Virtual power plant master-slave game-based electrical changing station participation electric energy market optimization scheduling method and system

Abstract

The invention relates to a virtual power plant master-slave game-based method for participating in optimal scheduling of an electric energy market by a battery swap station. The method comprises the following steps: step 1, collecting historical data of wind power and photovoltaic power generation by a virtual power plant; step 2, predicting wind power and photovoltaic power output of the second day by using a CNN-LSTM-based day-ahead wind and light output prediction model; 3, the virtual power plant sets the confidence level of the conditional value-at-risk; 4, the virtual power plant predicts the conditions of power over-generation, over-use, insufficient supply and the like in the area; and step 5, the virtual power plant analyzes the adjustment capability and the standby capability of the power generation side and the power utilization side, establishes and solves a real-time scheduling model based on the virtual power plant master-slave game, and obtains a charging and discharging service strategy. The method can fully consider the wind and light uncertainty, the charging and discharging requirements of the battery swap station and the power balance requirements of the power system.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system optimization and dispatching, and relates to a method and system for optimizing and dispatching battery swap stations participating in an electric energy market, and in particular to a method and system for optimizing and dispatching battery swap stations participating in an electric energy market based on a master-slave game of a virtual power plant. Background Art

[0002] With the large-scale access of new energy power generation in my country, the strong randomness and strong dispersion of power generation and consumption have made it more difficult for the power system to ensure the balance of power generation and consumption. The power system needs to be frequently dispatched to maintain the balance of power supply and demand, which in turn increases the complexity of power system optimization and dispatch. As part of the distributed energy storage system, the battery swap station needs to be efficiently coordinated with other distributed energy sources such as photovoltaic power generation and wind power generation. However, due to the large differences in the characteristics of various distributed energy sources, the complexity of coordinated dispatch has increased significantly.

[0003] Existing dispatching methods still face great challenges in the coordinated optimization of multi-energy systems, and it is difficult to achieve the global optimal energy configuration and dispatching. In addition, the optimal dispatching of battery swap stations relies on the real-time collection and processing of multi-source heterogeneous data, including the output of new energy sources, the load status of the power grid, meteorological data, etc. However, the processing capacity of the existing battery swap station participating in the optimization dispatching system of the electric energy market is still insufficient, and it is difficult to achieve efficient processing and accurate analysis of massive, real-time, and high-frequency data, which in turn affects the timeliness and accuracy of dispatching decisions.

[0004] With the rapid increase in the number of electric vehicles, the battery swapping infrastructure is also expanding and upgrading. As an important facility for electric vehicle energy replenishment, the increase in the number of battery swapping stations not only meets the growing charging demand, but also provides key dispatching resources for the power system.

[0005] Therefore, the present invention proposes a method and system for optimizing the scheduling of battery swap stations participating in the electric energy market based on the master-slave game of virtual power plants. Through reasonable optimization and scheduling, the battery swap station can serve as a distributed energy storage unit, releasing electric energy during peak electricity demand and storing electric energy during low demand, thereby smoothing the peak-to-valley difference of the power grid, promoting the consumption of new energy, and improving the operating efficiency and stability of the power system.

[0006] After searching, no public documents of the prior art that are identical or similar to the present invention were found. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and propose an optimal scheduling method for battery swap stations participating in the electric energy market based on the master-slave game of virtual power plant (VPP), which can fully consider the uncertainty of wind and solar power, the charging and discharging requirements of battery swap stations, and the power balance requirements of the power system.

[0008] The present invention solves the practical problem by adopting the following technical solutions:

[0009] A method for optimizing the dispatching of a power exchange station participating in an electric energy market based on a master-slave game of a virtual power plant comprises the following steps:

[0010] Step 1: In the day-ahead stage, the virtual power plant collects historical data on wind power and photovoltaic power generation, and the virtual power plant collects historical power generation and consumption information of internal market entities;

[0011] Step 2: Based on the historical data of wind power and photovoltaic power generation collected in step 1, the virtual power plant uses the CNN-LSTM-based day-ahead wind and photovoltaic output prediction model to predict the wind power and photovoltaic output of the next day.

[0012] Step 3: In the real-time stage, the virtual power plant sets the confidence level of the conditional risk value based on the forecast results of wind and solar power output in step 2;

[0013] Step 4: Based on the confidence level of the conditional risk value set in step 3, the virtual power plant obtains the power generation and consumption information of the region and the power demand of the day, and combines the historical power generation and consumption information to predict the possible power over-generation, over-consumption, and insufficient supply in the region;

[0014] Step 5. Based on the prediction results of step 4, the virtual power plant analyzes the regulation capacity and backup capacity of the power generation side and the power consumption side, establishes and solves the real-time scheduling model based on the master-slave game of the virtual power plant, obtains the charging and discharging service strategy, and guides the battery swap station to charge when the grid load is low and discharge when the grid load is high, so as to balance the supply and demand of electricity.

[0015] Moreover, the specific method of step 1 is:

[0016] In the previous stage, the virtual power plant collected historical data of wind power and photovoltaic power generation, including meteorological data and power generation data. The meteorological data included wind speed, temperature, humidity, and light intensity.

[0017] The virtual power plant collects the historical power generation and consumption information of internal market entities, including: the output on the power generation side, the basic load on the power consumption side, the standby capacity and the charging and discharging power of the battery swap station.

