. Therefore, it is often necessary to configure energy storage equipment for photovoltaic power generation.

Energy storage equipment can be divided into power-type energy storage and energy-type energy storage. Power-type energy storage has the advantages of high power density, fast response speed, and long cycle life, but its energy density is relatively low. Typical power-type energy storage includes flywheel energy storage, superconducting energy storage, and supercapacitor energy storage. Energy-type energy storage has high energy density and is suitable for storing a large amount of energy, but is not suitable for short-term high-power charging and discharging. It mainly includes hydrogen energy storage and compressed air energy storage. The hybrid energy storage system is equipped with a variety of energy storage devices at the same time. It can combine the advantages of power-type energy storage and energy-type energy storage, flexibly respond to power demands on different time scales, and smooth out photovoltaic power fluctuations caused by cloud obstruction and other reasons in a short period of time. At the same time, it participates in grid dispatching to promote photovoltaic consumption and reduce abandoned light.

However, the current construction cost of energy storage equipment is relatively high. In order to improve the economic efficiency of photovoltaic-energy storage systems, it is necessary to reasonably select energy storage equipment and configure appropriate capacity. The current hybrid energy storage capacity configuration method is to smooth out the fluctuations in new energy power. Usually, a typical photovoltaic operation day is selected. The data time scale is relatively small, usually 1 minute. Wavelet packet transform, Hilbert-Huang transform, variational mode decomposition and other methods are used to decompose photovoltaic output in the frequency domain. The frequency bands are divided according to indicators such as volatility. The high-frequency part of the photovoltaic output is allocated to power-type energy storage, and the low-frequency part is allocated to energy-type energy storage, so as to obtain the optimal configuration of the hybrid energy storage system. However, the above methods usually do not consider load power and time-of-use electricity prices, and the output of the hybrid energy storage system is not obtained through operation optimization, which lacks economy in practical applications. There are also hybrid energy storage configuration methods that focus more on the operational optimization of the photovoltaic-storage system. This type of method usually sets the time scale of renewable energy power data to 15 minutes or 1 hour, configures two or more energy storages with different response speeds, and reduces wind and solar power curtailment and electricity purchase costs through operational optimization, thereby improving the economy of the system. However, due to the large time scale, this type of method is difficult to consider the smoothing of short-term power fluctuations of renewable energy.

However, the above methods have the following defects: 1) The hybrid energy storage capacity configuration results obtained by the above methods are greatly affected by the typical days of photovoltaic operation, and the capacity configuration results obtained on different typical days often have large differences; 2) The above methods only consider smoothing short-term photovoltaic fluctuations, or only consider the optimized operation of hybrid energy storage in the power grid to improve the economy of the system, and lack a hybrid energy storage system configuration method that comprehensively considers smoothing photovoltaic power fluctuations and optimizing the operation of the photovoltaic-storage system; 3) When the operation optimization time scale is small, the traditional algorithm is difficult to solve and has poor convergence, making it difficult to solve the optimal output of the hybrid energy storage system.

Summary of the invention

The technical problem to be solved by the present invention is: to overcome the shortcomings of the prior art and provide a method and system for configuring the capacity and power of a hybrid energy storage system by reasonably configuring the power and capacity of supercapacitor energy storage and battery energy storage, thereby improving the economy of a photovoltaic-hybrid energy storage system, enabling the hybrid energy storage to smooth short-term fluctuations in photovoltaic power while optimizing the operation of the photovoltaic-hybrid energy storage and promoting photovoltaic consumption.

The technical solution adopted by the present invention to solve the technical problem is: the capacity and power configuration method of the hybrid energy storage system is characterized by comprising the following steps:

With the goal of minimizing the comprehensive cost of the hybrid energy storage system, a two-layer optimization model is constructed. The upper optimization model is used to solve the capacity and power configuration problems of the hybrid energy storage system, and the lower optimization model is used to solve the operation problems of the photovoltaic-hybrid energy storage system.

