Power distribution network load response regulation and control method and system based on data driving

Abstract

The invention relates to the technical field of power distribution network load response regulation and control, in particular to a power distribution network load response regulation and control method and system based on data driving. Firstly, power distribution network load data, charging station data and user behavior data are acquired; then, charging station data and user behavior data are input into a charging station load prediction module, and an output charging station load prediction value is corrected by using electricity price and environment evaluation information to obtain a final charging station load prediction value; then, acquiring a V2G discharge power prediction value by using a reinforcement learning-based V2G discharge power prediction module, and combining the V2G discharge power prediction value with the final charge station load prediction value to obtain a load reduction amount prediction value of the power distribution network; and finally, optimizing charging station load regulation and control by utilizing a power distribution network load response regulation and control module and combining the final charging station load prediction value, the load reduction quantity prediction value and the power distribution network load data. The effectiveness of load response regulation and control of the power distribution network can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of load response control of distribution network, and in particular to a data-driven load response control method and system for distribution network. Background Art

[0002] With the gradual popularization of electric vehicles, charging stations will become an important part of the power supply load source of the distribution network, which will cause a short-term peak in power demand in the distribution network, bringing severe challenges to the safe, stable and economical operation of the power grid. Therefore, distribution network load response regulation for charging stations is essential.

[0003] The existing technology still has some shortcomings in the load response control method of the distribution network based on charging stations. On the one hand, with the emergence of bidirectional charging stations, the existing technology only considers the impact of the charging power of the charging station on the load of the distribution network during the load control of the charging station, and does not see the regulating effect of the discharge power on the overall load; on the other hand, the existing technology does not optimize the load regulation of the distribution network based on the user-side charging demand obtained by analyzing the user behavior data, which will reduce the effectiveness of the load response control of the distribution network.

[0004] Therefore, a data-driven distribution network load response control method and system are proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a data-driven distribution network load response control method and system. First, the distribution network load data, charging station data and user behavior data are obtained; then, the charging station data and user behavior data are input into the charging station load prediction module, and the output charging station load prediction value is corrected using the electricity price and environmental assessment information to obtain the final charging station load prediction value; then, the V2G discharge power prediction value is obtained using the reinforcement learning-based V2G discharge power prediction module, and it is combined with the final charging station load prediction value to obtain the load reduction amount prediction value of the distribution network; finally, the distribution network load response control module is used and combined with the final charging station load prediction value, the load reduction amount prediction value and the distribution network load data to optimize the charging station load control. The present invention can improve the effectiveness of the distribution network load response control.

[0006] To achieve the above object, the present invention provides the following technical solutions: A data-driven distribution network load response control method, comprising: Obtain distribution network load data, charging station data, and user behavior data; Inputting the charging station data and the user behavior data into a charging station load prediction module to obtain a charging station load prediction value; modifying the charging station load prediction value using electricity price and environmental assessment information to obtain a final charging station load prediction value; According to the charging station data and the user behavior data, a V2G discharge power prediction module based on reinforcement learning is used to perform prediction to obtain a V2G discharge power prediction value; Obtaining a load reduction prediction value of the distribution network according to the final charging station load prediction value and the V2G discharge power prediction value; The final charging station load forecast value and the load reduction forecast value are input into the distribution network load response control module, and the charging station load control is optimized in combination with the distribution network load data, the charging station load real-time data and the charging station load real-time reduction amount.

[0007] Furthermore, the charging station data includes: charging station basic data, charging pile data, environmental data and electricity price data; the charging station basic data includes: charging station ID, geographical location, number of charging piles and type of charging piles; the charging pile data includes: charging pile status data, charging pile electrical parameter data, charging pile charging and discharging data and battery SOC data; the charging and discharging data includes: charging and discharging time, charging and discharging power and charging and discharging power; the user behavior data includes: user preference data, charging behavior data, user travel mode data and user discharge willingness data.