[0018] Moreover, the specific steps of step 3 include:

[0019] (1) Definition: Value at Risk (VaR) is the loss quantile at a specific confidence level. The formula is as follows:

[0020]

[0021] α β =min{α|P(s,x,α)≥β} (13)

[0022] Where: x∈X, X is a feasible decision set; y∈R m , represents the random factors that affect returns; the probability density function of the random variable y is p(y); the risk value at the confidence level β is α β ;

[0023] (2) Conditional Value at Risk not only focuses on the maximum potential loss at a specific confidence level, but also considers the average loss when the loss exceeds VaR. The definition of CVaR is:

[0024]

[0025] Where: (f(s,x,y)-α β ) + =max{f(s,x,y)-α β ,0},ψ β is the conditional value at risk;

[0026] (3) In order to solve ψ β , discretize equation (15), and we can get:

[0027]

[0028] Where: q is the number of samples; y is i is the mth sample of the random variable y;

[0029] (4) The CVaR value considering the uncertainty of wind and solar power is shown in formula (17).

[0030]

[0031] Where: VPP is the CVaR value of VPP; α β,VPP is the VaR value of VPP; i,s is the value in VPP that exceeds VaR. Relax equation (17) into equation (18) and equation (19):

[0032] z i,s ≥0 (18)

[0033]

[0034] Moreover, the specific steps of step 5 include:

[0035] (1) Establishing an upper-level virtual power plant optimization model:

[0036] 1) Objective function

[0037] Considering the charging and discharging service fees, electricity sales revenue, electricity purchase cost, operation and maintenance cost and conditional risk value of the virtual power plant, an upper-level virtual power plant optimization model is constructed, and the maximum profit of the virtual power plant is taken as the objective function of the upper-level model, as shown in formula (20):

[0038] maxR vpp =R vpp_Grid +R vpp_bss -C vpp_loss -C vpp_Grid -ξψ VPP (20)

[0039] Where: R vpp is the overall benefit of the virtual power plant, R vpp_Grid is the revenue gained by the virtual power plant from participating in the electricity market, R vpp_bss The charging and discharging service fee earned by the virtual power plant, C vpp_loss is the operation and maintenance cost of the virtual power plant, C vpp_Grid is the cost of purchasing electricity of the virtual power plant, ξ is the risk aversion coefficient of the VPP, and the larger the ξ is, the more the VPP is averse to potential risks;

[0040] The virtual power plant purchases electricity from the electricity market when the electricity price is low, and supplies electricity to the electricity market when the electricity price is high, making a profit through the price difference between the two, as shown in equations (21)-(23).

[0041]

[0042] Where: R vpp_Grid The benefits of virtual power plants participating in the electricity energy market; They are the time-of-use electricity prices in the electricity market during period t; is the power output and input of the internal resource n of the virtual power plant to the electric energy market during period t; The total charging and discharging demand reported by the battery swap station during period t;

[0043] The virtual power plant considers the load of the power grid and the charging and discharging demand of the battery swap station, formulates the charging and discharging service price of the battery swap station, and earns profits by providing charging and discharging services to the battery swap station and earning service fees. The charging and discharging service income can be expressed by formula (24):

[0044]

[0045] Where: R vpp_bss The income from VPP charging and discharging services. They represent the charging and discharging prices set by VPP during period t, They represent the upward and downward power of the VPP internal regulation resource n during period t. pv,t ,p wt,t , Respectively represent the output power of photovoltaic, wind power and energy storage;

[0046] Operation and maintenance costs of virtual power plants C vpp_loss As shown in formula (25):

[0047]

[0048] Where: m pv 、m wt and m es are the operation and maintenance cost coefficients of photovoltaic power, wind power and energy storage respectively; P pv,t and P wt,t are the output of photovoltaic and wind power respectively; Δt is the scheduling step; C ES The dispatch cost of energy storage; are the charging and discharging power of the energy storage at time t; η ES,cha ,η ES,dis are the charging and discharging efficiencies of the energy storage at time t, respectively.

[0049] 2) Constraints

[0050] 1. Adjust power constraints

[0051] The regulation service of the virtual power plant cannot exceed the upper and lower limits of the maximum output;

[0052]

[0053] Where: They are the upward and downward power regulation of the virtual power plant respectively. and are the maximum power for charging and discharging at the battery swap station, respectively;

[0054] 2. Electricity price constraints

[0055] The price of regulation services provided by the virtual power plant cannot exceed the minimum and maximum price limits for that period;

[0056]

[0057] Where: ξ is the price volatility coefficient;

[0058] 3. Output constraints

[0059]

[0060] Where: is the charging state and charging power of the energy storage during period t; are the discharge state and discharge power of energy storage respectively;

[0061] 4. Power balance constraints:

[0062]

[0063] 5. Adjust resource power mutual exclusion constraints:

[0064]

[0065] (2) Establishing an optimization model for lower-level battery swap stations

[0066] 1) Objective function

[0067] The battery swap station adjusts its charging and discharging behavior based on the charging and discharging service price set by the virtual power plant to minimize the total charging and discharging cost of the battery swap station in the region; the objective function is shown in formula (30):

[0068]

[0069] Where: θ dis is the unit discharge loss cost of EV discharge.

[0070] 2) Constraints

[0071] 1. Charge and discharge power constraints

[0072]

[0073] Where: They are the upper limits of charging and discharging power of the battery swap station respectively.

