Virtual power plant load prediction and dynamic response method and system integrated with vehicle network interaction

Abstract

The invention relates to a virtual power plant load prediction and dynamic response method and system integrated with vehicle network interaction. The method comprises the following steps: acquiring and processing historical charging data to generate a user data set; extracting and analyzing the user data set to generate a charging behavior prediction result; collecting traffic flow data in real time and matching to generate a traffic influence coefficient; constructing a regional charging demand prediction model, and generating regional charging demand distribution; acquiring external environment event data and processing the data to generate a sudden load correction value; a comprehensive load prediction value is generated by fusing multi-level factors, the power distribution and the charging priority of the charging pile are adjusted, and a scheduling instruction is output to complete closed-loop operation, so that the problem of load prediction deviation caused by user behavior randomness, traffic dynamic coupling and unpredictability of emergencies is solved; accurate modeling and collaborative optimization of multi-level dynamic factors are realized, and the scheduling efficiency of the virtual power plant to the vehicle network interaction resources is improved.

Description

Technical Field

[0001] The present invention relates to the field of virtual power plant load forecasting and response, and particularly to a virtual power plant load forecasting and dynamic response method and system integrating vehicle-grid interaction. Background Art

[0002] With the large-scale popularization of electric vehicles and the rapid development of vehicle-to-grid (V2G) technology, electric vehicles in the power grid operation have gradually changed from a single load to a dispatchable distributed resource. Under this background, as the core technology carrier for aggregating distributed energy, the virtual power plant (VPP) urgently needs to solve the problems of dynamic prediction and real-time response of electric vehicle loads. The virtual power plant load forecasting and dynamic response method and system integrating vehicle-grid interaction aims to achieve the collaborative optimization of high-precision load forecasting and power grid dispatching by integrating multi-source information such as the charging and discharging behaviors of electric vehicles, dynamic traffic changes, and external environmental events, so as to improve the power grid's ability to absorb high-proportion renewable energy and operation stability.

[0003] However, in the prior art, there are still significant defects in the virtual power plant load forecasting method when dealing with vehicle-grid interaction scenarios. At the micro level, the randomness and individual differences of user charging behaviors (such as charging time deviation and changes in location preferences) make it difficult for traditional statistical models to accurately capture dynamic demands. Especially when user habits are affected by factors such as electricity price fluctuations and charging facility availability, the prediction deviation is further aggravated. At the meso level, there is a strong coupling relationship between the dynamic changes in traffic flow (such as congestion, construction, and holiday travel peaks) and the spatio-temporal distribution of charging demands. However, existing models often rely on static traffic data and cannot reflect the impact of sudden changes in traffic conditions on charging demands in real time. At the macro level, the unpredictability of emergencies such as extreme weather and large-scale events and their significant impact on charging loads (such as a sharp increase in air-conditioning loads caused by high-temperature weather and a sudden increase in local charging demands after a concert) lack a rapid identification and dynamic correction mechanism, resulting in the prediction model lagging behind the actual scenario. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide a virtual power plant load forecasting and dynamic response method and system integrating vehicle-grid interaction that can process multi-level dynamic factors.

[0005] The purpose of the present invention is achieved by the following solutions:

[0006] In the first aspect, the present application provides a virtual power plant load forecasting and dynamic response method integrating vehicle-grid interaction, including the following steps:

[0007] Obtain the charging history data of electric vehicle users, clean and standardize the charging history data to generate a user dataset;

[0008] Extract features and perform behavior analysis on the user dataset to generate charging behavior prediction results;

[0009] Collect traffic flow data in real time, spatially match the traffic flow data with the electric vehicle density distribution to generate a traffic impact coefficient;

[0010] Based on the traffic impact coefficient, construct a regional charging demand prediction model, and process the traffic flow data based on the regional charging demand prediction model to generate a regional charging demand distribution;

[0011] Obtain external environmental event data and process the external environmental event data to generate a sudden load correction value;

[0012] Perform multi-source fusion processing on the charging behavior prediction results, regional charging demand distribution, and sudden load correction value to generate a comprehensive load prediction value;

[0013] Based on the comprehensive load prediction value, generate a virtual power plant load scheduling plan. The virtual power plant load scheduling plan is used to adjust the charging pile power distribution and electric vehicle charging priority, and output a scheduling instruction to complete the closed-loop operation.

[0014] In one embodiment, extracting features and performing behavior analysis on the user dataset to generate charging behavior prediction results includes:

[0015] Perform clustering analysis on the user dataset based on the clustering algorithm, divide user categories according to the charging time distribution and charging location preference to generate a user behavior classification model;

[0016] Based on the user behavior classification model, perform time series segmentation on the historical charging records of each category of users, and extract the charging period distribution characteristics and charging duration characteristics;

[0017] Perform time series prediction modeling on the charging period distribution characteristics and charging duration characteristics based on the long short-term memory network to generate charging behavior prediction results.

[0018] In one embodiment, collecting traffic flow data in real time, spatially matching the traffic flow data with the electric vehicle density distribution to generate a traffic impact coefficient includes:

[0019] Collect traffic flow data in real time, and based on the geographic coordinate mapping technology, associate and match the road congestion index in the traffic flow data with the geographical location of the charging station to generate spatial association data;

[0020] Calculate the charging demand attenuation ratio of the congested area and the charging demand transfer ratio of the detour route based on the spatial correlation data, and generate traffic impact weight parameters;

[0021] Perform multi-dimensional feature fusion on the traffic impact weight parameters based on the random forest algorithm to generate traffic impact coefficients.

[0022] In one embodiment, obtain external environmental event data and process the external environmental event data to generate a sudden load correction value, including:

[0023] Obtain external environmental event data, where the external environmental event data includes weather forecast data and real-time traffic monitoring data;

[0024] Based on the high-temperature warning signal in the weather forecast data, predict the air conditioner usage through the temperature change curve to generate an air conditioner load increment;

[0025] Based on the real-time traffic monitoring data, extract the electric vehicle density distribution characteristics to generate charging hot spot identifiers for the activity areas;

[0026] Perform a joint analysis on the air conditioner load increment and the charging hot spot identifiers for the activity areas based on the gradient boosting decision tree algorithm to generate a sudden load correction value.

