A method and system for predicting electric vehicle charging load based on data correlation
Through data correlation analysis and LSTM model, key features are identified and retained, and the problem of neglecting multidimensional data correlation and time series characteristics in the prior art is solved, and high-precision electric vehicle charging load prediction is achieved.
Patent Information
- Application Number
- CN202411266961.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-11
AI Technical Summary
The existing electric vehicle charging load prediction methods ignore the complex correlation between multidimensional data, lack effective feature selection and dimensionality reduction processing, and fail to fully utilize the characteristics of time series data, resulting in low prediction accuracy.
Using data correlation analysis method, key features are identified through Pearson correlation coefficient, mutual information method and time-delay correlation analysis, weighted principal component analysis and dimensionality reduction, and a charging load prediction model is established using the LSTM model to process the length of time series data. Short-term dependency.
It improves the accuracy and stability of charging load prediction, reduces redundant features, retains the main information in the data, and improves model training efficiency and prediction performance.
Smart Images
Figure CN118798440B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of charging load analysis, and in particular to a method and system for predicting charging load of an electric vehicle based on data correlation. Background Art
[0002] As the number of electric vehicles continues to increase, the prediction of electric vehicle charging load has become an important issue in power grid scheduling and planning. Existing charging load prediction methods mainly rely on statistical analysis of historical data and simple prediction models, but these methods have the following shortcomings:
[0003] Ignoring the complex correlation of data: Existing methods usually ignore the complex correlation between multidimensional data. Electric vehicle charging load is affected by multiple factors, including electric vehicle charging records, charging station location, grid load, weather conditions, traffic flow, etc. Existing technologies often do not fully consider the relationship between these factors, resulting in low prediction accuracy.
[0004] Lack of identification and retention of key features: Existing methods lack effective feature selection and dimensionality reduction in data preprocessing, and fail to accurately identify and retain key features that have an important impact on charging load prediction. Redundant and noisy features in the data may interfere with model training and prediction, reducing model performance.
[0005] Failure to fully utilize the characteristics of time series data: Electric vehicle charging load data has obvious time series characteristics, and existing methods often fail to effectively capture and utilize this characteristic. Traditional statistical analysis methods and simple machine learning models have difficulty in handling long-term and short-term dependencies in time series data, affecting the accuracy and stability of predictions. Summary of the invention
[0006] Based on the above objectives, the present invention provides a method and system for predicting electric vehicle charging load based on data correlation.
[0007] A method for predicting electric vehicle charging load based on data correlation comprises the following steps:
[0008] S100, collects multi-dimensional data from multiple data sources, the multi-dimensional data including electric vehicle charging records, charging station locations, grid load, weather data, and traffic flow;
[0009] S200, using a data correlation analysis method to analyze the correlation between multidimensional data, and marking the data with strong correlation;
[0010] S300, based on the result of data correlation analysis, a dimension reduction method is used to perform feature selection on the data and construct a feature vector, and the associated labeled data is considered in the process of constructing the feature vector;
[0011] S400, based on the machine learning model, uses the constructed feature vector as input to train historical data and establish a charging load prediction model;
[0012] S500, charging load prediction: input the multi-dimensional data collected in real time into an optimized charging load prediction model to obtain a charging load prediction result of the electric vehicle.
[0013] S600, result output: output the prediction results in the form of charts and reports for reference in power grid scheduling and planning.
[0014] Optionally, the collection of multidimensional data in S100 includes:
[0015] S110, acquiring an electric vehicle charging record from an electric vehicle management system, wherein the charging record includes a charging start time, a charging end time, a charging amount, a battery status, and vehicle location information;
[0016] S120, obtaining charging station location data from a charging station operator or a public charging station management system, where the location data includes the geographical coordinates of the charging station, the number of charging piles, and the charging power;
[0017] S130, obtaining power grid load data from a power grid operator, wherein the load data includes historical load records, real-time load data, peak and valley load distribution, and regional load information;
[0018] S140, obtaining weather data from a meteorological department or a meteorological data service provider, wherein the weather data includes rainfall, wind speed, and weather forecast information;
[0019] S150, obtaining traffic flow data from a traffic management department or a traffic data service provider, wherein the traffic flow data includes road congestion conditions, vehicle speeds, and traffic accident information;
[0020] S160, preprocessing the collected multidimensional data, the preprocessing including data cleaning operation and data normalization operation.