[0074] 2. Power demand mutual exclusion constraint

[0075] The battery swap station cannot report charging and discharging requirements at the same time during any period;

[0076]

[0077] 3. Battery capacity constraints

[0078] E m,min ≤SOC m ≤u m E m,max (37)

[0079] (3) The particle swarm algorithm is used to solve the proposed model. The electricity price is regarded as an intelligent agent, which guides the charging and discharging power of the battery swap station to approach the objective function of minimizing the charging and discharging cost, and then obtains the charging and discharging service strategy.

[0080] Advantages and beneficial effects of the present invention:

[0081] 1. The master-slave game scheduling strategy of the virtual power plant proposed in the present invention can guide the battery swap station to charge when the grid load is low and discharge when the grid load is high, so as to balance the supply and demand of electricity and significantly smooth the peak-to-valley difference of the grid load. This not only improves the operating efficiency of the power system, but also reduces the instability problem of the power grid caused by load fluctuations. Through optimized scheduling, the power system can use existing resources more efficiently and reduce the pressure and operating costs caused by peak loads. In addition, this method can also improve the reliability and stability of the power grid, reduce the risks caused by load fluctuations, and ensure the continuity and stability of power supply.

[0082] 2. The present invention can effectively promote the consumption of new energy sources such as wind power and photovoltaic power by optimizing the charging and discharging behavior of the battery swap station, and reduce the problem of wind and solar power abandonment caused by excess electricity during off-peak hours. During peak hours, the power demand is met by discharging at the battery swap station, which reduces the load pressure on the power grid. At the same time, by reasonably scheduling the charging time of the battery swap station, the power consumption during off-peak hours is increased, and the utilization rate of new energy is improved.

[0083] 3. The present invention uses convolutional neural networks (CNN) and long short-term memory networks (LSTM) models to efficiently predict the output of wind power and photovoltaic power. Through these advanced artificial intelligence technologies, valuable information can be extracted from massive, real-time, high-frequency data, improving the efficiency and accuracy of data processing and analysis. This method not only improves the timeliness and accuracy of scheduling decisions, but also better copes with the uncertainty of renewable energy generation, provides reliable data support for the power system, and optimizes scheduling decisions.

[0084] 4. Through reasonable charging and discharging optimization scheduling, the present invention can significantly reduce the charging and discharging costs of electric vehicle users. The battery swap station charges when the electricity price is low and discharges when the electricity price is high, which not only reduces the electricity cost of users, but also reduces the overall operating cost by optimizing the grid load management. This method enables users to use electric vehicles at a lower cost, increases the attractiveness of electric vehicles, and also reduces the operating pressure of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 It is a process flow chart of the method of the present invention;

[0086] Figure 2 It is a method framework diagram of the present invention;

[0087] Figure 3 Schematic diagram of the unit structure of the CNN model of the present invention;

[0088] Figure 4 Schematic diagram of the internal memory unit structure of the LSTM model of the present invention;

[0089] Figure 5It is a schematic diagram of the structure of the CNN-LSTM combined model of the present invention;

[0090] Figure 6 The particle swarm algorithm solution flow chart of the present invention;

[0091] Figure 7 It is a schematic diagram of wind power, photovoltaic and energy storage output of the present invention;

[0092] Figure 8 A schematic diagram of the purchase and sale of electricity by the virtual power plant of the present invention;

[0093] Fig. 9 This is a schematic diagram of the charging and discharging behavior of the battery swap station of the present invention;

[0094] Fig.10 This is a comparison diagram of power grid loads in different scenarios of the present invention. DETAILED DESCRIPTION

[0095] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings:

[0096] A method for optimizing the dispatching of a power exchange station participating in an electric energy market based on a master-slave game of a virtual power plant comprises the following steps:

[0097] Step 1: In the day-ahead stage, the virtual power plant collects historical data on wind power and photovoltaic power generation, and the virtual power plant collects historical power generation and consumption information of internal market entities;

[0098] The specific method of step 1 is:

[0099] In the previous stage, the virtual power plant collected historical data of wind power and photovoltaic power generation, including meteorological data and power generation data. The meteorological data included wind speed, temperature, humidity, and light intensity.

[0100] The virtual power plant collects the historical power generation and consumption information of internal market entities, including: the output on the power generation side, the basic load on the power consumption side, the standby capacity and the charging and discharging power of the battery swap station.

[0101] Step 2: Based on the historical data of wind power and photovoltaic power generation collected in step 1, the virtual power plant uses the CNN-LSTM-based day-ahead wind and photovoltaic output prediction model to predict the wind power and photovoltaic output of the next day.

[0102] In this embodiment, in the day-ahead stage, the virtual power plant predicts the wind power and photovoltaic output for the next day based on the collected wind power and photovoltaic power generation data using a convolutional neural network (CNN)-long short-term memory network (LSTM) combined model.

[0103] The day-ahead wind and solar power output prediction model based on CNN-LSTM is as follows:

[0104] 1) Convolutional Neural Network Model

[0105] CNN has a powerful feature extraction capability and can automatically extract multi-scale spatial features from raw data. CNN uses local perception and weight sharing mechanisms to perform multi-layer convolution and pooling operations on input data to extract high-level features of the data layer by layer. In wind and solar power output forecasting, historical output data and meteorological data (such as wind speed, light intensity, temperature, etc.) usually have a high degree of temporal and spatial correlation. The unit structure of the CNN model is as follows: Figure 3 shown.