[0027] In one embodiment, perform multi-source fusion processing on the charging behavior prediction results, regional charging demand distribution, and sudden load correction value to generate a comprehensive load prediction value, including:

[0028] Based on the dynamic weight allocation strategy, assign short-term prediction weights to the charging behavior prediction results and spatial distribution weights to the regional charging demand distribution;

[0029] Process the ratios of the short-term prediction weights and spatial distribution weights based on the type and influence range of the sudden load correction value to generate fusion weight parameters;

[0030] Perform weighted superposition on the charging behavior prediction results, regional charging demand distribution, and sudden load correction value based on the fusion weight parameters to generate a comprehensive load prediction value.

[0031] In a second aspect, the present application provides a virtual power plant load prediction and dynamic response system integrating vehicle-grid interaction, and the system is configured with the following modules:

[0032] A charging data processing module, configured to obtain the charging historical data of electric vehicle users, clean and standardize the charging historical data, and generate a user data set;

[0033] A charging behavior prediction module, configured to perform feature extraction and behavior analysis on the user data set to generate charging behavior prediction results;

[0034] A traffic space matching module, which is used to collect traffic flow data in real time, perform spatial matching on the traffic flow data and the electric vehicle density distribution, and generate a traffic impact coefficient;

[0035] A regional demand forecasting module, which is used to construct a regional charging demand forecasting model based on the traffic impact coefficient, and process the traffic flow data based on the regional charging demand forecasting model to generate a regional charging demand distribution;

[0036] An external data processing module, which is used to obtain external environmental event data and process the external environmental event data to generate a sudden load correction value;

[0037] A multi-source data fusion module, which is used to perform multi-source fusion processing on the charging behavior prediction result, the regional charging demand distribution and the sudden load correction value to generate a comprehensive load prediction value;

[0038] A scheduling strategy generation module, which is used to generate a virtual power plant load scheduling plan based on the comprehensive load prediction value. The virtual power plant load scheduling method is used to adjust the charging pile power distribution and the electric vehicle charging priority, and output a scheduling instruction to complete the closed-loop operation.

[0039] In one embodiment, the charging behavior prediction module is configured with the following units:

[0040] A user clustering analysis unit, which is used to analyze the user data set based on a clustering algorithm, divide user categories according to the charging time distribution and charging location preference, and generate a user behavior classification model;

[0041] A charging feature extraction unit, which is used to perform time series segmentation on the historical charging records of each type of user based on the user behavior classification model, and extract the charging period distribution feature and the charging duration feature;

[0042] A time series prediction model construction unit, which is used to perform time series prediction modeling on the charging period distribution feature and the charging duration feature based on a long short-term memory network to generate a charging behavior prediction result.

[0043] In one embodiment, the traffic space matching module is configured with the following units:

[0044] A traffic data spatial association unit, which is used to collect traffic flow data in real time, and perform association matching on the road congestion index in the traffic flow data and the geographical location of the charging station based on the geographical coordinate mapping technology to generate spatial association data;

[0045] A traffic impact weight calculation unit, which is used to calculate the charging demand attenuation ratio of the congested area and the charging demand transfer ratio of the detour path based on the spatial association data, and generate a traffic impact weight parameter;

[0046] A multi-dimensional feature fusion unit is used to perform multi-dimensional feature fusion on traffic impact weight parameters based on a random forest algorithm to generate a traffic impact coefficient.

[0047] In one embodiment, the external data processing module is configured with the following units:

[0048] An external data acquisition and integration unit is used to acquire external environmental event data, and the external environmental event data includes weather forecast data and real-time traffic monitoring data;

[0049] An air-conditioning load prediction unit is used to predict the air-conditioning usage based on the high-temperature warning signal of the weather forecast data and generate an air-conditioning load increment through a temperature change curve;

[0050] A charging hot spot identification unit is used to extract the electric vehicle density distribution characteristics based on the real-time traffic monitoring data and generate an activity area charging hot spot identifier;

[0051] A joint analysis and correction unit is used to perform joint analysis on the air-conditioning load increment and the activity area charging hot spot identifier based on a gradient boosting decision tree algorithm to generate a sudden load correction value.

[0052] In one embodiment, the multi-source data fusion module is configured with the following units:

[0053] A dynamic weight allocation unit is used to assign a short-term prediction weight to the charging behavior prediction result and a spatial distribution weight to the regional charging demand distribution based on a dynamic weight allocation strategy;

[0054] A weight ratio adjustment unit is used to process the ratio of the short-term prediction weight and the spatial distribution weight based on the type and influence range of the sudden load correction value to generate a fusion weight parameter;

[0055] A weighted superposition fusion unit is used to perform weighted superposition on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value based on the fusion weight parameter to generate a comprehensive load prediction value.

[0056] In summary, a virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application realizes accurate prediction and comprehensive analysis of electric vehicle charging loads by cleaning and standardizing the charging historical data of electric vehicle users, extracting features and analyzing behavior patterns, and combining real-time traffic flow data, electric vehicle density distribution, and external environmental event data. Based on the comprehensive load prediction value generated by multi-source data fusion, it can provide a scientific and reasonable load scheduling plan for the virtual power plant, optimize the power distribution of charging piles and the charging priority of electric vehicles, effectively balance the grid load, reduce the operation cost, improve the reliability of the system and user satisfaction, and promote the coordinated development of electric vehicles and the energy system.

[0057] For better understanding and implementation, the present invention will be described in detail below with reference to the accompanying drawings. Description of the Drawings

[0058] Figure 1 It is a schematic flow chart of a virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by an embodiment of the present application;

[0059] Figure 2 It is a schematic flow chart of generating a comprehensive load prediction value in a virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by an embodiment of the present application;

[0060] Figure 3 It is a schematic structural diagram of a virtual power plant load prediction and dynamic response system integrating vehicle-grid interaction provided by another embodiment of the present application. Detailed Embodiment

[0061] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Preferred embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the understanding of the disclosure of the invention more thorough and comprehensive.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0063] In one embodiment, as Figure 1 shown, a virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction is provided. In this embodiment, taking the application of this method to a terminal as an example, it can be understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0064] S110. Obtain the charging historical data of electric vehicle users, clean and standardize the charging historical data, and generate a user data set.