[0021] Optionally, the S200 specifically includes:
[0022] S210, using the Pearson correlation coefficient method to analyze the linear correlation between each characteristic variable and the electric vehicle charging load, and calculating the correlation coefficient between each characteristic variable and the charging load, where the characteristic variable is a specific data item in the multidimensional data;
[0023] S220, using a mutual information method to analyze the nonlinear correlation between each characteristic variable and the charging load of the electric vehicle, and calculating a mutual information value between each characteristic variable and the charging load;
[0024] S230, performing a time lag correlation analysis, evaluating the lag relationship between each characteristic variable and the charging load, and calculating a time lag correlation coefficient;
[0025] S240, comprehensively evaluating the results of the linear correlation analysis, the nonlinear correlation analysis, and the time-lag correlation analysis to determine characteristic variables that have a strong correlation with the electric vehicle charging load;
[0026] S250, associate and mark the feature variables determined to be strongly correlated, and record their positions and correlation strengths in the data set, so that these feature variables can be given priority in subsequent feature selection and model training processes.
[0027] Optionally, the Pearson correlation coefficient analysis is expressed as: ,in, and represent the characteristic variables and the electric vehicle charging load respectively, Indicates the number of data points;
[0028] The mutual information analysis is expressed as:
[0029] ,in, and represent the characteristic variables and the electric vehicle charging load respectively, express and The joint probability distribution of and Respectively and The marginal probability distribution of
[0030] The time-lag correlation analysis is expressed as:
[0031] ,in, and Denote the characteristic variable and lag time respectively After the charging load, Represents the number of data points.
[0032] Optionally, the comprehensive evaluation includes standardizing the Pearson correlation coefficient, the mutual information value, and the time-lag correlation coefficient so that their value ranges are consistent, and the standardization is expressed as:
[0033] ;
[0034] Assume that the Pearson correlation coefficient, mutual information value and time-lag correlation coefficient are standardized as , and , then the comprehensive relevance score for:
[0035] ,in, , and is the weight coefficient, reflecting the relative importance of linear, nonlinear and time-lag correlation;
[0036] Related tags: score the comprehensive evaluation results Sort the data, select feature variables for association marking based on the sorting, and record the position and correlation strength of the feature variables in the data set.
[0037] Optionally, the S300 specifically includes:
[0038] S310, the original data matrix is expressed as , whose rows represent samples and columns represent feature variables, assign high weights to the associated marked feature variables, and construct a weighted data matrix , the weighted expression is: ,in, It is The weight of each feature, giving high weight to the features of the associated mark , for other features, assign weights ;
[0039] S320, Calculate weighted data matrix using improved principal component analysis The covariance matrix of :
[0040] ,in, is the sample size, is the transpose of the weighted data matrix, and the covariance matrix Perform eigenvalue decomposition and obtain the eigenvalue and the corresponding eigenvector : , sort the eigenvectors by eigenvalue and select the first eigenvectors, construct the eigenvector matrix :
[0041] ,in, The number of main components to be retained;
[0042] S330, construct feature vector: weighted data matrix Projection to the eigenvector matrix On the above, we get the eigenvector matrix after dimensionality reduction : ,During the process of constructing the feature vector, it is ensured that the associated labeled feature variables occupy an important position in the feature vector after dimensionality reduction and retain their strong correlation.
[0043] Optionally, the establishing of the charging load prediction model in S400 specifically includes:
[0044] S410, based on the characteristics of the feature vector and the requirements of the prediction task, select the LSTM (Long Short-Term Memory Network) model as the machine learning model to process the long-term and short-term dependencies of the time series data;
[0045] S420, performing data segmentation on the constructed feature vector, dividing the feature vector and the corresponding historical charging load data into a training set and a verification set;
[0046] S430, constructing the network structure of the LSTM model, setting the input layer, multiple LSTM layers, a fully connected layer and an output layer, and initializing the model parameters;
[0047] S440, the feature vector matrix of the training set As input, historical charging load data As output, the LSTM model is trained and the model parameters are adjusted to optimize the model performance; the feature vector matrix of the training set is As input, historical charging load data As output, the LSTM model is trained and the training process is expressed as:
[0048] ;
[0049] S450, use the validation set to evaluate the trained LSTM model, calculate the prediction error and model performance indicators, including the mean square error (MSE), and optimize the model based on the evaluation results. The mean square error loss function measures the error between the model prediction value and the actual value: , and finally a charging load prediction model with excellent performance is obtained.