[0008] Furthermore, the process of obtaining the final charging station load prediction value by using the charging station load prediction module includes: Constructing the charging station load prediction module, training it using historical charging station data, historical user behavior data, and historical load data to obtain a pre-trained charging station load prediction module; Inputting the charging pile data in the charging station data and the charging behavior data and the user travel mode data in the user behavior data into the pre-trained charging station load prediction module to obtain the charging station load prediction value; Input environmental data, electricity price data and user preference data into the load impact assessment module to obtain the load impact assessment weight of the charging station; The charging station load prediction value is weighted by using the charging station load impact assessment weight to obtain the final charging station load prediction value.

[0009] Furthermore, according to the charging station data and the user behavior data, a V2G discharge power prediction module based on reinforcement learning is used to perform prediction, and the process of obtaining the V2G discharge power prediction value includes: Constructing the V2G discharge power prediction module based on reinforcement learning, and defining the state space, action space and reward function of the intelligent agent according to the charging station data and the user behavior data; The V2G discharge power prediction module is trained using a reinforcement learning algorithm and historical data, and the model parameters are updated using a gradient descent algorithm to obtain a pre-trained model; The real-time data of the charging station and the real-time user behavior data are input into the pre-training model to obtain the V2G discharge power prediction value.

[0010] Furthermore, the process of optimizing the load control of the charging station by using the distribution network load response control module includes: Compare the final charging station load prediction value with the V2G discharge power prediction value to obtain a load reduction prediction value of the distribution network; Compare the final charging station load prediction value and the load reduction amount prediction value with the charging station load real-time data and the charging station load real-time reduction amount, respectively, to obtain a load prediction error and a reduction amount prediction error; The objective function of the distribution network load response control module is constructed by using the distribution network load data, the load prediction error and the reduction amount prediction error; the charging station load constraint, the V2G constraint and the distribution network constraint are used as the constraint conditions of the distribution network load response control module; The load control of the charging station is optimized by combining the distribution network load response control module and the optimization algorithm.

[0011] A data-driven distribution network load response control system, comprising: a distribution network control data acquisition unit, a charging station load prediction unit, a V2G discharge power prediction unit, a charging station load reduction prediction unit and a distribution network load response control unit; The distribution network control data acquisition unit is used to obtain distribution network load data, charging station data and user behavior data; The charging station load prediction unit is used to input the charging station data and the user behavior data into the charging station load prediction module to obtain a charging station load prediction value; and to modify the charging station load prediction value using electricity price and environmental assessment information to obtain a final charging station load prediction value; The V2G discharge power prediction unit is used to perform prediction based on the charging station data and the user behavior data using a V2G discharge power prediction module based on reinforcement learning to obtain a V2G discharge power prediction value; The charging station load reduction amount prediction unit is used to obtain a load reduction amount prediction value of the distribution network according to the final charging station load prediction value and the V2G discharge power prediction value; The distribution network load response control unit is used to input the final charging station load prediction value and the load reduction amount prediction value into the distribution network load response control module, and optimize the charging station load control in combination with the distribution network load data, the charging station load real-time data and the charging station load real-time reduction amount.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention proposes a charging station load forecasting method for assisting the distribution network in regulating the charging station load response; the method uses a charging station load forecasting module to process the charging station data and user behavior data to obtain a charging station load forecast value; then, the charging station load forecast value is weighted by using the charging station load impact assessment weight obtained by the load impact assessment module to obtain a final charging station load forecast value; the method takes into account the impact of the environment, electricity prices and user preferences on the charging station load, so that it can adapt to different scenarios and needs, thereby improving the effectiveness of the distribution network load response regulation.

[0013] 2. The present invention proposes a V2G discharge power prediction method for providing data support for distribution network control optimization; this method uses a V2G discharge power prediction module based on reinforcement learning for prediction, which can continuously learn through the interaction between the intelligent agent and the environment, and can obtain the operating law of V2G discharge power by combining charging station data and user behavior data. In this way, the load reduction of the charging station can be more accurately predicted, thereby improving the effectiveness of load response control of the distribution network.