[0065] Specifically, historical data such as the user's charging time, charging duration, charging power, charging location, etc. can be collected by connecting to devices such as the communication system of the electric vehicle charging pile and the in-vehicle terminal. After collecting the charging historical data, the collected charging historical data is cleaned to remove duplicate, incorrect, incomplete, and abnormal data records. Preferably, data cleaning techniques such as missing value filling, outlier detection and processing, etc. can be used. Then, the cleaned data is standardized to have a unified format and dimension. For example, the charging time is converted into a unified time format, and numerical data such as charging power and charge amount are normalized to be within the range of [0, 1] for subsequent model training and prediction. Finally, the cleaned and standardized charging historical data is organized into a structured data set, with each user corresponding to a record, containing multiple features of their charging behavior, such as charging frequency, average charging duration, charging location distribution, etc., providing a basis for subsequent feature extraction and behavior analysis.

[0066] S120. Extract features and analyze the behavior of the user data set to generate a charging behavior prediction result.

[0067] Specifically, features related to charging behavior are extracted from the user data set, such as the distribution characteristics of charging time (morning rush hour, evening rush hour, etc.), charging frequency, charging interval time, the change trend of charging power, etc. Feature engineering methods such as clustering analysis and principal component analysis can be used to reduce the dimension and select features of the original data, and extract the most representative and distinguishable features; and based on the extracted features, the charging behavior patterns of users are analyzed to identify the charging habits and rules of different types of users. For example, users are divided into categories such as regular charging type and random charging type, and the behavioral differences of different types of users in terms of charging time, charging location, etc. are analyzed. Preferably, methods such as time series analysis and Markov chain can be used to model and analyze the time series data of users' charging behavior.

[0068] Preferably, machine learning or deep learning algorithms such as support vector machine, long short-term memory network (LSTM), etc. can be used to predict the charging demand and behavior pattern of users in a specific future time period according to the historical charging behavior data and extracted features of users. For example, predict the charging probability, charging time window, charging power, etc. of users on a certain day to generate a charging behavior prediction result.

[0069] S130. Real-time collect traffic flow data, spatially match the traffic flow data with the electric vehicle density distribution, and generate a traffic impact coefficient.

[0070] Specifically, the system can obtain real-time traffic flow data in the road network through devices such as traffic sensors, cameras, and floating cars, including information such as traffic volume, vehicle speed, and vehicle type. Combining with the registration information of electric vehicles, the distribution data of charging piles, and historical charging data, etc., estimate the density distribution of electric vehicles in different regions. Preferably, Geographic Information System (GIS) technology can be used to combine the location information of electric vehicles with geospatial data to generate an electric vehicle density distribution map.

[0071] Associate the road segments or regions in the traffic flow data with the corresponding electric vehicle density values. For example, according to the geographical location of the road segments and the electric vehicle density distribution map, determine the electric vehicle density of each road segment, so as to analyze the spatial correlation relationship between traffic flow and electric vehicle density. Preferably, a mathematical model can be established, such as regression analysis, neural network, etc., to calculate the traffic impact coefficient according to the spatial matching results of traffic flow data and electric vehicle density distribution. This coefficient reflects the degree of influence of traffic conditions on the charging demand of electric vehicles. For example, in traffic congestion sections, the driving speed of electric vehicles decreases, which may lead to an increase in their energy consumption, thus affecting their charging demand.

[0072] S140. Based on the traffic impact coefficient, construct a regional charging demand prediction model, and process the traffic flow data based on the regional charging demand prediction model to generate a regional charging demand distribution.

[0073] Specifically, the traffic impact coefficient can be used as one of the input variables, combined with other relevant factors, such as regional population density, economic development level, charging pile distribution density, etc., to construct a regional charging demand prediction model. Preferably, machine learning algorithms, such as random forest, gradient boosting tree, etc., or model structures in deep learning such as Convolutional Neural Network (CNN) can be used to predict the regional charging demand.

[0074] Specifically, use historical data to train the constructed regional charging demand prediction model, optimize the performance of the model by adjusting the parameters and hyperparameters of the model, and improve the accuracy and stability of the prediction. Methods such as cross-validation and grid search can be used to determine the best combination of model parameters. After training is completed, input the real-time collected traffic flow data into the trained regional charging demand prediction model, combine with the traffic impact coefficient, process and transform the traffic flow data, and generate a regional charging demand distribution. For example, according to the size of the traffic flow and the traffic impact coefficient, predict the charging demand and its distribution of electric vehicles in different regions, providing a basis for subsequent load forecasting and scheduling.

[0075] S150. Obtain external environment event data, and process the external environment event data to generate a sudden load correction value.

[0076] Specifically, collect external environment event data related to electric vehicle charging, such as weather changes (rainfall, snowfall, high temperature, etc.), traffic accidents, road construction, large-scale events, etc. This data can be obtained through channels such as meteorological departments, traffic management departments, and news media, and usually exists in the form of event records. After the collection is completed, process the obtained external environment event data, including data cleaning, feature extraction, and coding conversion, etc. For example, normalize numerical data such as rainfall and temperature in weather data, and encode information such as the location and type of traffic accidents to convert them into input features recognizable by the model; and establish an association model between external environment events and electric vehicle charging loads, and calculate the sudden load correction value according to the processed external environment event data. For example, when a large-scale event occurs, the charging demand for electric vehicles around the event area may increase. By analyzing factors such as the scale and duration of the event through the model, a corresponding sudden load correction value is generated to adjust the comprehensive load prediction value.

[0077] S160. Perform multi-source fusion processing on the charging behavior prediction result, regional charging demand distribution, and sudden load correction value to generate a comprehensive load prediction value.

[0078] Specifically, fuse and process multiple data sources such as the charging behavior prediction result, regional charging demand distribution, and sudden load correction value, and comprehensively consider the impact of user individual charging behavior, regional overall charging demand, and external environment events on the load. Fusion algorithms such as weighted fusion and Bayesian network can be used to integrate the information of different data sources. Through multi-source fusion processing, the electric vehicle charging load prediction value considering various factors is obtained. This prediction value not only reflects the charging demand of individual users, but also considers the impact of regional traffic conditions and external environment events on the overall load, and can more accurately reflect the distribution of electric vehicle charging load in time and space, providing an accurate basis for the load dispatching of the virtual power plant.