A deep deterministic policy gradient algorithm is used to solve the lower-level optimization model to obtain the operating cost. Based on the operating cost, a particle swarm algorithm is used to solve the upper-level optimization model to obtain the optimal power and capacity configuration scheme of the supercapacitor and battery.

Preferably, the comprehensive cost of the hybrid energy storage system includes investment and construction costs and operating costs, wherein the investment and construction costs include the power and capacity investment and construction costs of supercapacitors and the power and capacity investment and construction costs of batteries; the operating costs include the output cost of the photovoltaic-hybrid energy storage system, the electricity purchase cost, the abandoned light cost, and the power fluctuation penalty cost.

Preferably, the constraints of the upper optimization model include supercapacitor energy storage configuration power and capacity constraints, battery energy storage configuration power and capacity constraints;

The constraints of the lower-level optimization model include power balance constraints, battery energy storage charging and discharging power constraints, battery energy storage charging and discharging climbing power constraints, battery energy storage power constraints, supercapacitor energy storage charging and discharging power constraints, and supercapacitor energy storage power constraints.

Preferably, the specific process of using the deep deterministic policy gradient algorithm to solve the lower-level optimization model is as follows: the objective function of the lower-level optimization model is the annual operating cost of the hybrid energy storage system, and the photovoltaic and load power data within the historical set time are used as input. The optimization problem is converted into a Markov decision model and solved using the deep deterministic policy gradient algorithm.

Preferably, the action space of the Markov decision model is the current charging and discharging power of the supercapacitor and battery, the state space is the photovoltaic power, load power, supercapacitor charge state, battery charge state and electricity price information at the current moment, and the reward function is obtained by transforming the objective function of the lower-level optimization model, and the objective function minimization problem is transformed into the cumulative reward function maximization problem.

Preferably, a deep deterministic policy gradient algorithm is used to solve the Markov decision model, and four fully connected neural networks are constructed. The four fully connected neural networks are Actor network, Actor target network, Critic network and Critic target network. The Actor network calculates the current action space, and the environment feeds back the current state space and reward function to calculate the cumulative reward function.

The current action, state, reward function and the state at the next moment are stored in the experience replay pool, and random sampling is used from the experience replay pool for parameter update of the Actor network and the Critic network. The parameters of the Actor target network and the Critic target network are updated using the soft update method. The optimal hybrid energy storage charging and discharging control strategy is obtained through continuous training, and the operating cost is calculated.

Preferably, the process of solving the upper optimization model includes the following steps:

The upper model initializes the particle swarm parameters, and the power and capacity of the battery and supercapacitor are used as the position parameters of each particle;

Adjust the lower-level constraints according to the particle position parameters, solve the lower-level optimization model, initialize the deep deterministic policy gradient algorithm environment, train the hybrid energy storage charging and discharging control strategy, and then calculate the annual operating cost;

According to the obtained annual operating cost, the fitness value of each particle is calculated and the global optimum is obtained. If the maximum number of iterations of the upper layer is reached, the minimum total cost is output. The position parameters of the global optimal particle are the optimal configuration of the capacity and power of the hybrid energy storage system. Otherwise, the speed and position of each particle are updated, and the lower-level constraints are adjusted according to the updated particle position parameters, and the lower-level model is solved again.

A hybrid energy storage system capacity and power configuration system, characterized by comprising:

The modeling module is configured to construct a two-layer optimization model with the goal of minimizing the comprehensive cost of the hybrid energy storage system. The upper optimization model is used to solve the capacity configuration problem of the hybrid energy storage system, and the lower optimization model is used to solve the operation problem of the photovoltaic-hybrid energy storage system;

The solution module is configured to use a deep deterministic policy gradient algorithm to solve the lower-level optimization model to obtain the operating cost, and based on the operating cost, use a particle swarm algorithm to solve the upper-level optimization model to obtain the optimal power and capacity configuration scheme of the supercapacitor and the battery.