[0014] 3. The present invention proposes a distribution network load response control method for optimizing the control of the distribution network on the charging station load response; the method constructs a multi-objective function of the distribution network load response control module, and incorporates the load prediction error and the reduction amount prediction error into the distribution network load response control module to cope with the complex distribution network operating environment, thereby improving the robustness of the control; and adaptively adjusts the control strategy according to real-time data and the changing environment, thereby improving the effectiveness of the distribution network load response control. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a flow chart of a data-driven distribution network load response control method of the present invention; Figure 2 It is a structural schematic diagram of the charging station load prediction module of the present invention; Figure 3 It is a structural schematic diagram of a data-driven distribution network load response control system of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] See also Figures 1 to 3 The present invention provides a data-driven distribution network load response control method and system, and the technical solution is as follows: Embodiment 1 In order to optimize the regulation of the distribution network's response to the charging station load, a company used a data-driven distribution network load response regulation method proposed in the present invention. The process of the method is shown in the following figure: Figure 1 As shown, specifically including: Obtain distribution network load data, charging station data, and user behavior data; Furthermore, the distribution network load data and charging station data are obtained through backend calls from the distribution automation system and the charging station management system, respectively; the user behavior data is obtained based on the third-party APP platform and user surveys; Furthermore, charging station data includes: charging station basic data, charging pile data, environmental data and electricity price data; charging station basic data includes: charging station ID, geographical location, number of charging piles and charging pile type; charging pile data includes: charging pile status data, charging pile electrical parameter data, charging pile charging and discharging data and battery SOC data; charging and discharging data includes: charging and discharging time, charging and discharging power and charging and discharging power; user behavior data includes: user preference data, charging behavior data, user travel mode data and user discharge willingness data; Furthermore, the environmental data includes load-affecting environmental data and non-load-affecting environmental data; the load-affecting environmental data includes: temperature, precipitation, humidity, wind speed, etc., and the non-load-affecting environmental data includes: air quality index, ultraviolet index, air pressure, etc.; Furthermore, user preference data include: charging price preference, charging time preference and energy demand preference; charging behavior data include: charging frequency, single charging time, single charging power, charging time period, etc.; user travel mode data include: travel frequency, travel route, appearance time, etc.

[0018] By introducing charging station data and user behavior data, a solid data foundation can be provided for charging station load forecasting and V2G discharge power forecasting, ensuring the rationality of the forecast results; and combining it with distribution network load data can improve the effectiveness of distribution network load response regulation.

[0019] Input charging station data and user behavior data into the charging station load prediction module to obtain the charging station load prediction value; use electricity price and environmental assessment information to correct the charging station load prediction value to obtain the final charging station load prediction value; Furthermore, the process of obtaining the final charging station load forecast value includes: Construct a charging station load prediction module, and train it using historical charging station data, historical user behavior data, and historical load data to obtain a pre-trained charging station load prediction module; Input the charging pile data in the charging station data and the charging behavior data and user travel mode data in the user behavior data into the pre-trained charging station load prediction module to obtain the charging station load prediction value; Input environmental data, electricity price data and user preference data into the load impact assessment module to obtain the load impact assessment weight of the charging station; The charging station load impact assessment weight is used to weight the charging station load forecast value to obtain the final charging station load forecast value; Furthermore, the input data of the charging station load prediction module, i.e., charging pile data, charging behavior data, and user travel pattern data, are divided into time scale data, space scale data, and user scale data; Furthermore, the charging station load prediction module is a multi-dimensional feature attention network, the structure of which is as follows: Figure 2 As shown, it includes: input layer, feature attention mechanism layer, feature fusion layer, recurrent neural network layer and fully connected output layer; Furthermore, the input layer of the charging station load prediction module is used to map the multidimensional data into feature vectors to obtain time features, spatial features and user features; the feature attention mechanism layer is used to use time attention, spatial attention and user attention to process the time features, spatial features and user features respectively to obtain the time features, spatial features and user features after attention weighting; the feature fusion layer is used to use the global attention mechanism to learn the importance of different types of features and perform adaptive fusion; the recurrent neural network layer is used to use the LSTM network to extract the temporal dependency from the fused features to improve the accuracy of load prediction; the fully connected output layer is used to use the fully connected layer to convert the output of the recurrent neural network layer into a charging station load prediction value; By using the attention mechanism to weight the time scale data, space scale data and user scale data respectively, the importance of different time points, different spaces and different users can be learned from the data, thereby improving the accuracy of charging station load prediction.