[0106] 2) Long Short-Term Memory Neural Network Model

[0107] Long short-term memory network (LSTM) is a special recurrent neural network designed to process and predict long-term dependencies in time series data. LSTM overcomes the gradient vanishing and gradient exploding problems of traditional RNN in the long sequence training process by introducing the forget gate, input gate and output gate mechanism, and can more accurately capture the dynamic change characteristics of wind power and photovoltaic output data. LSTM mainly consists of three stages:

[0108] 1. Forgetting stage: determines which information in the cell state needs to be discarded. By setting the sigmoid function, the forgetting gate receives the hidden state of the previous moment and the current input, outputs a value between 0 and 1, and selectively retains or deletes each information element in the characteristic data such as wind speed and light intensity.

[0109] 2. Update phase: determines how the current input information updates the cell state. The input gate is also calculated through a sigmoid function and a tanh function. The sigmoid function determines which information needs to be updated, and the tanh function generates a new candidate information vector to add important information to the cell state, such as the current wind speed change, light intensity change, etc.

[0110] 3. Output stage: determines how the cell state affects the hidden state at the current moment, thereby affecting the final output. The output gate calculates the weight through a sigmoid function, and then adjusts the cell state through the tanh function to output the hidden state at the current moment. The hidden state reflects the predicted value of wind and solar power output at the current moment.

[0111] The relevant calculation formulas for each layer are shown in formulas (1)-(6).

[0112] f t =σ(W f ×[h t-1 ,xt ]+b f ) (1)

[0113] i t =σ(W i ×(x t ,h t-1 )+b i ) (2)

[0114] C t =tanh(W c ×(x t ,h t-1 )+b c ) (3)

[0115] U t =i t ×C t +f t ×C t-1 (4)

[0116] O t =σ(W o ×(x t ,h t-1 )+b o ) (5)

[0117] h t =O t ×tanh(U t ) (6)

[0118] Where: f t 、i t , O t , C t They are forget gate, input gate, output gate and control gate respectively; W f , W i , W o Respectively represent the weight matrix of each gate; b f , b i , b o are the bias values ​​of each door; U t 、h t , σ are the unit state, hidden layer output and Sigmoid function respectively; tanh is the activation function. The unit structure of the LSTM model is as follows Figure 4 shown.

[0119] In the CNN module, the input time series data such as wind speed and light intensity are processed through convolutional layers and pooling layers to extract their spatial features. These feature maps contain important patterns and trends in the original data, such as fluctuations in wind speed changes and periodicity in light intensity. These extracted spatial features are then input into the LSTM module.

[0120] In the LSTM module, the temporal dynamic changes in the data are captured through its unique memory unit and gating mechanism. The LSTM layer processes and stores time series information, such as historical changes in wind speed and light intensity, thereby enhancing the ability to predict future wind and solar power output. The output data processed by LSTM is further processed by the fully connected layer to generate the final prediction result, which is usually the wind and solar power generation at a certain moment or multiple moments in the future.

[0121] The whole process includes data preprocessing, feature extraction, time series modeling and final prediction output. In order to ensure the efficiency and accuracy of the model, historical wind and solar output data are used for model training, and the model parameters are optimized through the back propagation algorithm. The output calculation of one-dimensional convolution for time series feature extraction is shown in formula (7).

[0122] C=σ(W×X+b) (7)

[0123] In the formula, C is the extracted feature, σ is the activation function, W is the weight matrix, X is the time series, and b is the bias vector. By combining the convolutional neural network (CNN) to extract spatial features and the long short-term memory network (LSTM) to capture temporal dynamic features, the CNN-LSTM model can effectively process the complex patterns in wind and solar output data and improve the accuracy and robustness of the prediction. The structure of the CNN-LSTM combined model is as follows: Figure 5 shown.

[0124] After obtaining the prediction results, in order to evaluate the prediction effect of the model, MAE, RMSE and MAPE are used to evaluate the prediction accuracy of the model. The calculation process is shown in formulas (8)-(10) respectively.

[0125]

[0126] In the formula, n is the number of data in the test set, is the predicted carbon emission trading price result of the model, y i is the true value of carbon emission trading price data.

[0127] The root mean square error is more sensitive to abnormal data, so it can evaluate the degree of dispersion of the predicted data to the real data, thereby judging the fitting effect of the predicted data. The smaller the RMSE value, the closer the predicted data curve of the model is to the real data curve; the mean absolute error and the mean absolute percentage error can avoid the mutual cancellation of errors, which can reflect the accuracy of the prediction. The smaller the values ​​of MAE and MAPE, the higher the prediction accuracy of the model. The weighted average is used to obtain the comprehensive prediction accuracy, as shown in formula (11).

[0128]

[0129] Step 3: In the real-time stage, the virtual power plant sets the confidence level of the conditional risk value based on the forecast results of wind and solar power output in step 2;

[0130] The specific steps of step 3 include:

[0131] The main risks brought by wind and solar uncertainty are reflected in the volatility of power generation and forecast errors. In order to deal with these risks, the power system needs to consider the potential worst-case losses. Value at Risk (VaR) is a risk measurement method that represents the maximum loss that may be incurred within a given confidence level and time frame.