[0050] Optionally, the network structure of constructing the LSTM model in S430 specifically includes:
[0051] Input layer: Set the dimension of the input layer to the length of the feature vector. If the feature vector matrix after dimensionality reduction is The dimension is ,in is the sample size, is the number of features, then the dimension of the input layer is ;
[0052] LSTM layer: Add one or more LSTM layers, each containing several LSTM units. The number of units in the LSTM layer is adjusted according to the size and complexity of the data set. The more units there are, the stronger the learning ability of the model is, but the more computation is required:
[0053] ;
[0054] Fully connected layer: The output of the LSTM layer is converted into a prediction result through a fully connected layer. The fully connected layer is used to continuously extract features and perform linear transformations, which can be expressed as:
[0055] ,in, is the weight matrix, is the bias vector;
[0056] Output layer: Set the output layer as a node to output the predicted charging load value.
[0057] Optionally, the S600 specifically includes:
[0058] S610, visualizing the charging load prediction result to generate relevant charts, including a charging load curve chart and a load distribution chart;
[0059] S620, write a forecast report, and record the forecast results in detail, including the forecast value, forecast error, and forecast time information;
[0060] S630 integrates the generated charts and reports into complete PDF output files for reference by power grid dispatching and planning departments.
[0061] A data-correlation-based electric vehicle charging load prediction system, used to implement the above-mentioned data-correlation-based electric vehicle charging load prediction method, includes the following modules:
[0062] The acquisition module collects multi-dimensional data including electric vehicle charging records, charging station locations, power grid load, weather data, and traffic flow from multiple data sources;
[0063] The preprocessing module is used to clean and normalize the collected multi-dimensional data;
[0064] A correlation analysis module is used to analyze the correlation between multidimensional data using Pearson correlation coefficient, mutual information method and time-lag correlation analysis, and to associate and mark characteristic variables that have a strong correlation with electric vehicle charging load;
[0065] The feature selection and dimension reduction module is used to select features and reduce dimensions of data based on the results of data correlation analysis using an improved principal component analysis method to construct feature vectors;
[0066] The prediction model training module, based on the LSTM model, takes the constructed feature vector as input, trains historical data, and establishes a charging load prediction model;
[0067] The real-time data processing and prediction module is used to input the multi-dimensional data collected in real time into the optimized charging load prediction model to generate the charging load prediction results of electric vehicles;
[0068] The result output module is used to output the prediction results.
[0069] Beneficial effects of the present invention:
[0070] The present invention collects multidimensional data (including electric vehicle charging records, charging station locations, power grid load, weather data and traffic flow), adopts an improved principal component analysis method, performs dimensionality reduction processing on the characteristic variables after comprehensive evaluation, and constructs a characteristic vector. In the dimensionality reduction process, the characteristic variables with strong correlation are associated and marked and given higher weights, so that the key features are fully retained and the validity of the model input data is improved. The long short-term memory network (LSTM) model is selected to make full use of its advantages in processing the long-term and short-term dependencies of time series data to perform high-precision prediction of the charging load.
[0071] The present invention proposes a comprehensive data correlation analysis method, combining Pearson correlation coefficient, mutual information, time-lag correlation and other analysis methods to comprehensively evaluate multidimensional data and identify characteristic variables that have a significant impact on charging load. In the dimensionality reduction process, the characteristic variables after comprehensive evaluation are associated and marked, and the weighted principal component analysis method is used to assign higher weights to the associated marked characteristic variables to ensure that these key features are fully expressed in the feature vector after dimensionality reduction, effectively reducing redundant features, retaining the main information in the data, and improving the efficiency of model training and the accuracy of prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0073] Figure 1 A schematic diagram of a prediction method flow chart of an embodiment of the present invention;
[0074] Figure 2 Schematic diagram of the prediction system functional modules according to an embodiment of the present invention. DETAILED DESCRIPTION
[0075] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.
[0076] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).
[0077] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.