[0020] Furthermore, the process of the load impact assessment module obtaining the load impact assessment weight of the charging station is as follows: using a multi-layer perceptron to extract the features of the input data; using channel splicing to fuse the extracted different features; using a fully connected layer to convert the fused features into a weight value, and then using the Sigmoid function to limit the weight to the range of [0,1].

[0021] By using the charging station load prediction module to predict the charging station load, and combining the attention mechanism to learn the importance of different types of features, and adaptively fusing these features, the prediction accuracy is further improved; the impact of the environment, electricity prices and user preferences on the charging station load is also considered, so that it can adapt to different scenarios and needs, thereby improving the effectiveness of distribution network load response regulation.

[0022] According to the charging station data and user behavior data, the V2G discharge power prediction module based on reinforcement learning is used to predict the V2G discharge power prediction value; Furthermore, according to the charging station data and user behavior data, a V2G discharge power prediction module based on reinforcement learning is used to make a prediction, and the process of obtaining the V2G discharge power prediction value includes: Construct a V2G discharge power prediction module based on reinforcement learning, and define the state space, action space and reward function of the agent according to the charging station data and user behavior data; The V2G discharge power prediction module is trained using a reinforcement learning algorithm and historical data, and the model parameters are updated using a gradient descent algorithm to obtain a pre-trained model. Input the real-time data of charging stations and real-time user behavior data into the pre-training model to obtain the V2G discharge power prediction value; Furthermore, the state space includes: time, electricity price, battery SOC, user willingness and charging station status; the action space includes V2G discharge power and V2G subsidy electricity price; the reward function includes: charging station revenue, user satisfaction and new energy consumption rate; Furthermore, the reinforcement learning algorithm can adopt Q-Learning algorithm, DQN algorithm and PPO algorithm; Furthermore, the training process of the V2G discharge power prediction module includes: initializing the V2G environment, selecting an action using the ε-greedy strategy according to the current state information of the V2G system; sending the selected action to the V2G environment for execution, and the environment will update the state according to the state transition rule to obtain the reward and the next state; storing the state, action, reward and next state in the experience pool, and randomly extracting a batch of experience from it; updating the model parameters using the extracted experience and the reinforcement learning algorithm; updating the current state to the next state to prepare for the next iteration; judging whether the loop termination condition is met, if so, jumping out of the loop, otherwise executing the next loop.

[0023] By using the V2G discharge power prediction module based on reinforcement learning to predict the discharge power, it is possible to continuously learn through the interaction between the intelligent agent and the environment. Combining the charging station data and user behavior data, the operating law of the V2G discharge power can be obtained. This can make a more accurate prediction of the load reduction of the charging station, thereby improving the effectiveness of the load response regulation of the distribution network.

[0024] Obtain the load reduction forecast value of the distribution network according to the final charging station load forecast value and the V2G discharge power forecast value; Furthermore, the load reduction prediction value of the distribution network is obtained by: first, obtaining the minimum value of the final charging station load prediction value and the V2G discharge power prediction value to obtain the V2G offset; then, weighting the V2G offset using a preset transmission loss rate to obtain the load reduction prediction value of the distribution network; Furthermore, the calculation formula for the predicted value of the load reduction of the distribution network is: ; in, is the load reduction forecast value of the distribution network, indicating Load carrying capacity at a point in time; is the minimum function; is the final charging station load prediction value, indicating The total electricity demand expected at the charging station at a point in time; is the predicted value of V2G discharge power, indicating The total discharge capacity of the charging station at a point in time; is the transmission loss rate, which is set to 5%.