[0132] (1) Definition: Value at Risk (VaR) is the loss quantile at a specific confidence level. The formula is as follows:

[0133]

[0134] α β =min{α|P(s,x,α)≥β} (13)

[0135] Where: x∈X, X is a feasible decision set; y∈R m , represents the random factors that affect returns; the probability density function of the random variable y is p(y); the risk value at the confidence level β is α β ;

[0136] (2) Conditional Value at Risk (CVaR) not only focuses on the maximum potential loss at a specific confidence level, but also considers the average loss when the loss exceeds VaR. The definition of CVaR is:

[0137]

[0138] Where: (f(s,x,y)-α β ) + =max{f(s,x,y)-α β ,0},ψ β is the conditional value at risk;

[0139] (3) In order to solve ψ β , discretize equation (15), and we can get:

[0140]

[0141] Where: q is the number of samples; y is i is the mth sample of the random variable y;

[0142] Wind power and photovoltaic power generation are subject to natural conditions and have significant randomness and volatility. This uncertainty poses a challenge to the planning and scheduling of power systems. In this case, relying solely on VaR may not fully reflect the risks in extreme situations. Therefore, using CVaR as a risk measurement tool can more effectively describe the uncertainty of wind power and photovoltaic power generation and provide a more robust risk management tool. Using the prediction result accuracy of the CNN-LSTM combined model as the confidence level of the conditional value at risk (CVaR), further quantifying and managing the risks brought by prediction errors can provide strong support for the stable operation and optimized scheduling of power systems.

[0143] (4) The CVaR value considering the uncertainty of wind and solar power is shown in formula (17).

[0144]

[0145] Where: VPP is the CVaR value of VPP; α β,VPP is the VaR value of VPP; i,s is the value in VPP that exceeds VaR. Relax equation (17) into equation (18) and equation (19):

[0146] z i,s ≥0 (18)

[0147]

[0148] Step 4: Based on the confidence level of the conditional risk value set in step 3, the virtual power plant obtains the power generation and consumption information of the region and the power demand of the day by time, and then combines the historical power generation and consumption information to predict the possible power overgeneration, overconsumption, and insufficient supply in the region.

[0149] Step 5. Based on the prediction results of step 4, the virtual power plant analyzes the regulation capacity and backup capacity of the power generation side and the power consumption side, establishes and solves the real-time scheduling model based on the master-slave game of the virtual power plant, obtains the charging and discharging service strategy, and guides the battery swap station to charge when the grid load is low and discharge when the grid load is high, so as to balance the supply and demand of electricity.

[0150] The virtual power plant analyzes the regulation and backup capabilities of the power generation side and the power consumption side, formulates corresponding charging and discharging strategies, and guides the battery swap stations to charge when the grid load is low and discharge when the grid load is high, so as to balance the supply and demand of electricity.

[0151] The specific steps of step 5 include:

[0152] (1) Establishing an upper-level virtual power plant optimization model:

[0153] 1) Objective function

[0154] Considering the charging and discharging service fees, electricity sales revenue, electricity purchase costs, operation and maintenance costs and conditional risk value of the virtual power plant, an upper-level virtual power plant optimization model is constructed, and the maximum profit of the virtual power plant is taken as the objective function of the upper-level model, as shown in Formula (20).

[0155] maxR vpp =R vpp_Grid +R vpp_bss -C vpp_loss -C vpp_Grid -ξψ VPP (20)

[0156] Where: R vpp is the overall benefit of the virtual power plant, R vpp_Grid is the revenue gained by the virtual power plant from participating in the electricity market, R vpp_bss The charging and discharging service fee earned by the virtual power plant, C vpp_loss is the operation and maintenance cost of the virtual power plant, C vpp_Grid is the cost of purchasing electricity for the virtual power plant, ξ is the risk aversion coefficient of VPP, and the larger the ξ is, the more VPP averse to potential risks.

[0157] The virtual power plant purchases electricity from the electricity market when the electricity price is low, and supplies electricity to the electricity market when the electricity price is high, making a profit through the price difference between the two, as shown in equations (21)-(23).

[0158]

[0159] Where: R vpp_Grid The benefits of virtual power plants participating in the electricity energy market; They are the time-of-use electricity prices in the electricity market during period t; is the power output and input of the internal resource n of the virtual power plant to the electric energy market during period t; The total charging and discharging demand reported by the battery swap station during period t;

[0160] The virtual power plant considers the load of the power grid and the charging and discharging demand of the battery swap station, formulates the charging and discharging service price of the battery swap station, and earns profits by providing charging and discharging services to the battery swap station and earning service fees. The charging and discharging service income can be expressed by formula (24):

[0161]

[0162] Where: R vpp_bss The income from VPP charging and discharging services. They represent the charging and discharging prices set by VPP during period t, They represent the upward and downward power of the VPP internal regulation resource n during period t. pv,t ,p wt,t, Respectively represent the output power of photovoltaic, wind power and energy storage.

[0163] Operation and maintenance costs of virtual power plants C vpp_loss As shown in formula (25).

[0164]

[0165] Where: m pv 、m wt and m es are the operation and maintenance cost coefficients of photovoltaic power, wind power and energy storage respectively; P pv,t and P wt,t are the output of photovoltaic and wind power respectively; Δt is the scheduling step; C ES The dispatch cost of energy storage; are the charging and discharging power of the energy storage at time t; η ES,cha ,η ES,dis are the charging and discharging efficiencies of the energy storage at time t, respectively.

[0166] 2) Constraints

[0167] 1. Adjust power constraints

[0168] The regulation service of the virtual power plant cannot exceed the upper and lower limits of the maximum output.

[0169]

[0170] Where: They are the upward and downward power regulation of the virtual power plant respectively. and The maximum power for charging and discharging at the battery swap station respectively.