A computer-readable storage medium, characterized in that: a plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the steps in the above method.

A terminal device, characterized in that it includes a processor and a computer-readable storage medium, the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the steps in the above method.

Compared with the prior art, the present invention has the following beneficial effects:

Compared with the traditional hybrid energy storage power and capacity configuration method, the present invention uses supercapacitors and batteries to form a hybrid energy storage system, suppresses short-term photovoltaic power fluctuations by setting power fluctuation costs, and introduces time-of-use electricity prices and abandoned light costs, taking into account the economic efficiency of photovoltaic consumption and photovoltaic-hybrid energy storage systems. Compared with existing methods, the present invention simultaneously considers the smoothing of photovoltaic power fluctuations and the optimized operation of photovoltaic-hybrid energy storage systems.

The lower-level optimization model of the present invention simulates the operation of the photovoltaic-hybrid energy storage system. Due to the small time scale of the operation optimization, the model construction is complex and difficult to solve. Therefore, the present invention adopts a deep deterministic policy gradient algorithm, takes photovoltaic, load power data and time-of-use electricity prices as algorithm inputs, and solves the optimization problem by converting it into a Markov decision model, thus overcoming the problems of traditional algorithms being difficult to solve and having poor convergence. In addition, the input data set of the deep deterministic policy gradient algorithm consists of one year of photovoltaic minute-level output data. Compared with the optimization operation method based on a typical day, the present invention fully considers the uncertainty of photovoltaic output, making the hybrid energy storage capacity and power configuration results more in line with actual needs.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

FIG1 is a schematic diagram of a Markov decision model in an example verification of the present invention;

FIG2 is a schematic diagram of the principle of a deep deterministic policy gradient algorithm in an example verification of the present invention;

FIG3 is a schematic diagram of the algorithm flow of solving the lower layer optimization model based on the particle swarm-deep deterministic policy gradient algorithm in the example verification of the present invention;

FIG4 is a schematic diagram of photovoltaic power fluctuations on a certain day before and after hybrid energy storage is configured in an example verification of the present invention;

FIG. 5 is a schematic diagram of the charge state of the battery and the supercapacitor in the example verification of the present invention.

DETAILED DESCRIPTION

1 to 5 are the best embodiments of the present invention, and the present invention will be further described below in conjunction with FIGS. 1 to 5 .

Example 1

Taking into account the installation conditions and environmental requirements of energy storage equipment, a hybrid energy storage system is composed of popular supercapacitors and batteries. Supercapacitors are used to smooth short-term photovoltaic power fluctuations, and batteries are used to participate in system operation optimization to improve system economy.

The technical solution of the present invention aims to minimize the sum of the construction cost and the operating cost of the hybrid energy storage system, and proposes a double-layer optimization model for the hybrid energy storage power and capacity configuration. The upper optimization model solves the capacity configuration problem of the hybrid energy storage system, and the lower optimization model solves the operation problem of the photovoltaic-hybrid energy storage system.

A method for configuring capacity and power of a hybrid energy storage system comprises the following steps:

With the goal of minimizing the comprehensive cost of the hybrid energy storage system, a two-layer optimization model is constructed. The upper optimization model is used to solve the capacity and power configuration problems of the hybrid energy storage system, and the lower optimization model is used to solve the operation problems of the photovoltaic-hybrid energy storage system.

A deep deterministic policy gradient algorithm is used to solve the lower-level optimization model to obtain the operating cost. Based on the operating cost, a particle swarm algorithm is used to solve the upper-level optimization model to obtain the optimal power and capacity configuration scheme of the supercapacitor and battery.

The upper optimization model objective function is:

The upper optimization model uses the configuration capacity of battery energy storage and supercapacitor energy storage as decision variables, and aims to minimize the sum of the construction cost and operating cost of the photovoltaic-hybrid energy storage system:

minC _total =C _invest +C _op ; (1)

Where C _total is the total cost of the photovoltaic-hybrid energy storage system; C _invest is the investment and construction cost of the hybrid energy storage system; C _op is the annual operating cost of the photovoltaic-hybrid energy storage system, which is given by the lower optimization model.