[0079] S170. Generate a virtual power plant load dispatching plan based on the comprehensive load prediction value. The virtual power plant load dispatching plan is used to adjust the power distribution of charging piles and the charging priority of electric vehicles, and output dispatching instructions to complete the closed-loop operation.

[0080] Specifically, according to the comprehensive load prediction value, combined with the operation constraints and objective functions of the virtual power plant, such as grid load balance, cost minimization, maximization of user satisfaction, etc., a load scheduling scheme for the virtual power plant is formulated. Optimization algorithms, such as linear programming, integer programming, genetic algorithms, etc., can be used to solve the optimal scheduling strategy and determine the power distribution of each charging pile and the charging priority of electric vehicles in different time periods. And according to the load scheduling scheme, the power distribution of the charging piles is dynamically adjusted, and the charging sequence and charging power of electric vehicles are reasonably arranged. For example, during peak load periods, the power output of some charging piles is appropriately reduced to give priority to meeting the charging needs of high-priority users; during off-peak load periods, the power utilization rate of the charging piles is increased to encourage users to charge.

[0081] After the power distribution is completed, the system converts the formulated load scheduling scheme into specific scheduling instructions, sends them to each charging pile and related control devices, and executes the power distribution and charging priority adjustment operations. At the same time, the operation status of the system is monitored in real time, and the scheduling scheme is fed back and adjusted according to the actual operation situation to realize the closed-loop operation of the virtual power plant load scheduling and ensure the stable and efficient operation of the system.

[0082] In summary, the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application realizes the accurate prediction and comprehensive analysis of electric vehicle charging load by cleaning and standardizing the charging historical data of electric vehicle users, extracting features and analyzing behavior patterns, and combining real-time traffic flow data, electric vehicle density distribution, and external environment event data. Based on the comprehensive load prediction value generated by multi-source data fusion, it can provide a scientific and reasonable load scheduling scheme for the virtual power plant, optimize the power distribution of charging piles and the charging priority of electric vehicles, effectively balance the grid load, reduce the operation cost, improve the reliability of the system and user satisfaction, and promote the coordinated development of electric vehicles and the energy system.

[0083] In one embodiment, the steps for generating the charging behavior prediction result in the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application are as follows:

[0084] S121. Perform clustering analysis on the user dataset based on the clustering algorithm, divide user categories according to the charging time distribution and charging location preference, and generate a user behavior classification model.

[0085] Preferably, appropriate clustering algorithms such as K-means, DBSCAN, and hierarchical clustering can be selected to perform clustering analysis on the user dataset. The K-means algorithm is widely used in user behavior clustering due to its simplicity and efficiency, but the number of clusters needs to be specified in advance. The DBSCAN algorithm can automatically discover clusters of any shape and is robust to noise data, but it is more sensitive to parameter selection. Select the most suitable algorithm according to the data characteristics and clustering objectives.

[0086] Specifically, extract features related to charging time and location from the user dataset, such as the start time of charging, the end time of charging, the charging frequency, the longitude and latitude of the charging location, etc. Normalize these features to eliminate the differences in dimensions and magnitude levels, making the data suitable for processing by clustering algorithms; input the processed feature data into the clustering algorithm, and cluster users according to the charging time distribution and charging location preference. By analyzing the clustering results, determine the charging behavior patterns of different types of users, such as regular charging on weekdays, concentrated charging on weekends, and charging preference for specific areas, etc. Save the clustering results as a user behavior classification model to provide a basis for subsequent charging behavior prediction.

[0087] Preferably, indicators such as the silhouette coefficient and the Davies-Bouldin index can be used to evaluate the clustering results and judge the rationality and compactness of the clustering. If the evaluation results are not satisfactory, adjust the parameters of the clustering algorithm or select other clustering algorithms to re-perform clustering analysis to improve the clustering quality and the accuracy of the model.

[0088] S122. Based on the user behavior classification model, perform time series segmentation on the historical charging records of each type of user, and extract the charging period distribution characteristics and charging duration characteristics.

[0089] Preferably, appropriate time series segmentation methods such as the sliding window method and the segmentation method can be used to segment the historical charging records of each type of user. The sliding window method can capture the continuity and change trend of charging behavior in time, and by setting appropriate window sizes and step lengths, the time series can be segmented into multiple subsequences. The segmentation method divides the time series into different segments according to the significant change points of the charging behavior, and each segment represents a relatively stable charging stage.

[0090] Specifically, extract the charging period distribution features from the segmented time series, such as the probability distribution of the charging start time, the charging peak period, and the changes in the charging frequency in different time periods. Statistical methods such as probability density functions and cumulative distribution functions can be used to model and describe the charging period distribution, and analyze the possibility and pattern of users charging in different time periods. After extracting the features, calculate the duration of each charge, and extract the features related to the charging duration, such as the average charging duration, the standard deviation of the charging duration, and the distribution pattern of the charging duration. Integrate the extracted charging period distribution features and charging duration features to form a feature vector that can comprehensively describe the charging behavior characteristics of users. Adopt appropriate feature representation methods, such as one-hot encoding, embedding vectors, etc., to convert the feature vector into a form suitable for input to the subsequent prediction model.

[0091] S123. Perform a time series prediction modeling on the charging period distribution features and charging duration features based on the long short-term memory network to generate the charging behavior prediction results.

[0092] Specifically, LSTM is a special recurrent neural network (RNN) that can effectively handle the long-term and short-term dependencies in time series data. Its core is the memory unit, which determines the flow and preservation of information through the control of the input gate, forget gate, and output gate. The input gate controls the writing of new information, the forget gate controls the degree of retention of old information in the memory unit, and the output gate controls the output of information in the memory unit. This structure enables LSTM to automatically capture the complex patterns and trends in the time series during the learning process.