[0171] 2. Electricity price constraints

[0172] The price of regulation services provided by the virtual power plant cannot exceed the minimum and maximum price limits for that period.

[0173]

[0174] Where: ξ is the price volatility coefficient.

[0175] 3. Output constraints

[0176]

[0177] Where: is the charging state and charging power of the energy storage during period t; are the discharge state and discharge power of the energy storage respectively.

[0178] 4. Power balance constraints

[0179]

[0180] 5. Adjust resource power mutual exclusion constraints

[0181]

[0182] (2) Establishing an optimization model for lower-level battery swap stations

[0183] 1) Objective function

[0184] The battery swap station adjusts its charging and discharging behavior based on the charging and discharging service price set by the virtual power plant to minimize the total charging and discharging cost of the battery swap station in the region. The objective function is shown in formula (30).

[0185]

[0186] Where: θ dis is the unit discharge loss cost of EV discharge.

[0187] 2) Constraints

[0188] 1. Charge and discharge power constraints

[0189]

[0190] Where: They are the upper limits of charging and discharging power of the battery swap station respectively.

[0191] 2. Power demand mutual exclusion constraint

[0192] The battery swap station cannot report charging and discharging demands at the same time during any period of time.

[0193]

[0194] 3. Battery capacity constraints

[0195] E m,min ≤SOC m ≤u m E m,max (37)

[0196] (3) The particle swarm algorithm is used to solve the proposed model. The electricity price is regarded as an intelligent agent, which guides the charging and discharging power of the battery swap station to approach the objective function of minimizing the charging and discharging cost, and then obtains the charging and discharging service strategy.

[0197] The particle swarm algorithm solution process is as follows Figure 6 shown.

[0198] Step 6. The battery swap station performs charging and discharging operations at appropriate times according to the charging and discharging service strategy formulated by the virtual power plant, thereby smoothing out the peak and valley differences in the power grid, promoting the consumption of new energy, and improving the operating efficiency and stability of the power system.

[0199] A system for optimizing the dispatching of power exchange stations participating in the electric energy market based on the master-slave game of a virtual power plant, comprising:

[0200] Data collection module: In the day-ahead phase, the virtual power plant collects historical data on wind power and photovoltaic power generation, and the virtual power plant collects historical power generation and consumption information of internal market entities;

[0201] Prediction module: The virtual power plant uses the CNN-LSTM-based day-ahead wind and photovoltaic output prediction model to predict the wind and photovoltaic output of the next day based on the collected historical data of wind power and photovoltaic power generation;

[0202] The confidence level setting module of the conditional risk value, in the real-time stage, the virtual power plant sets the confidence level of the conditional risk value based on the prediction results of the wind and solar power output in step 2;

[0203] The power over-generation, over-usage and under-supply prediction module, based on the confidence level of the conditional risk value set, after obtaining the power generation and consumption information of the region and the power demand of the day by time, combines the historical power generation and consumption information to predict the power over-generation, over-usage and under-supply in the region;

[0204] The charging and discharging service strategy acquisition module analyzes the regulation and backup capabilities of the power generation and power consumption sides based on the prediction results, establishes and solves a real-time scheduling model based on the master-slave game of the virtual power plant, and obtains the charging and discharging service strategy.

[0205] The present invention is further described below by means of specific examples:

[0206] Taking the charging and discharging pilot project of battery swap stations in a certain area as an example, the battery swap stations based on the master-slave game of virtual power plants are simulated to participate in the optimal scheduling of the electricity market.

[0207] (1) In the day-ahead stage, the virtual power plant collects historical data on wind power and photovoltaic power generation, including meteorological data (such as wind speed, temperature, humidity, light intensity, etc.) and power generation data. In addition, the virtual power plant collects historical power generation and consumption information of internal market entities, such as the output of the power generation side, the basic load on the power consumption side, the backup capacity, the charging and discharging power of the battery swap station, etc. In order to simplify the calculation, the present invention selects 2×1.5MW wind turbines, 6×0.5MW photovoltaic generators, 2×0.5MW energy storage, and 300 batteries in 3 battery swap stations. The capacity of a single battery involved in the scheduling is 65kWh, the charging and discharging efficiency is 95%, the initial state of charge is 20kWh, the charging and discharging power is ±12.5kW, the utility function value is 90%, and the off-grid SOC value is 90%. The equipment-related parameters are as shown in the following table.

[0208] Table 1 Equipment related parameters

[0209]

[0210]

[0211] The present invention divides the dispatching cycle into 24 hours, and the time-of-use electricity price data is shown in the following table.

[0212] Table 2 Time-of-use electricity prices

[0213]

[0214] Based on the above basic data, the present invention conducts case analysis on the proposed method. Matlab (R2022b) software is used to model and solve the problem. The particle swarm related parameters are shown in the following table.

[0215] Table 3 Particle swarm algorithm related parameters

[0216]

[0217] (2) The virtual power plant predicts the wind power and photovoltaic power output of the next day based on the collected wind power and photovoltaic power generation data. After obtaining the prediction results, MAE, RMSE and MAPE are used to evaluate the prediction accuracy of the model.