The investment and construction cost of the hybrid energy storage system is expressed as:

Cinvest＝Ccons,ba+Ccons,sc; (2)

Where r is the discount rate; n _ba and n _sc represent the planned years of battery energy storage and supercapacitor energy storage respectively; P _ba and P _sc represent the rated power of battery energy storage and supercapacitor energy storage respectively; S _ba and S _sc represent the planned capacity of battery energy storage and supercapacitor energy storage respectively; c _ba1 and c _ba2 represent the unit power and capacity investment and construction cost of battery energy storage respectively; c _sc1 and c _sc2 represent the unit power and capacity investment and construction cost of battery energy storage respectively.

The upper-level optimization model specifically includes the following constraints:

The energy storage configuration power constraint is expressed as:

0≤P _ba ≤P _ba,max ； (5)

0≤P _sc ≤P _sc,max ； (6)

Where P _ba,max and P _sc,max represent the configurable upper limits of battery energy storage and supercapacitor energy storage power, respectively.

The energy storage configuration capacity constraint is expressed as:

0≤S _ba ≤S _ba,max ； (7)

Where S _ba,max and S _sc,max represent the upper limits of the configurable battery energy storage and supercapacitor energy storage capacity, respectively.

The objective function of the lower optimization model is:

Based on the upper optimization model, the lower optimization model aims to minimize the annual operating cost of the system, and achieves the goals of reducing system operating costs, reducing abandoned light, and smoothing power fluctuations at the grid connection point by controlling the charging and discharging power of the hybrid energy storage. The specific objective function is:

minC _op ＝C _op,pv,hess +C _buy +C _cut,pv +C _penalty (9)

Where C _op,pv,hess is the output cost of the photovoltaic-hybrid energy storage system, C _buy is the electricity purchase cost, C _cut,pv is the cost of abandoned light, and C _penalty is the power fluctuation penalty cost. Each cost is described in detail below.

The output cost of the photovoltaic-hybrid energy storage system is:

分别表示光伏、电池储能、超级电容储能单位出力成本；

分别表示光伏、电池储能、超级电容储能运行功率。In the formula,

Respectively represent the unit output costs of photovoltaic, battery energy storage, and supercapacitor energy storage;

Respectively represent the operating power of photovoltaic, battery energy storage, and supercapacitor energy storage.

The cost of purchasing electricity is:

表示单位购电成本，

表示上级电网供电功率。In the formula,

represents the unit electricity purchase cost,

Indicates the power supply of the upper power grid.

The cost of abandoning light is:

为单位弃光成本，由于本专利要求光伏发电就地消纳，不允许返送电网，因此上级电网供电功率

若

则表示该部分光伏发电功率无法被系统消纳，需要支付弃光成本。In the formula,

is the unit cost of abandoned light. Since this patent requires that photovoltaic power be consumed locally and is not allowed to be fed back to the grid, the power supply power of the upper grid is

like

This means that this part of the photovoltaic power generation cannot be absorbed by the system and the cost of abandonment needs to be paid.

The power fluctuation penalty cost is:

Wherein, ω is the power fluctuation penalty coefficient; ΔP _t represents the power fluctuation amount of the grid connection point within the Δt time; ΔP _lim represents the maximum limit of power fluctuation. If |ΔP _t |＞ΔP _lim , the power fluctuation penalty cost needs to be paid.

The lower-level optimization model constraints specifically include the following constraints:

The power balance constraint is:

为t时刻负载功率。In the formula,

is the load power at time t.

The battery energy storage charging and discharging power constraints are:

表示电池储能充放电功率上限，由上层优化模型配置的电池储能额定功率决定。In the formula,

Indicates the upper limit of battery energy storage charging and discharging power, which is determined by the rated power of the battery energy storage configured by the upper-level optimization model.