[0025] In order to illustrate the load reduction prediction value of the distribution network proposed in the present invention, the collected data at three different time points are randomly selected for the same charging station for testing. The collected data are the charging station data and user behavior data used for charging station load prediction and V2G discharge power prediction. Each group of data is recorded as data one, data two and data three respectively; the pre-trained charging station load prediction module and V2G discharge power prediction module are used to process each group of data to obtain the final charging station load prediction value and V2G discharge power prediction value of each group, and combined with the calculation formula of the load reduction prediction value of the distribution network, the load reduction prediction value test result of the distribution network is obtained, as shown in Table 1.

[0026] Table 1. Test results of load reduction prediction value of distribution network

[0027] The final charging station load forecast value and load reduction forecast value are input into the distribution network load response control module, and the charging station load control is optimized by combining the distribution network load data, the charging station load real-time data and the charging station load real-time reduction.

[0028] Furthermore, the process of optimizing the load control of the charging station by using the distribution network load response control module includes: Compare the final charging station load forecast value with the V2G discharge power forecast value to obtain the load reduction forecast value of the distribution network; The final charging station load prediction value and load reduction prediction value are compared with the charging station load real-time data and the charging station load real-time reduction amount, respectively, to obtain the load prediction error and the reduction amount prediction error; The objective function of the distribution network load response control module is constructed using the distribution network load data, load forecast error and reduction forecast error; the charging station load constraint, V2G constraint and distribution network constraint are used as the constraint conditions of the distribution network load response control module; Combine the distribution network load response control module and optimization algorithm to optimize the load control of charging stations; Furthermore, the distribution network load response control module is a multi-objective optimization model, and the objective function of the multi-objective optimization model can be expressed as: ; ; in, is the objective function; is the total length of the time series; is the load weight of the distribution network; Indicates time point Total load of the distribution network under represents the average load of the distribution network under the total length of the time series; is the charging station load prediction error weight; represents the charging station load prediction error; The prediction error weight for the reduction of charging stations; represents the prediction error of charging station reduction; is the total active power loss weight of the distribution network; is the total active power loss of the distribution network; Furthermore, the time point The total length of the time series An example is as follows: Assuming that the load response control of the distribution network within the next hour is optimized, with every 10 minutes as a time point, the total length of the time series is 6; Furthermore, the total active power loss of the distribution network is obtained by calling the real-time distribution network data through the distribution automation system and using the power flow calculation method; Furthermore, the charging station load prediction error is the difference between the real-time data of the charging station load and the final charging station load prediction value; the charging station reduction amount prediction error is the difference between the real-time reduction amount of the charging station load and the predicted value of the charging station load reduction amount; Furthermore, the distribution network load weight, charging station load prediction error weight, charging station reduction prediction error weight and distribution network total active power loss weight are all set to 0.25; the weight setting can be flexibly adjusted according to actual needs and is not unique.

[0029] By constructing a multi-objective function of the distribution network load response control module, and incorporating the load prediction error and the reduction amount prediction error into the distribution network load response control module to cope with the complex distribution network operating environment, the robustness of the control can be improved; and the control strategy can be adaptively adjusted according to real-time data and changing environment, thereby improving the effectiveness of distribution network load response control.

[0030] Embodiment 2 The present invention also proposes a data-driven distribution network load response control system, the structure of which can be referred to Figure 3 , specifically including: distribution network control data acquisition unit, charging station load prediction unit, V2G discharge power prediction unit, charging station load reduction prediction unit and distribution network load response control unit; The distribution network control data acquisition unit is used to obtain distribution network load data, charging station data and user behavior data; The charging station load prediction unit is used to input the charging station data and user behavior data into the charging station load prediction module to obtain the charging station load prediction value; the charging station load prediction value is corrected by using the electricity price and environmental assessment information to obtain the final charging station load prediction value; Furthermore, the specific process of the charging station load prediction unit obtaining the final charging station load prediction value includes: Construct a charging station load prediction module, and train it using historical charging station data, historical user behavior data, and historical load data to obtain a pre-trained charging station load prediction module; Input the charging pile data in the charging station data and the charging behavior data and user travel mode data in the user behavior data into the pre-trained charging station load prediction module to obtain the charging station load prediction value; Input environmental data, electricity price data and user preference data into the load impact assessment module to obtain the load impact assessment weight of the charging station; The charging station load impact assessment weight is used to weight the charging station load forecast value to obtain the final charging station load forecast value; Furthermore, the structure of the charging station load prediction module includes: an input layer, a feature attention mechanism layer, a feature fusion layer, a recurrent neural network layer and a fully connected output layer.