[0093] Specifically, the system takes the extracted charging period distribution characteristics and charging duration characteristics as inputs to construct an LSTM time series prediction model. Define the network structure of the model including the input layer, LSTM layer, fully connected layer, etc., and determine the number of neurons and activation functions in each layer. Use historical data to train the model, and adjust the parameters of the model through the backpropagation algorithm and optimizer (such as Adam) to minimize the error between the predicted value and the true value, so that the model can accurately learn the time series law of charging behavior. Preferably, metrics such as mean squared error (MSE) and mean absolute error (MAE) can be used to evaluate the trained LSTM model to judge the prediction performance of the model. Through methods such as cross-validation and hyperparameter tuning, optimize the structure and parameters of the model to improve the generalization ability and prediction accuracy of the model. For example, adjust hyperparameters such as the number of LSTM layers, the number of neurons, and the learning rate to find the optimal model configuration. Apply the evaluated and optimized LSTM model to actual charging behavior prediction, input the current charging period distribution characteristics and charging duration characteristics, and output the prediction results of charging behavior within a specific future time period, including the predicted charging start time, charging duration, charging frequency, etc. These prediction results can provide an accurate basis for the load scheduling of the virtual power plant and help it formulate a reasonable scheduling strategy in advance.

[0094] In summary, in the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application, the steps for generating the traffic impact coefficient are as follows: By performing clustering analysis on the electric vehicle user dataset, classify users according to their charging time and location preferences to generate a user behavior classification model. On this basis, perform time series segmentation on the historical charging records of each type of user, extract the charging period distribution characteristics and charging duration characteristics, and provide rich feature information for subsequent prediction modeling. Finally, use the long short-term memory network to perform time series prediction modeling on these features to generate accurate charging behavior prediction results. This series of processes helps the virtual power plant understand the charging demand patterns of users in advance, reasonably arrange the power distribution of charging piles and the charging priorities of electric vehicles, optimize the load scheduling plan, improve the operation efficiency and stability of the system, reduce the operation cost, enhance the user experience, and promote the coordinated development of electric vehicles and the energy system.

[0095] In the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application, the steps for generating the traffic impact coefficient are as follows:

[0096] S131. Collect traffic flow data in real time, and based on the geographic coordinate mapping technology, associate and match the road congestion index in the traffic flow data with the geographical location of the charging station to generate spatial association data.

[0097] Specifically, information such as traffic flow, vehicle speed, and vehicle type can be obtained in real time through devices such as sensors, cameras, and floating cars deployed on the road. These data are usually continuously updated in the form of time series, reflecting the dynamic changes in traffic conditions. Preferably, the geographic information system (GIS) technology can be used to match the road information in the traffic flow data with the geographical location information of the charging stations. Through coordinate transformation and spatial analysis, the spatial relationship between each road and the charging station is determined, and spatial association data is generated. For example, the congestion index of each road within a certain range around the charging station can be calculated and associated with the location of the charging station.

[0098] S132. Calculate the charging demand attenuation ratio of the congested area and the charging demand transfer ratio of the detour path based on the spatial association data, and generate traffic impact weight parameters.

[0099] Specifically, analyze the traffic conditions in the congested area in the spatial association data to determine the impact of congestion on the charging demand of electric vehicles. Generally speaking, congestion will lead to a decrease in the driving speed of electric vehicles and an increase in energy consumption, but at the same time, it may also cause some users to choose to delay charging or reduce the number of charging times. By establishing a mathematical model, such as a regression analysis model, according to the relationship between the congestion degree (such as the congestion index) and the change in charging demand, calculate the charging demand attenuation ratio of the congested area. When a road is congested, some vehicles will choose to take a detour on other paths. Analyze the traffic flow changes and charging station distribution on the detour path to determine the charging demand transfer ratio of the detour path. This is achieved by statistically analyzing the traffic flow data and charging station usage on the detour path and establishing a transfer ratio model. Integrate the charging demand attenuation ratio of the congested area and the charging demand transfer ratio of the detour path to generate traffic impact weight parameters. This parameter reflects the comprehensive impact degree of traffic congestion on the charging demand in different regions.

[0100] S133. Perform multi-dimensional feature fusion on the traffic impact weight parameters based on the random forest algorithm to generate traffic impact coefficients.

[0101] Specifically, the random forest is an ensemble learning algorithm composed of multiple decision trees. Through random sampling and feature selection, each decision tree is trained on a subset of the training data, and the final prediction result is determined by the voting or average result of all decision trees. The random forest has advantages such as handling high-dimensional data and preventing overfitting. Specifically, the traffic impact weight parameter is integrated with other relevant features (such as time, weather, regional population density, etc.) as the input features of the random forest algorithm. By training the random forest model, these multi-dimensional features are fused and learned to uncover the complex relationships and interactions between different features. Using the trained random forest model, new traffic impact weight parameters and relevant features are predicted to generate a traffic impact coefficient. This coefficient comprehensively considers the contributions of various factors to traffic impact and can more accurately reflect the degree of influence of traffic conditions on the charging demand of electric vehicles.

[0102] In summary, the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application realizes the association and matching of traffic flow data and the geographical location of charging stations through the collection and processing of real-time traffic flow data, combined with geographical coordinate mapping technology, and generates spatial association data. On this basis, the charging demand attenuation ratio in the congested area and the charging demand transfer ratio of the detour route are calculated to generate the traffic impact weight parameter. Finally, the random forest algorithm is used to perform multi-dimensional feature fusion on the traffic impact weight parameter to generate the traffic impact coefficient. This series of processes helps the virtual power plant to more accurately understand the impact of traffic conditions on the charging demand of electric vehicles, thereby optimizing the charging load prediction and scheduling scheme, improving the operation efficiency and stability of the system, reducing the operation cost, enhancing the user experience, and promoting the coordinated development of the traffic and energy systems.

[0103] In one embodiment, the steps for generating the sudden load correction value in the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application are as follows:

[0104] S151. Obtain external environment event data, where the external environment event data includes weather forecast data and real-time traffic monitoring data.

[0105] Specifically, weather forecast data for a future period, including information such as temperature, humidity, wind speed, and precipitation probability, can be obtained through weather data API interfaces such as OpenWeatherMap and Fengyun Weather. These data are usually returned in JSON or XML format and need to be parsed and processed to extract the temperature data related to the prediction of air conditioner usage. And the real-time traffic monitoring data interface provided by the traffic management department can be used to obtain information such as the traffic flow, vehicle speed, and congestion index of the road. These data can be collected by traffic sensors, cameras, and other devices and reflect the real-time state of the current traffic conditions.