[0218] (3) In the real-time stage, the virtual power plant sets the confidence level of the conditional risk value based on the forecast results of wind and solar power output. After obtaining the power generation and consumption information of the region and the power demand of the day, the virtual power plant predicts the possible power over-generation, over-consumption, and insufficient supply in the region based on the historical power generation and consumption trend. VPP gives priority to wind power and photovoltaic output. However, as the load increases, the power generation output of wind power and photovoltaic power gradually cannot meet all the power demand on the load side. The remaining demand needs to be met by the discharge of energy storage units and battery swap stations. This mode can promote the consumption of wind power and photovoltaic power, solve the problem of excessive power abandonment during the valley period, and the problem of difficulty in ensuring power supply during peak hours. This operating mode helps to promote the consumption of wind power and photovoltaic power, solve the problem of wind power abandonment caused by excessive power during the valley period, and the problem of difficulty in power supply during peak hours.

[0219] Wind power, photovoltaic and energy storage output Figure 7 As shown in Figure 2, the charging and discharging behavior of the battery swap station is as follows: Fig. 9 As shown in Figure 2, the VPP's power grid purchase and sales are as follows: Figure 8 As shown in the figure, when there is large-scale wind power generation at night from 0:00 to 8:00 and sufficient renewable energy generation during the day from 11:00 to 18:00, the VPP sells excess electricity to the grid to earn profits.

[0220] (4) The battery swap station performs charging and discharging operations at appropriate times according to the guidance of the virtual power plant, thereby smoothing out the peak and valley differences in the power grid, promoting the consumption of new energy, and improving the operating efficiency and stability of the power system. Fig. 9 shown.

[0221] The load peak-valley difference, charging and discharging cost of the battery swap station and VPP income are selected as evaluation indicators to evaluate the load management effect under different scenarios, as shown in the following table. Fig.10 As shown, scenario 1 does not perform optimized scheduling, and scenario 2 uses the optimized scheduling method of the present invention to guide the charging and discharging behavior of the battery swap station. Compared with scenario 1, scenario 2 can effectively reduce the load peak-to-valley difference by 11.82%, and the charging and discharging cost of the battery swap station is reduced by 28.79%. Through example analysis, it can be concluded that the present invention can significantly reduce the peak-to-valley difference of the power grid load and the electricity cost of EV users, and at the same time can effectively promote the consumption of wind power and photovoltaic power, solve the problem of wind and solar power abandonment caused by excess electricity during the valley, and the problem of power supply difficulties during peak hours.

[0222] Table 4 Comparison of load levels in different scenarios

[0223]

[0224] It should be emphasized that the embodiments of the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation modes. Any other implementation modes derived by those skilled in the art based on the technical solutions of the present invention also fall within the scope of protection of the present invention.

Claims

1. A method for optimizing the dispatching of power exchange stations participating in the electric energy market based on the master-slave game of a virtual power plant, characterized by: The following steps are involved: Step 1: In the day-ahead stage, the virtual power plant collects historical data on wind power and photovoltaic power generation, and the virtual power plant collects historical power generation and consumption information of internal market entities; Step 2: The virtual power plant uses the day-ahead wind and photovoltaic output prediction model based on CNN-LSTM to predict the wind power and photovoltaic output of the next day based on the historical data of wind power and photovoltaic power generation collected in step 1; Step 3: In the real-time stage, the virtual power plant sets the confidence level of the conditional risk value based on the forecast results of wind and solar power output in step 2; Step 4: Based on the confidence level of the conditional risk value set in step 3, the virtual power plant obtains the power generation and consumption information of the region and the power demand of the day, and combines the historical power generation and consumption information to predict the power over-generation, over-consumption, and insufficient supply in the region; Step 5: Based on the prediction results of step 4, the virtual power plant analyzes the regulation capacity and backup capacity of the power generation side and the power consumption side, establishes and solves the real-time scheduling model based on the master-slave game of the virtual power plant, and obtains the charging and discharging service strategy.

2. According to claim 1, a method for optimizing the dispatching of power exchange stations participating in the electric energy market based on the master-slave game of a virtual power plant, characterized in that: The specific method of step 1 is: In the previous stage, the virtual power plant collected historical data of wind power and photovoltaic power generation, including meteorological data and power generation data. The meteorological data included wind speed, temperature, humidity, and light intensity. The virtual power plant collects the historical power generation and consumption information of internal market entities, including: the output on the power generation side, the basic load on the power consumption side, the standby capacity and the charging and discharging power of the battery swap station.