[0031] In order to verify the effectiveness of the charging station load forecasting scheme proposed in the present invention, the historical data of a charging station in the past three years was used to conduct a comparative test on the charging station load forecasting schemes; the charging station load forecasting schemes are: the scheme proposed in the present invention, which combines load forecasting and load impact assessment, recorded as Scheme 1; the feature attention mechanism layer in the charging station load forecasting module in the present invention is changed to a single feature attention mechanism layer, that is, only the time feature attention is selected for processing, so that the feature fusion layer loses its function and is removed, recorded as Scheme 2; the load impact assessment operation of the present invention is removed, and the output of the charging station load forecasting module is directly used as the final charging station load forecast value, recorded as Scheme 3; the proportion of the final charging station load forecast value within a reasonable range is obtained through manual proofreading, and the effectiveness test results of the charging station load forecasting scheme are shown in Table 2.

[0032] Table 2. Effectiveness test results of charging station load prediction scheme

[0033] It can be seen from Table 2 that the prediction results obtained by using the charging station load prediction scheme proposed in this application are more reasonable and effective than those obtained by using other schemes; among them, the charging station load prediction module combines time attention, space attention and user attention, and uses the weights output by the load impact evaluation module for correction, which can effectively improve the prediction accuracy of the charging station load, thereby improving the effectiveness of the load response regulation of the distribution network.

[0034] The V2G discharge power prediction unit is used to predict the V2G discharge power prediction value according to the charging station data and the user behavior data using the V2G discharge power prediction module based on reinforcement learning; Furthermore, the process of the V2G discharge power prediction unit obtaining the V2G discharge power prediction value includes: Construct a V2G discharge power prediction module based on reinforcement learning, and define the state space, action space and reward function of the agent according to the charging station data and user behavior data; The V2G discharge power prediction module is trained using a reinforcement learning algorithm and historical data, and the model parameters are updated using a gradient descent algorithm to obtain a pre-trained model. The real-time data of charging stations and real-time user behavior data are input into the pre-training model to obtain the predicted value of V2G discharge power.

[0035] The charging station load reduction amount prediction unit is used to obtain the load reduction amount prediction value of the distribution network according to the final charging station load prediction value and the V2G discharge power prediction value; The distribution network load response control unit is used to input the final charging station load forecast value and the load reduction forecast value into the distribution network load response control module, and optimize the charging station load control in combination with the distribution network load data, the charging station load real-time data and the charging station load real-time reduction.

[0036] Furthermore, the process of optimizing the load control of the charging station by the distribution network load response control unit using the distribution network load response control module includes: Compare the final charging station load forecast value with the V2G discharge power forecast value to obtain the load reduction forecast value of the distribution network; The final charging station load prediction value and load reduction prediction value are compared with the charging station load real-time data and the charging station load real-time reduction amount, respectively, to obtain the load prediction error and the reduction amount prediction error; The objective function of the distribution network load response control module is constructed using the distribution network load data, load forecast error and reduction forecast error; the charging station load constraint, V2G constraint and distribution network constraint are used as the constraint conditions of the distribution network load response control module; The load control of charging stations is optimized by combining the distribution network load response control module and optimization algorithm.

[0037] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data-driven distribution network load response control method, characterized in that: include: Obtain distribution network load data, charging station data, and user behavior data; Inputting the charging station data and the user behavior data into a charging station load prediction module to obtain a charging station load prediction value; The charging station load forecast value is corrected using the electricity price and environmental assessment information to obtain a final charging station load forecast value; According to the charging station data and the user behavior data, a V2G discharge power prediction module based on reinforcement learning is used to perform prediction to obtain a V2G discharge power prediction value; Obtaining a load reduction prediction value of the distribution network according to the final charging station load prediction value and the V2G discharge power prediction value; The final charging station load forecast value and the load reduction forecast value are input into the distribution network load response control module, and the charging station load control is optimized in combination with the distribution network load data, the charging station load real-time data and the charging station load real-time reduction amount.