[0078] like Figure 1 As shown, a method for predicting electric vehicle charging load based on data correlation includes the following steps:
[0079] S100, collects multi-dimensional data from multiple data sources, the multi-dimensional data including electric vehicle charging records, charging station locations, grid load, weather data, and traffic flow;
[0080] S200, using a data correlation analysis method to analyze the correlation between multidimensional data, and marking the data with strong correlation;
[0081] S300, based on the result of data correlation analysis, a dimension reduction method is used to perform feature selection on the data and construct a feature vector, and the associated labeled data is considered in the process of constructing the feature vector;
[0082] S400, based on the machine learning model, uses the constructed feature vector as input to train historical data and establish a charging load prediction model;
[0083] S500, charging load prediction: inputting the multi-dimensional data collected in real time into an optimized charging load prediction model to obtain a charging load prediction result of the electric vehicle;
[0084] Collect multi-dimensional data such as electric vehicle charging records, charging station locations, power grid load, weather data and traffic flow in real time; pre-process the multi-dimensional data collected in real time; input the pre-processed feature vector into the optimized LSTM charging load prediction model; the optimized LSTM model processes the input data to generate the charging load prediction result of the electric vehicle.
[0085] S600, result output: output the prediction results in the form of charts and reports for reference in power grid scheduling and planning.
[0086] The collection of multi-dimensional data in S100 includes:
[0087] S110, obtaining an electric vehicle charging record from an electric vehicle management system, the charging record including charging start time, charging end time, charging amount, battery status, and vehicle location information;
[0088] S120, obtaining charging station location data from a charging station operator or a public charging station management system, where the location data includes the geographical coordinates of the charging station, the number of charging piles, and the charging power;
[0089] S130, obtaining power grid load data from the power grid operator, the load data including historical load records, real-time load data, peak and valley load distribution, and regional load information;
[0090] S140, obtaining weather data from a meteorological department or a meteorological data service provider, where the weather data includes rainfall, wind speed, and weather forecast information;
[0091] S150, obtaining traffic flow data from a traffic management department or a traffic data service provider, where the traffic flow data includes road congestion conditions, vehicle speeds, and traffic accident information;
[0092] S160, preprocessing the collected multidimensional data, including data cleaning operations and data normalization operations to remove noise and outliers to obtain a high-quality data set.
[0093] Data cleaning is an important step to ensure data quality, which mainly includes the following operations:
[0094] Missing value processing: For data with few missing values and random distribution, directly delete the records containing missing values. For data with many missing values, use the mean and median to fill them.
[0095] Duplicate value processing: Identify and delete identical duplicate records to ensure that each piece of data is unique.
[0096] Data consistency check: Ensure that the data field types (such as integers, floating-point numbers, and strings) are consistent, correct incorrect data types, ensure that data values are within a reasonable range, and correct data that is out of range.
[0097] Data normalization is the operation of converting data of different scales to the same scale range. By using minimum and maximum normalization, it is suitable for situations where the data distribution is relatively uniform.
[0098] By standardization: , which is applicable to the case where the data distribution is close to the normal distribution, where is the mean, is the standard deviation.
[0099] S200 specifically includes:
[0100] S210, using the Pearson correlation coefficient method to analyze the linear correlation between each characteristic variable and the charging load of the electric vehicle, and calculating the correlation coefficient between each characteristic variable and the charging load, the characteristic variable is a specific data item in the multidimensional data (charging start time, charging end time, charging amount, battery status and vehicle location information; geographical coordinates of the charging station, number of charging piles, charging power; historical load records, real-time load data, peak and valley load distribution and regional load information; rainfall, wind speed and weather forecast information; road congestion, vehicle speed, traffic accident information);
[0101] S220, using a mutual information method to analyze the nonlinear correlation between each characteristic variable and the charging load of the electric vehicle, and calculating a mutual information value between each characteristic variable and the charging load;
[0102] S230, performing a time lag correlation analysis, evaluating the lag relationship between each characteristic variable and the charging load, and calculating a time lag correlation coefficient;
[0103] S240, comprehensively evaluating the results of the linear correlation analysis, the nonlinear correlation analysis, and the time-lag correlation analysis to determine characteristic variables that have a strong correlation with the electric vehicle charging load;
[0104] S250, associate and mark the feature variables determined to be strongly correlated, and record their positions and correlation strengths in the data set, so that these feature variables can be given priority in subsequent feature selection and model training processes.