3. According to claim 1, a method for optimizing the dispatching of power exchange stations participating in the electric energy market based on the master-slave game of a virtual power plant, characterized in that: The specific steps of step 3 include: (1) Definition: Value at Risk (VaR) is the loss quantile at a specific confidence level. The formula is as follows: a β =min{α|P(s,x,α)≥β} (13) Where: x∈X, X is a feasible decision set; y∈R m , represents the random factors that affect returns; the probability density function of the random variable y is p(y); the risk value at the confidence level β is α β ; (2) Conditional Value at Risk not only focuses on the maximum potential loss at a specific confidence level, but also considers the average loss when the loss exceeds VaR. The definition of CVaR is: Where: (f(s,x,y)-α β ) + =max{f(s,x,y)-α β ,0},ψ β is the conditional value at risk; (3) In order to solve ψ β , discretize equation (15), and we can get: Where: q is the number of samples; y is i is the mth sample of the random variable y; (4) The CVaR value considering the uncertainty of wind and solar power is shown in formula (17); Where: VPP is the CVaR value of VPP; α β,VPP is the VaR value of VPP; i,s is the value in VPP that exceeds VaR; relax equation (17) into equation (18) and equation (19): z i,s ≥0 (18) 4. According to claim 1, a method for optimizing the dispatching of power exchange stations participating in the electric energy market based on the master-slave game of a virtual power plant, characterized in that: The specific steps of step 5 include: (1) Establishing an upper-level virtual power plant optimization model: 1) Objective function Considering the charging and discharging service fees, electricity sales revenue, electricity purchase cost, operation and maintenance cost and conditional risk value of the virtual power plant, an upper-level virtual power plant optimization model is constructed, and the maximum profit of the virtual power plant is taken as the objective function of the upper-level model, as shown in formula (20): maxR vpp =R vpp_Grid +R vpp_bss -C vpp_loss -C vpp_Grid -ξψ VPP (20) Where: R vpp is the overall benefit of the virtual power plant, R vpp_Grid is the revenue gained by the virtual power plant from participating in the electricity market, R vpp_bss The charging and discharging service fee earned by the virtual power plant, C vpp_loss is the operation and maintenance cost of the virtual power plant, C vpp_Grid is the cost of purchasing electricity of the virtual power plant, ξ is the risk aversion coefficient of the VPP, and the larger the ξ is, the more the VPP is averse to potential risks; The virtual power plant purchases electricity from the electricity market when the electricity price is low, and delivers electricity to the electricity market when the electricity price is high, making a profit through the price difference between the two, as shown in equations (21)-(23); Where: R vpp_Grid The benefits of virtual power plants participating in the electricity energy market; are the time-of-use electricity prices in the electricity market during period t; is the power output and input of the internal resource n of the virtual power plant to the electric energy market during period t; The total charging and discharging demand reported by the battery swap station during period t; The virtual power plant considers the load of the power grid and the charging and discharging demand of the battery swap station, formulates the charging and discharging service price of the battery swap station, and earns profit by providing charging and discharging services to the battery swap station and earning service fees; the charging and discharging service income can be expressed by formula (24): Where: R vpp_bss The income from VPP charging and discharging services. They represent the charging and discharging prices set by VPP during period t, They represent the upward and downward power of the VPP internal regulation resource n during period t respectively; Respectively represent the output power of photovoltaic, wind power and energy storage; Operation and maintenance costs of virtual power plants C vpp_loss As shown in formula (25): Where: m pv 、m wt and m es are the operation and maintenance cost coefficients of photovoltaic power, wind power and energy storage respectively; P pv,t and P wt,t are the output of photovoltaic and wind power respectively; Δt is the scheduling step; C ES The dispatch cost of energy storage; are the charging and discharging power of the energy storage at time t respectively; η ES,cha ,η ES,dis are the charging and discharging efficiencies of the energy storage at time t respectively; 2) Constraints 1. Adjust power constraints The regulation service of the virtual power plant cannot exceed the upper and lower limits of the maximum output; Where: They are the power increase and power decrease of the virtual power plant respectively; and are the maximum power for charging and discharging at the battery swap station, respectively; 2. Electricity price constraints The price of regulation services provided by the virtual power plant cannot exceed the minimum and maximum price limits for that period; Where: ξ is the price volatility coefficient; 3. Output constraints Where: is the charging state and charging power of the energy storage during period t; are the discharge state and discharge power of energy storage respectively; 4. Power balance constraints:

5. Adjust resource power mutual exclusion constraints: (2) Establishing an optimization model for lower-level battery swap stations 1) Objective function The battery swap station adjusts its charging and discharging behavior based on the charging and discharging service price set by the virtual power plant to minimize the total charging and discharging cost of the battery swap station in the region; the objective function is shown in formula (30): Where: θ dis is the unit discharge loss cost of EV discharge; 2) Constraints 1. Charge and discharge power constraints Where: They are the upper limits of charging and discharging power of the battery swap station; 2. Power demand mutual exclusion constraint The battery swap station cannot report charging and discharging requirements at the same time at any time; 3. Battery capacity constraints AND m,min ≤SOC m ≤u m AND m,max (37) (3) The particle swarm algorithm is used to solve the proposed model. The electricity price is regarded as an intelligent agent, which guides the charging and discharging power of the battery swap station to approach the objective function of minimizing the charging and discharging cost, and then obtains the charging and discharging service strategy.

5. A system for optimizing the dispatching of power exchange stations participating in the electric energy market based on the master-slave game of virtual power plants, characterized by: include: Data collection module: In the day-ahead phase, the virtual power plant collects historical data on wind power and photovoltaic power generation, and the virtual power plant collects historical power generation and consumption information of internal market entities; Prediction module: The virtual power plant uses the CNN-LSTM-based day-ahead wind and photovoltaic output prediction model to predict the wind and photovoltaic output of the next day based on the collected historical data of wind power and photovoltaic power generation; The confidence level setting module of the conditional risk value, in the real-time stage, the virtual power plant sets the confidence level of the conditional risk value based on the prediction results of the wind and solar power output in step 2; The power over-generation, over-usage and under-supply prediction module, based on the confidence level of the conditional risk value set, after obtaining the power generation and consumption information of the region and the power demand of the day by time, combines the historical power generation and consumption information to predict the power over-generation, over-usage and under-supply in the region; The charging and discharging service strategy acquisition module analyzes the regulation and backup capabilities of the power generation and power consumption sides based on the prediction results, establishes and solves a real-time scheduling model based on the master-slave game of the virtual power plant, and obtains the charging and discharging service strategy.

Images