The battery energy storage charging and discharging ramp power constraint is:

表示电池储能最大充放电爬坡功率。In the formula,

Indicates the maximum charging and discharging ramp power of battery energy storage.

The battery energy storage capacity constraint is:

In the formula, E _ba,min and E _ba,max represent the minimum and maximum values of the amount of electricity allowed to be stored in the battery energy storage, respectively, which are determined by the battery energy storage capacity configured by the upper-level optimization model; E _ba,ini and E _ba,end represent the electricity at the initial and end times of the battery energy storage, respectively.

The supercapacitor energy storage power constraint is:

表示超级电容储能充放电功率上限，由上层优化模型配置的超级电容储能额定功率决定。In the formula,

Indicates the upper limit of supercapacitor energy storage charging and discharging power, which is determined by the supercapacitor energy storage rated power configured by the upper optimization model.

The supercapacitor energy storage capacity constraint is:

Where E _sc,min and E _sc,max represent the minimum and maximum values of the amount of electricity allowed to be stored by the supercapacitor energy storage, respectively, which are determined by the supercapacitor energy storage capacity configured by the upper optimization model; E _sc,ini and E _sc,end represent the amount of electricity at the initial and end times of the supercapacitor energy storage, respectively.

The model solving method based on particle swarm-deep deterministic policy gradient algorithm includes the following steps:

The particle swarm algorithm is:

The particle swarm algorithm is used for the upper optimization problem. Formula (1) is set as the fitness function of the particle swarm algorithm. The operating cost is provided by the lower optimization model. The power and capacity of battery energy storage and supercapacitor energy storage are set as the position of the particle swarm. The particle swarm algorithm is used to solve the minimum comprehensive cost of the hybrid energy storage system.

The Markov decision model is:

In order to transform the underlying optimization problem into a problem that can be solved by deep reinforcement learning methods, we first need to build a Markov decision model. The Markov decision model is shown in Figure 1.

Among them, s _t represents the state observed by the agent in the environment at time t, which is defined as:

Among them, a _t represents the action of the agent at time t, which is defined as:

Among them, r( _at | _st ) is the reward value fed back by the environment under the action of _at , which is defined as:

In the formula, K is a constant. Through the reward function r(a _t |s _t ), the problem of minimizing the operating cost of the system at the lower level is transformed into the problem of maximizing the reward function.

The cumulative reward _Rt corresponding to an exploration process starting from state _st can be expressed as:

R _t =r(a _t |s _t )+γr(a _t+1 |s _t+1 )+…+γ ^Tt r(a _T |s _T ); (24)

Where γ is the discount coefficient.

The action-value function ^Qπ (s,a) represents the expected cumulative reward obtained by following the strategy π in an exploration process of performing action _at in state _st , specifically:

Q ^π (s,a)=E ^π [R _t |s _t =s,a _t =a]; (25)

The deep deterministic policy gradient algorithm is:

The principle of the deep deterministic policy gradient algorithm (DDPG) is shown in Figure 2. The DDPG algorithm consists of four fully connected neural networks, namely the Actor network μ(s|θ ^μ ), the Actor target network μ'(s|θ ^μ '), the Critic network Q(s,a|θ ^Q ), and the Critic target network Q'(s,a|θ ^Q ').

The DDPG algorithm adopts the experience replay mechanism to store the explored samples (s _t , a _t , r _t , st ₊₁ ) in the experience replay pool and continuously update them. The Actor network and the Critic network extract small batches of samples from the experience replay pool for training and update parameters to maximize the reward, thereby obtaining the optimal operating state of the hybrid energy storage system.

The Actor network determines the action a _t performed by the agent in state s _t , specifically:

为随机噪声。Where θ ^μ is the parameter of the Actor network,

is random noise.