[0105] The Pearson correlation coefficient analysis is expressed as: ,in, and represent the characteristic variables and the electric vehicle charging load respectively, Indicates the number of data points, Pearson correlation coefficient It is used to measure the linear correlation between two variables. The value range is between -1 and 1. The larger the absolute value, the stronger the correlation. The correlation coefficient between each characteristic variable and the charging load is calculated to screen out the characteristic variables with higher correlation coefficients.
[0106] The mutual information analysis is expressed as:
[0107] ,in, and represent the characteristic variables and the electric vehicle charging load respectively, express and The joint probability distribution of and Respectively and The marginal probability distribution, mutual information It is used to measure the nonlinear correlation between two variables. The larger the value, the stronger the correlation. The mutual information value between each characteristic variable and the charging load is calculated to select the characteristic variables with higher mutual information values.
[0108] The time-lag correlation analysis is expressed as:
[0109] ,in, and Denote the characteristic variable and lag time respectively After the charging load, Indicates the number of data points, time-lag correlation coefficient It is used to evaluate the time lag effect between characteristic variables and charging load, identify characteristic variables with significant time lag correlation, and screen out characteristic variables with significant time lag correlation by calculating the correlation between characteristic variables and charging load at different lag times.
[0110] The comprehensive evaluation includes standardizing the Pearson correlation coefficient, mutual information value, and time-lag correlation coefficient to make their value ranges consistent. The standardized expression is:
[0111] ;
[0112] Assume that the Pearson correlation coefficient, mutual information value and time-lag correlation coefficient are standardized as , and , then the comprehensive relevance score for:
[0113] ,in, , and is the weight coefficient, which reflects the relative importance of linear, nonlinear and time-lag correlation, and its value is 0.33; through comprehensive evaluation, it is determined which characteristic variables have a significant impact on the charging load, and the correlation scores of these variables are Sort and select feature variables with high scores;
[0114] Related tags: score the comprehensive evaluation results Sort, select feature variables for association marking based on the sort, record the position and relevance strength of the feature variable in the data set, and add tag information to the metadata of the data set so that these feature variables can be identified and prioritized in subsequent steps to ensure that highly relevant features can be fully utilized during model training. Select feature variables based on sorting According to the percentage method, sort from high to low by comprehensive score S, and select the feature variables with the top 20% comprehensive score for association marking.
[0115] S300 specifically includes:
[0116] S310, the original data matrix is expressed as , whose rows represent samples and columns represent feature variables, assign high weights to the associated marked feature variables, and construct a weighted data matrix , the weighted expression is: ,in, It is The weight of each feature, giving high weight to the features of the associated mark , for other features, assign weights ;
[0117] S320, Calculate weighted data matrix using improved principal component analysis The covariance matrix of :
[0118] ,in, is the sample size, is the transpose of the weighted data matrix, and the covariance matrix Perform eigenvalue decomposition and obtain the eigenvalue and the corresponding eigenvector : , sort the eigenvectors by eigenvalue and select the first eigenvectors, construct the eigenvector matrix :
[0119] ,in, The number of main components to be retained;
[0120] S330, construct feature vector: weighted data matrix Projection to the eigenvector matrix On the above, we get the eigenvector matrix after dimensionality reduction : ,During the process of constructing the feature vector, it is ensured that the associated labeled feature variables occupy an important position in the feature vector after dimensionality reduction and retain their strong correlation.
[0121] The establishment of the charging load prediction model in S400 specifically includes:
[0122] S410, based on the characteristics of the feature vector and the requirements of the prediction task, select the LSTM (Long Short-Term Memory Network) model as the machine learning model to process the long-term and short-term dependencies of the time series data;
[0123] S420, performing data segmentation on the constructed feature vector, dividing the feature vector and the corresponding historical charging load data into a training set and a verification set;
[0124] S430, constructing the network structure of the LSTM model, setting the input layer, multiple LSTM layers, a fully connected layer and an output layer, and initializing the model parameters;
[0125] S440, the feature vector matrix of the training set As input, historical charging load data As output, the LSTM model is trained and the model parameters are adjusted to optimize the model performance; the feature vector matrix Contains the main components extracted after dimensionality reduction. Each row represents a sample and each column represents a new feature. After dimensionality reduction, these features retain most of the information of the original data while removing redundancy and noise. The feature vector matrix of the training set As input, historical charging load data As output, the LSTM model is trained and the training process is expressed as:
[0126] ;
[0127] S450, use the validation set to evaluate the trained LSTM model, calculate the prediction error and model performance indicators, including the mean square error (MSE), and optimize the model based on the evaluation results. The mean square error loss function measures the error between the model prediction value and the actual value: , and finally a charging load prediction model with excellent performance is obtained.