[0106] S152. The high - temperature warning signal based on weather forecast data predicts the air - conditioner usage through the temperature change curve and generates the air - conditioner load increment.

[0107] Specifically, analyze the obtained weather forecast data to identify the high - temperature warning signal. That is, when the forecast temperature exceeds a certain threshold (such as 35°C), the high - temperature warning is triggered. The system constructs a temperature change curve based on historical temperature data and weather forecast data, and analyzes the change trend of temperature over time. Time - series analysis methods such as moving average and exponential smoothing can be used to smooth the temperature data to obtain a more accurate temperature change trend. Based on the temperature change curve, establish a relationship model between air - conditioner usage and temperature, such as a linear regression model or a polynomial regression model. Train the model with historical data to predict the air - conditioner usage under different temperature conditions, and then calculate the air - conditioner load increment. For example, when the temperature rises, the air - conditioner usage increases, and the air - conditioner load increment also increases.

[0108] S153. Based on real - time traffic monitoring data, extract the density distribution characteristics of electric vehicles and generate charging hot - spot identifiers for active areas.

[0109] Specifically, analyze the real - time traffic monitoring data, and combine the characteristics of electric vehicles (such as license - plate recognition, vehicle type, etc.) to extract the density distribution characteristics of electric vehicles on the road. Preferably, clustering algorithms such as DBSCAN and K - means can be used to cluster the distribution of electric vehicles in different regions to obtain high - density regions and low - density regions; mark the regions with a high density of electric vehicles as charging hot - spots, and at the same time combine the location information of charging stations to generate charging hot - spot identifiers for active areas, providing a basis for subsequent load correction.

[0110] S154. Based on the gradient - boosting decision - tree algorithm, jointly analyze the air - conditioner load increment and the charging hot - spot identifiers for active areas to generate a sudden - load correction value.

[0111] Specifically, the gradient - boosting decision tree (GBDT) is an ensemble learning algorithm based on the gradient - boosting framework. By iteratively training decision trees, each iteration fits the residuals of the previous - round model, and finally sums the prediction results of multiple decision trees with weights to obtain the final prediction result. This algorithm has the advantages of processing high - dimensional data and capturing non - linear relationships.

[0112] Specifically, the air-conditioning load increment and the charging hot spot identification in the active area are used as input features, combined with other relevant features (such as time, regional population density, etc.), and input into the GBDT model for joint analysis. By training the model, the complex relationships and interactions between different features are mined to achieve feature fusion and learning. Using the trained GBDT model, features such as the new air-conditioning load increment and the charging hot spot identification in the active area are predicted to generate a sudden load correction value. This correction value comprehensively considers the combined effects of weather and traffic factors on the load, can more accurately reflect the actual load changes, and provides a more accurate basis for the load dispatching of the virtual power plant.

[0113] In summary, the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application obtains and processes weather forecast data and real-time traffic monitoring data, predicts the air-conditioning load increment and extracts the electric vehicle density distribution characteristics respectively, and generates the charging hot spot identification in the active area. On this basis, the gradient boosting decision tree algorithm is used to conduct joint analysis on the two to generate a sudden load correction value. This series of processes helps the virtual power plant to more accurately predict and respond to sudden load changes, optimize the charging load dispatching plan, improve the operation efficiency and stability of the system, reduce the operation cost, enhance the user experience, and promote the coordinated development of the energy and transportation systems.

[0114] In one embodiment, as Figure 2 shown, the steps for generating the comprehensive load prediction value in the virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application are as follows:

[0115] S161. Based on the dynamic weight allocation strategy, assign a short-term prediction weight to the charging behavior prediction result and a spatial distribution weight to the regional charging demand distribution.

[0116] Preferably, methods such as the analytic hierarchy process (AHP) can be used to dynamically assign weights to the charging behavior prediction result and the regional charging demand distribution according to the uncertainty of the prediction result and the reliability of the data. For example, when the confidence level of the charging behavior prediction result is high, assign a larger short-term prediction weight to it, and vice versa, reduce the weight; for the regional charging demand distribution, dynamically adjust the spatial distribution weight according to its spatial coverage and data accuracy to reflect its relative importance in the comprehensive load prediction.

[0117] S162. Process the ratio of the short-term prediction weight and the spatial distribution weight based on the type and influence range of the sudden load correction value to generate a fusion weight parameter.

[0118] Specifically, according to the type of sudden load correction value (such as weather changes, traffic accidents, etc.) and the influence range (such as local area or the whole network), determine its influence degree on the short-term prediction weight and the spatial distribution weight. For example, if the sudden load correction value is caused by weather changes in a local area, it mainly affects the short-term prediction weight and has less influence on the spatial distribution weight; if it is caused by traffic congestion across the whole network, it may have a greater impact on both the short-term prediction weight and the spatial distribution weight. According to the type and influence range of the sudden load correction value, adjust the ratio of the short-term prediction weight and the spatial distribution weight to generate a fused weight parameter. For example, a linear combination or a non-linear mapping method can be used to incorporate the influence of the sudden load correction value into the calculation of the weight ratio, so that the fused weight parameter can more accurately reflect the key points of load prediction in the current situation.

[0119] S163. Based on the fused weight parameter, perform weighted superposition on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to generate a comprehensive load prediction value.

[0120] Specifically, the system performs weighted summation on the pre-set weight parameters of the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to obtain a comprehensive load prediction value that fuses the three, and verifies the generated comprehensive load prediction value, compares it with the actual load data for analysis, and evaluates the prediction accuracy. If there is a deviation in the prediction result, further optimize the weight allocation strategy and the weighted superposition method to improve the accuracy and reliability of the comprehensive load prediction.

[0121] In summary, a virtual power plant load prediction and dynamic response method integrating vehicle-grid interaction provided by this application uses a dynamic weight allocation strategy to flexibly assign weights to the charging behavior prediction result and the regional charging demand distribution, and adjusts the weight ratio according to the characteristics of the sudden load correction value to generate a fused weight parameter. Finally, based on the fused weight parameter, perform weighted superposition on each prediction result and correction value to generate a comprehensive load prediction value. This series of processes helps the virtual power plant to more accurately grasp the dynamic changes of the electric vehicle charging load, optimize the load scheduling plan, improve the operation efficiency and stability of the system, reduce the operation cost, enhance the user experience, and promote the coordinated development of the energy and transportation systems.