The specific update method of its network parameter θ ^μ is:

Where N is the number of random sampling batches, Q(s _i ,a _i |θ ^Q ) is the output of the Critic network, and α ^μ is the learning rate of the Actor network.

The Critic network is used to estimate the action value function Q(s _i ,a _i |θ ^Q ) to evaluate the strategy generated by the Actor network. The parameter θ ^Q is updated as follows:

y _i =r _i +γQ'(s _i+1 ,μ'(s _i+1 |θ ^μ ')|θ ^Q '); (29)

Where μ'(s _i+1 |θ ^μ' ) is the output of the Actor target network, Q'(s _i+1 ,μ'(s _i+1 |θ ^μ' )|θ ^Q' ) is the output of the Critic target network, and α ^Q is the learning rate of the Critic network.

The Actor target network and the Critic target network update network parameters through soft update, specifically:

θ ^μ' ←τθ ^μ +(1-τ)θ ^μ' ; (32)

θ ^Q' ←τθ ^Q +(1-τ)θ ^Q' ; (33)

Where θ ^μ' is the parameter of the Actor target network, θ ^Q' is the parameter of the Critic target network, and τ is the soft update coefficient.

The model solving process based on the particle swarm-deep deterministic policy gradient algorithm is as follows:

The model solving algorithm flow based on the particle swarm-deep deterministic policy gradient algorithm is shown in Figure 3, which includes the following steps:

S1 initializes the particle swarm parameters, and the power and capacity of the battery and supercapacitor are used as the position parameters of each particle;

S2 initializes the DDPG environment, loads PV output data and load data from the data set, and initializes the state space;

The S3DDPG algorithm calculates the current action a _t based on the current state space s _t , which is the input/output power of the battery energy storage and supercapacitor energy storage.

The S4 environment is updated according to the data at the next moment, and the reward r _t of the action in S3 is calculated to obtain the state space s _t+1 at the next moment;

S5 stores (s _t ,a _t ,r _t ,s _t+1 ) into the experience replay pool, randomly samples from the experience replay pool for training DDPG, and updates the parameters of the Actor network and the Critic network according to equations (28) and (31);

S6 updates the parameters of the Actor target network and the Critic target network according to equations (32) and (33);

If the S7 environment is loaded with the photovoltaic and load data at the set time, the termination condition of this training is met, the accumulated reward is calculated, and the process jumps to S8, otherwise it jumps to S3;

If the maximum number of training times is reached in S8, the current round of training is terminated, and the annual operating cost is calculated using the hybrid energy storage charging and discharging control strategy obtained through training and the process jumps to S9; otherwise, the process jumps to S2;

S9 calculates the fitness value of each particle based on the annual operating cost calculated by DDPG and obtains the global optimum. If the maximum number of iterations is reached, the minimum total cost obtained by the PSO-DDPG algorithm is output. Otherwise, the speed and position of each particle are updated and jump to S2.

In this embodiment, the specific algorithm analysis is as follows:

Case Study

The present invention uses the output data of a photovoltaic demonstration project in a certain place for one year as an example to analyze and verify the feasibility and effectiveness of the proposed method. The installed capacity of the photovoltaic demonstration project is 1MW, the data sampling time interval is 1 minute, and 6:00-18:00 every day is regarded as a charge and discharge cycle of the hybrid energy storage system. The parameter settings in the hybrid energy storage configuration strategy of the present invention are shown in Tables 1, 2, and 3.

Table 1 Battery and supercapacitor parameters

Table 2 Time-of-use electricity prices

Table 3 DDPG algorithm hyperparameters

Comparison of different energy storage configuration strategies

In order to verify the necessity of configuring hybrid energy storage, the following four energy storage configuration schemes are compared: no energy storage, only battery energy storage, only supercapacitor energy storage, and the hybrid energy storage configuration strategy proposed by the present invention. The results obtained from different energy storage configuration schemes are shown in Table 4.