[0128] The network structure of constructing the LSTM model in S430 specifically includes:
[0129] Input layer: Set the dimension of the input layer to the length of the feature vector. If the feature vector matrix after dimensionality reduction is The dimension is ,in is the sample size, is the number of features, then the dimension of the input layer is ;
[0130] LSTM layer: Add one or more LSTM layers, each containing several LSTM units. The number of units in the LSTM layer is adjusted according to the size and complexity of the data set. The more units there are, the stronger the learning ability of the model is, but the more computation is required:
[0131] ;
[0132] Fully connected layer: The output of the LSTM layer is converted into a prediction result through a fully connected layer. The fully connected layer is used to continuously extract features and perform linear transformations, which can be expressed as:
[0133] ,in, is the weight matrix, is the bias vector;
[0134] Output layer: Set the output layer as a node to output the predicted charging load value.
[0135] S600 specifically includes:
[0136] S610, visualizing the charging load prediction result to generate relevant charts, including a charging load curve chart and a load distribution chart;
[0137] S620, write a forecast report, and record the forecast results in detail, including the forecast value, forecast error, and forecast time information;
[0138] S630 integrates the generated charts and reports into complete PDF output files for reference by power grid dispatching and planning departments.
[0139] Visualization:
[0140] Charging load curve chart: Draw a curve chart of the predicted charging load changes over time to show the load change trend.
[0141] Horizontal axis: time (hours, days);
[0142] Vertical axis: predicted charging load;
[0143] Load distribution diagram: Plot the distribution of predicted charging load in different time periods or regions.
[0144] Horizontal axis: time period or region;
[0145] Vertical axis: load value or load distribution density;
[0146] Writing forecast reports:
[0147] Forecast value: Displays the specific forecast value and its change over time.
[0148] Prediction Error: Calculates and displays the prediction error (mean squared error MSE).
[0149] Forecast time: record the forecast time and forecast period.
[0150] like Figure 2 As shown, a system for predicting electric vehicle charging load based on data correlation is used to implement the above-mentioned method for predicting electric vehicle charging load based on data correlation, and includes the following modules:
[0151] The acquisition module collects multi-dimensional data including electric vehicle charging records, charging station locations, power grid load, weather data, and traffic flow from multiple data sources;
[0152] The preprocessing module is used to clean and normalize the collected multi-dimensional data;
[0153] A correlation analysis module is used to analyze the correlation between multidimensional data using Pearson correlation coefficient, mutual information method and time-lag correlation analysis, and to associate and mark characteristic variables that have a strong correlation with electric vehicle charging load;
[0154] The feature selection and dimension reduction module is used to select features and reduce dimensions of data based on the results of data correlation analysis using an improved principal component analysis method to construct feature vectors;
[0155] The prediction model training module, based on the LSTM model, takes the constructed feature vector as input, trains historical data, and establishes a charging load prediction model;
[0156] The real-time data processing and prediction module is used to input the multi-dimensional data collected in real time into the optimized charging load prediction model to generate the charging load prediction results of electric vehicles;
[0157] The result output module is used to output the prediction results.
[0158] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.