[0122] Preferably, as Figure 3 shown, this application also provides a virtual power plant load prediction and dynamic response system 200 integrating vehicle-grid interaction, and this system is configured with the following modules:

[0123] A charging data processing module 210, which is used to obtain the charging historical data of electric vehicle users, clean and standardize the charging historical data, and generate a user data set;

[0124] The charging behavior prediction module 220 is used to extract features and analyze behaviors from the user dataset, and generate a charging behavior prediction result;

[0125] The traffic space matching module 230 is used to collect traffic flow data in real time, spatially match the traffic flow data with the electric vehicle density distribution, and generate a traffic impact coefficient;

[0126] The regional demand prediction module 240 is used to build a regional charging demand prediction model based on the traffic impact coefficient, and process the traffic flow data based on the regional charging demand prediction model to generate a regional charging demand distribution;

[0127] The external data processing module 250 is used to obtain external environment event data and process the external environment event data to generate a sudden load correction value;

[0128] The multi-source data fusion module 260 is used to perform multi-source fusion processing on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to generate a comprehensive load prediction value;

[0129] The scheduling strategy generation module 270 is used to generate a virtual power plant load scheduling plan based on the comprehensive load prediction value. The virtual power plant load scheduling method is used to adjust the charging pile power distribution and the electric vehicle charging priority, and output a scheduling instruction to complete the closed-loop operation.

[0130] In summary, the virtual power plant load prediction and dynamic response system integrating vehicle-grid interaction provided by this application realizes the accurate prediction and comprehensive analysis of the electric vehicle charging load by cleaning and standardizing the charging historical data of electric vehicle users, extracting features and analyzing behavior patterns, and combining real-time traffic flow data, electric vehicle density distribution, and external environment event data. Based on the comprehensive load prediction value generated by multi-source data fusion, it can provide a scientific and reasonable load scheduling plan for the virtual power plant, optimize the charging pile power distribution and the electric vehicle charging priority, effectively balance the grid load, reduce the operation cost, improve the reliability of the system and the user satisfaction, and promote the coordinated development of electric vehicles and the energy system.

[0131] Preferably, in one embodiment, the charging behavior prediction module 220 is configured with the following units:

[0132] The user clustering analysis unit 221 is used to analyze the user dataset based on the clustering algorithm, divide user categories according to the charging time distribution and charging location preference, and generate a user behavior classification model;

[0133] The charging feature extraction unit 222 is used to perform time series segmentation on the historical charging records of each type of user based on the user behavior classification model, and extract the charging period distribution feature and the charging duration feature;

[0134] The time series prediction model construction unit 223 is used to perform time series prediction modeling on the charging period distribution characteristics and charging duration characteristics based on the long short-term memory network, and generate a charging behavior prediction result.

[0135] Preferably, in one embodiment, the traffic space matching module 230 is configured with the following units:

[0136] The traffic data space association unit 231 is used to collect traffic flow data in real time, and perform association matching on the road congestion index in the traffic flow data and the geographical location of the charging station based on the geographical coordinate mapping technology, and generate space association data;

[0137] The traffic impact weight calculation unit 232 is used to calculate the charging demand attenuation ratio in the congested area and the charging demand transfer ratio of the detour path based on the space association data, and generate a traffic impact weight parameter;

[0138] The multi-dimensional feature fusion unit 233 is used to perform multi-dimensional feature fusion on the traffic impact weight parameter based on the random forest algorithm, and generate a traffic impact coefficient.

[0139] Preferably, in one embodiment, the external data processing module 250 is configured with the following units:

[0140] The external data acquisition and integration unit 251 is used to acquire external environment event data, and the external environment event data includes weather forecast data and real-time traffic monitoring data;

[0141] The air conditioning load prediction unit 252 is used to predict the air conditioning usage amount through the temperature change curve based on the high temperature warning signal in the weather forecast data, and generate an air conditioning load increment;

[0142] The charging hot spot identification unit 253 is used to extract the electric vehicle density distribution characteristics based on the real-time traffic monitoring data, and generate an active area charging hot spot identifier;

[0143] The joint analysis and correction unit 254 is used to perform joint analysis on the air conditioning load increment and the active area charging hot spot identifier based on the gradient boosting decision tree algorithm, and generate a sudden load correction value.

[0144] Preferably, in one embodiment, the multi-source data fusion module 260 is configured with the following units:

[0145] The dynamic weight allocation unit 261 is used to assign a short-term prediction weight to the charging behavior prediction result and a spatial distribution weight to the regional charging demand distribution based on the dynamic weight allocation strategy;

[0146] The weight ratio adjustment unit 262 is configured to process the ratios of the short-term prediction weight and the spatial distribution weight based on the type and influence range of the burst load correction value, and generate a fused weight parameter;

[0147] The weighted superposition fusion unit 263 is configured to perform weighted superposition on the charging behavior prediction result, the regional charging demand distribution, and the burst load correction value based on the fused weight parameter, and generate a comprehensive load prediction value.

[0148] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0149] In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present application, "a plurality" means two or more, unless otherwise specifically defined.

[0150] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed in the present application, and these should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A virtual power plant load forecasting and dynamic response method integrating vehicle-grid interaction, characterized in that It includes the following steps: Obtain the charging history data of electric vehicle users, clean and standardize the charging history data to generate a user dataset; Perform feature extraction and behavior analysis on the user dataset to generate a charging behavior prediction result; Collect traffic flow data in real time, spatially match the traffic flow data with the electric vehicle density distribution to generate a traffic impact coefficient; Based on the traffic impact coefficient, construct a regional charging demand prediction model, and process the traffic flow data based on the regional charging demand prediction model to generate a regional charging demand distribution; Obtain external environment event data and process the external environment event data to generate a sudden load correction value; Perform multi-source fusion processing on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to generate a comprehensive load prediction value; Based on the comprehensive load prediction value, generate a virtual power plant load scheduling plan, where the virtual power plant load scheduling plan is used to adjust the charging pile power distribution and the charging priority of electric vehicles, and output a scheduling instruction to complete the closed-loop operation.