2. The method for controlling load response of a distribution network based on data drive according to claim 1, characterized in that: The charging station data includes: charging station basic data, charging pile data, environmental data and electricity price data; the charging station basic data includes: charging station ID, geographical location, number of charging piles and type of charging piles; the charging pile data includes: charging pile status data, charging pile electrical parameter data, charging pile charging and discharging data and battery SOC data; the charging and discharging data includes: charging and discharging time, charging and discharging power and charging and discharging power; the user behavior data includes: user preference data, charging behavior data, user travel mode data and user discharge willingness data.

3. The method for controlling load response of a distribution network based on data drive according to claim 1, characterized in that: The process of obtaining the final charging station load prediction value by using the charging station load prediction module includes: Constructing the charging station load prediction module, training it using historical charging station data, historical user behavior data, and historical load data to obtain a pre-trained charging station load prediction module; Inputting the charging pile data in the charging station data and the charging behavior data and the user travel mode data in the user behavior data into the pre-trained charging station load prediction module to obtain the charging station load prediction value; Input environmental data, electricity price data and user preference data into the load impact assessment module to obtain the load impact assessment weight of the charging station; The charging station load prediction value is weighted by using the charging station load impact assessment weight to obtain the final charging station load prediction value.

4. The method for controlling load response of a distribution network based on data drive according to claim 1, characterized in that: The process of using a V2G discharge power prediction module based on reinforcement learning to perform prediction according to the charging station data and the user behavior data to obtain a V2G discharge power prediction value includes: Constructing the V2G discharge power prediction module based on reinforcement learning, and defining the state space, action space and reward function of the intelligent agent according to the charging station data and the user behavior data; The V2G discharge power prediction module is trained using a reinforcement learning algorithm and historical data, and the model parameters are updated using a gradient descent algorithm to obtain a pre-trained model; The real-time data of the charging station and the real-time user behavior data are input into the pre-training model to obtain the V2G discharge power prediction value.

5. The method for controlling load response of a distribution network based on data drive according to claim 1, characterized in that: The process of optimizing the load control of the charging station by using the distribution network load response control module includes: Compare the final charging station load prediction value with the V2G discharge power prediction value to obtain a load reduction prediction value of the distribution network; Compare the final charging station load prediction value and the load reduction amount prediction value with the charging station load real-time data and the charging station load real-time reduction amount, respectively, to obtain a load prediction error and a reduction amount prediction error; The objective function of the distribution network load response control module is constructed by using the distribution network load data, the load prediction error and the reduction amount prediction error; the charging station load constraint, the V2G constraint and the distribution network constraint are used as the constraint conditions of the distribution network load response control module; The load control of the charging station is optimized by combining the distribution network load response control module and the optimization algorithm.

6. A data-driven distribution network load response control system, characterized in that: include: Distribution network control data acquisition unit, charging station load prediction unit, V2G discharge power prediction unit, charging station load reduction prediction unit and distribution network load response control unit; The distribution network control data acquisition unit is used to obtain distribution network load data, charging station data and user behavior data; The charging station load prediction unit is used to input the charging station data and the user behavior data into the charging station load prediction module to obtain a charging station load prediction value; and to modify the charging station load prediction value using electricity price and environmental assessment information to obtain a final charging station load prediction value; The V2G discharge power prediction unit is used to perform prediction based on the charging station data and the user behavior data using a V2G discharge power prediction module based on reinforcement learning to obtain a V2G discharge power prediction value; The charging station load reduction amount prediction unit is used to obtain a load reduction amount prediction value of the distribution network according to the final charging station load prediction value and the V2G discharge power prediction value; The distribution network load response control unit is used to input the final charging station load prediction value and the load reduction amount prediction value into the distribution network load response control module, and optimize the charging station load control in combination with the distribution network load data, the charging station load real-time data and the charging station load real-time reduction amount.

Images