Table 4 Comparison of different energy storage configuration strategies

As can be seen from Table 4, the total cost converted to one year after configuring hybrid energy storage is the lowest. Although the installation cost of hybrid energy storage is the highest, the configuration of hybrid energy storage significantly reduces the power fluctuation penalty cost and the cost of abandoning light. This is because supercapacitors suppress power fluctuations through rapid charging and discharging, reducing the power fluctuation penalty cost. In addition, the battery participates in operation scheduling, and stores excess electricity when the photovoltaic power is greater than the load, reducing the cost of abandoning light. When the photovoltaic output is low and the electricity price is high, the battery releases electricity to meet the load demand, further reducing the cost of purchasing electricity.

When no energy storage is configured, the power fluctuation of photovoltaic power generation is large due to the influence of cloud obstruction, and the power fluctuation penalty cost of the system is high. At the same time, since there is no energy storage to participate in the consumption of new energy, when the photovoltaic output is higher than the load, the penalty cost of abandoning light increases. When the photovoltaic output is lower than the load, the load demand can only be met by purchasing electricity, which increases the cost of purchasing electricity. When only batteries are configured, the battery's adjustment of photovoltaic power fluctuations is relatively limited due to the battery's climbing power constraint, so the penalty cost is high. When only supercapacitors are configured, due to the high cost per unit capacity of supercapacitors, supercapacitors are difficult to participate in cross-time scheduling, so the cost of abandoning light and the cost of purchasing electricity are high.

By comparing the four configuration strategies, it can be seen that the hybrid energy storage system can combine the advantages of both power-type energy storage and energy-type energy storage, and has obvious advantages in smoothing the power fluctuations of renewable energy power generation and cross-time scheduling.

Analysis of system operation optimization results:

Taking the photovoltaic output curve of a certain day as an example, Figure 4 shows the photovoltaic power fluctuation before and after the configuration of hybrid energy storage. It can be seen from the figure that the photovoltaic power fluctuation has been significantly improved after the configuration of hybrid energy storage. When energy storage is not configured, the power fluctuation of the grid-connected point is large and frequent. The high-frequency and instantaneous power fluctuation with large amplitude is suppressed by the rapid charging and discharging of supercapacitors. At the same time, battery energy storage also participates in the suppression of some low-frequency power fluctuations. Figure 5 shows the charge state of the battery and supercapacitor on that day. The supercapacitor has frequent charge and discharge state transitions, while the battery energy storage has fewer charge and discharge state transitions, which is conducive to extending the service life of battery energy storage and reducing the operation and maintenance costs of hybrid energy storage. At the same time, battery energy storage absorbs electricity when the photovoltaic output is higher than the load at noon, and releases the stored electricity when the electricity price is high, reducing the cost of purchasing electricity and improving the economy of the photovoltaic-hybrid energy storage system.

The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims (10)