2. The method according to claim 1, characterized in that, The performing feature extraction and behavior analysis on the user dataset to generate a charging behavior prediction result includes: Perform clustering analysis on the user dataset based on the clustering algorithm, divide user categories according to the charging time distribution and charging location preference to generate a user behavior classification model; Based on the user behavior classification model, perform time series segmentation on the historical charging records of each type of user, and extract the charging period distribution feature and the charging duration feature; Perform time series prediction modeling on the charging period distribution feature and the charging duration feature based on the long short-term memory network to generate the charging behavior prediction result.

3. The method according to claim 1, characterized in that The collecting traffic flow data in real time, spatially matching the traffic flow data with the electric vehicle density distribution to generate a traffic impact coefficient includes: Collect traffic flow data in real time, and based on the geographic coordinate mapping technology, associate and match the road congestion index in the traffic flow data with the geographical location of the charging station to generate spatial association data; Calculate the charging demand attenuation ratio in the congested area and the charging demand transfer ratio of the detour path based on the spatial association data to generate a traffic impact weight parameter; Perform multi-dimensional feature fusion on the traffic impact weight parameter based on the random forest algorithm to generate the traffic impact coefficient.

4. The method according to claim 1, characterized in that, The obtaining external environment event data and processing the external environment event data to generate a sudden load correction value includes: Obtain external environment event data, where the external environment event data includes weather forecast data and real-time traffic monitoring data; Based on the high temperature warning signal in the weather forecast data, predict the air conditioner usage through the temperature change curve to generate an air conditioner load increment; Based on the real-time traffic monitoring data, extract the electric vehicle density distribution feature to generate an activity area charging hot spot identifier; Perform joint analysis on the air conditioner load increment and the activity area charging hot spot identifier based on the gradient boosting decision tree algorithm to generate the sudden load correction value.

5. The method according to claim 1, characterized in that, Performing multi-source fusion processing on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to generate a comprehensive load prediction value includes: Based on a dynamic weight allocation strategy, assigning a short-term prediction weight to the charging behavior prediction result and a spatial distribution weight to the regional charging demand distribution; Based on the type and influence range of the sudden load correction value, processing the ratio of the short-term prediction weight and the spatial distribution weight to generate a fusion weight parameter; Based on the fusion weight parameter, performing weighted superposition on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to generate the comprehensive load prediction value.

6. A virtual power plant load forecasting and dynamic response system integrating vehicle-grid interaction, characterized in that, Including: A charging data processing module for obtaining the charging historical data of electric vehicle users, cleaning and standardizing the charging historical data to generate a user dataset; A charging behavior prediction module for performing feature extraction and behavior analysis on the user dataset to generate a charging behavior prediction result; A traffic space matching module for collecting traffic flow data in real time, spatially matching the traffic flow data with the electric vehicle density distribution to generate a traffic impact coefficient; A regional demand prediction module for constructing a regional charging demand prediction model based on the traffic impact coefficient and processing the traffic flow data based on the regional charging demand prediction model to generate a regional charging demand distribution; An external data processing module for obtaining external environment event data and processing the external environment event data to generate a sudden load correction value; A multi-source data fusion module for performing multi-source fusion processing on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value to generate a comprehensive load prediction value; A scheduling strategy generation module for generating a virtual power plant load scheduling plan based on the comprehensive load prediction value. The virtual power plant load scheduling method is used to adjust the charging pile power distribution and the charging priority of electric vehicles, and output a scheduling instruction to complete closed-loop operation.

7. The system according to claim 6, wherein The charging behavior prediction module is configured with the following units: A user clustering analysis unit for analyzing the user dataset based on a clustering algorithm, dividing user categories according to the charging time distribution and charging location preference, and generating a user behavior classification model; A charging feature extraction unit for performing time series segmentation on the historical charging records of each type of user based on the user behavior classification model, and extracting the charging period distribution feature and the charging duration feature; A time series prediction model construction unit for performing time series prediction modeling on the charging period distribution feature and the charging duration feature based on a long short-term memory network to generate the charging behavior prediction result.

8. The system according to claim 6, wherein The traffic space matching module is configured with the following units: A traffic data spatial association unit for collecting traffic flow data in real time, and associating and matching the road congestion index in the traffic flow data with the geographical location of the charging station based on geographical coordinate mapping technology to generate spatial association data; A traffic impact weight calculation unit, configured to calculate a charging demand attenuation ratio of a congested area and a charging demand transfer ratio of a detour path based on the spatial association data, and generate a traffic impact weight parameter; A multi-dimensional feature fusion unit, configured to perform multi-dimensional feature fusion on the traffic impact weight parameter based on a random forest algorithm to generate the traffic impact coefficient.

9. The system according to claim 6, wherein The external data processing module is configured with the following units: An external data acquisition and integration unit, configured to acquire external environmental event data, where the external environmental event data includes weather forecast data and real-time traffic monitoring data; An air-conditioning load prediction unit, configured to predict the air-conditioning usage amount through a temperature change curve based on a high-temperature warning signal in the weather forecast data, and generate an air-conditioning load increment; A charging hot spot identification unit, configured to extract an electric vehicle density distribution feature based on the real-time traffic monitoring data, and generate an activity area charging hot spot identifier; A joint analysis and correction unit, configured to perform joint analysis on the air-conditioning load increment and the activity area charging hot spot identifier based on a gradient boosting decision tree algorithm to generate the sudden load correction value.

10. The system according to claim 6, characterized in that, The multi-source data fusion module is configured with the following units: A dynamic weight allocation unit, configured to assign a short-term prediction weight to the charging behavior prediction result and a spatial distribution weight to the regional charging demand distribution based on a dynamic weight allocation strategy; A weight ratio adjustment unit, configured to process the ratio of the short-term prediction weight and the spatial distribution weight based on the type and influence range of the sudden load correction value to generate a fusion weight parameter; A weighted superposition fusion unit, configured to perform weighted superposition on the charging behavior prediction result, the regional charging demand distribution, and the sudden load correction value based on the fusion weight parameter to generate the comprehensive load prediction value.

Images