1. A method for configuring capacity and power of a hybrid energy storage system, characterized in that it comprises the following steps: With the goal of minimizing the comprehensive cost of the hybrid energy storage system, a two-layer optimization model is constructed. The upper optimization model is used to solve the capacity and power configuration problems of the hybrid energy storage system, and the lower optimization model is used to solve the operation problems of the photovoltaic-hybrid energy storage system. A deep deterministic policy gradient algorithm is used to solve the lower-level optimization model to obtain the operating cost. Based on the operating cost, a particle swarm algorithm is used to solve the upper-level optimization model to obtain the optimal power and capacity configuration scheme of the supercapacitor and battery. 2. The capacity and power configuration method of the hybrid energy storage system according to claim 1 is characterized in that the comprehensive cost of the hybrid energy storage system includes investment and construction costs and operating costs, wherein the investment and construction costs include the power and capacity investment and construction costs of supercapacitors and the power and capacity investment and construction costs of batteries, and the operating costs include the output costs of photovoltaic-hybrid energy storage systems, electricity purchase costs, abandoned light costs, and power fluctuation penalty costs. 3. The capacity and power configuration method of the hybrid energy storage system according to claim 1 is characterized in that: the constraints of the upper optimization model include supercapacitor energy storage configuration power and capacity constraints, battery energy storage configuration power and capacity constraints; The constraints of the lower-level optimization model include power balance constraints, battery energy storage charging and discharging power constraints, battery energy storage charging and discharging climbing power constraints, battery energy storage power constraints, supercapacitor energy storage charging and discharging power constraints, and supercapacitor energy storage power constraints. 4. The capacity and power configuration method of the hybrid energy storage system according to claim 1 is characterized in that: the specific process of using the deep deterministic policy gradient algorithm to solve the lower-level optimization model is: the objective function of the lower-level optimization model is the annual operating cost of the hybrid energy storage system, and the photovoltaic and load power data within the historical set time are used as input, the optimization problem is converted into a Markov decision model, and the deep deterministic policy gradient algorithm is used to solve it. 5. The capacity and power configuration method of the hybrid energy storage system according to claim 4 is characterized in that: the action space of the Markov decision model is the current charging and discharging power of the supercapacitor and the battery, the state space is the photovoltaic power, load power, supercapacitor state of charge, battery state of charge and electricity price information at the current moment, and the reward function is obtained by converting the objective function of the lower-level optimization model, and the objective function minimization problem is converted into the cumulative reward function maximization problem. 6. The capacity and power configuration method of the hybrid energy storage system according to claim 4 or 5 is characterized in that: a deep deterministic policy gradient algorithm is used to solve the Markov decision model, and four fully connected neural networks are constructed, the four fully connected neural networks are Actor network, Actor target network, Critic network and Critic target network, the Actor network calculates the current action space, the environment feeds back the current state space and reward function, and calculates the cumulative reward function; The current action, state, reward function and the state at the next moment are stored in the experience replay pool, and random sampling is used from the experience replay pool for parameter update of the Actor network and the Critic network. The parameters of the Actor target network and the Critic target network are updated using the soft update method. The optimal hybrid energy storage charging and discharging control strategy is obtained through continuous training, and the operating cost is calculated. 7. The hybrid energy storage system capacity and power configuration method according to claim 1 is characterized in that: the process of solving the upper optimization model includes the following steps: The upper model initializes the particle swarm parameters, and the power and capacity of the battery and supercapacitor are used as the position parameters of each particle; Adjust the lower-level constraints according to the particle position parameters, solve the lower-level optimization model, initialize the deep deterministic policy gradient algorithm environment, train the hybrid energy storage charging and discharging control strategy, and then calculate the annual operating cost; According to the obtained annual operating cost, the fitness value of each particle is calculated and the global optimum is obtained. If the maximum number of iterations of the upper layer is reached, the minimum total cost is output. The position parameters of the global optimal particle are the optimal configuration of the capacity and power of the hybrid energy storage system. Otherwise, the speed and position of each particle are updated, and the lower-level constraints are adjusted according to the updated particle position parameters, and the lower-level model is solved again. 8. A hybrid energy storage system capacity and power configuration system, characterized by: comprising: The modeling module is configured to construct a two-layer optimization model with the goal of minimizing the comprehensive cost of the hybrid energy storage system. The upper optimization model is used to solve the capacity configuration problem of the hybrid energy storage system, and the lower optimization model is used to solve the operation problem of the photovoltaic-hybrid energy storage system; The solution module is configured to use a deep deterministic policy gradient algorithm to solve the lower-level optimization model to obtain the operating cost, and based on the operating cost, use a particle swarm algorithm to solve the upper-level optimization model to obtain the optimal power and capacity configuration scheme of the supercapacitor and the battery. 9. A computer-readable storage medium, characterized in that: a plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the steps in the method according to any one of claims 1 to 7. 10. A terminal device, characterized in that it includes a processor and a computer-readable storage medium, the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the steps in the method described in any one of claims 1 to 7.

Images