[0159] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A method for predicting electric vehicle charging load based on data correlation, characterized in that: The following steps are involved: S100, collecting multidimensional data from multiple data sources, the multidimensional data including electric vehicle charging records, charging station locations, power grid load, weather data, and traffic flow, the collection of the multidimensional data including: S110, acquiring an electric vehicle charging record from an electric vehicle management system, wherein the charging record includes a charging start time, a charging end time, a charging amount, a battery status, and vehicle location information; S120, obtaining charging station location data from a charging station operator or a public charging station management system, where the location data includes the geographical coordinates of the charging station, the number of charging piles, and the charging power; S130, obtaining power grid load data from a power grid operator, wherein the load data includes historical load records, real-time load data, peak and valley load distribution, and regional load information; S140, obtaining weather data from a meteorological department or a meteorological data service provider, wherein the weather data includes rainfall, wind speed, and weather forecast information; S150, obtaining traffic flow data from a traffic management department or a traffic data service provider, wherein the traffic flow data includes road congestion conditions, vehicle speeds, and traffic accident information; S160, preprocessing the collected multidimensional data, the preprocessing including data cleaning operation and data normalization operation; S200, using a data correlation analysis method to analyze the correlation between multidimensional data, and marking the data with strong correlation, specifically including: S210, using the Pearson correlation coefficient method to analyze the linear correlation between each characteristic variable and the electric vehicle charging load, and calculating the correlation coefficient between each characteristic variable and the charging load, where the characteristic variable is a specific data item in the multidimensional data; S220, using a mutual information method to analyze the nonlinear correlation between each characteristic variable and the charging load of the electric vehicle, and calculating a mutual information value between each characteristic variable and the charging load; S230, performing a time lag correlation analysis, evaluating the lag relationship between each characteristic variable and the charging load, and calculating a time lag correlation coefficient; S240, comprehensively evaluating the results of the linear correlation analysis, the nonlinear correlation analysis, and the time-lag correlation analysis to determine characteristic variables that have a strong correlation with the electric vehicle charging load; S250, associate marking the feature variables determined to be strongly correlated, and record their positions and correlation strengths in the data set; The Pearson correlation coefficient analysis is expressed as: ,in, and represent the characteristic variables and the electric vehicle charging load respectively, Indicates the number of data points; The mutual information analysis is expressed as: ,in, and represent the characteristic variables and the electric vehicle charging load respectively, express and The joint probability distribution of and Respectively and The marginal probability distribution of The time-lag correlation analysis is expressed as: ,in, and represent the characteristic variables and lag time respectively After the charging load, Indicates the number of data points; The comprehensive evaluation includes standardizing the Pearson correlation coefficient, the mutual information value and the time-lag correlation coefficient so that their value ranges are consistent; Assume that the Pearson correlation coefficient, mutual information value and time-lag correlation coefficient are standardized as , and , then the comprehensive relevance score for: ,in, , and is the weight coefficient, reflecting the relative importance of linear, nonlinear and time-lag correlation; Related tags: score the comprehensive evaluation results Sorting is performed, and characteristic variables are selected according to the sorting for association marking, and the position and correlation strength of the characteristic variables in the data set are recorded; S300, based on the result of data correlation analysis, adopts dimension reduction method to select features of data and construct feature vectors. In the process of constructing feature vectors, the data with associated tags are considered, which specifically includes: S310, the original data matrix is expressed as , whose rows represent samples and columns represent feature variables, assign high weights to the associated marked feature variables, and construct a weighted data matrix , the weighted expression is: ,in, It is The weight of each feature, giving high weight to the features of the associated mark , for other features, assign weights ; S320, Calculate weighted data matrix using improved principal component analysis The covariance matrix of : ,in, is the sample size, is the transpose of the weighted data matrix, and the covariance matrix Perform eigenvalue decomposition and obtain the eigenvalue and the corresponding eigenvector : , sort the eigenvectors by eigenvalue and select the first eigenvectors, construct the eigenvector matrix : ,in, The number of main components to be retained; S330, construct feature vector: weighted data matrix Projection to the eigenvector matrix On the above, we get the eigenvector matrix after dimensionality reduction : ,In the process of constructing the feature vector, ensure that the associated marked feature variables occupy an important position in the feature vector after dimensionality reduction and retain their strong correlation; S400, based on the machine learning model, and using the constructed feature vector as input, training the historical data to establish a charging load prediction model, wherein establishing the charging load prediction model specifically includes: S410, selecting an LSTM model as a machine learning model to process long-term and short-term dependencies of time series data; S420, performing data segmentation on the constructed feature vector, dividing the feature vector and the corresponding historical charging load data into a training set and a verification set; S430, constructing the network structure of the LSTM model, setting the input layer, multiple LSTM layers, a fully connected layer and an output layer, and initializing the model parameters; S440, the feature vector matrix of the training set As input, historical charging load data As output, the LSTM model is trained and the model parameters are adjusted to optimize the model performance; the feature vector matrix of the training set is As input, historical charging load data As output, the LSTM model is trained and the training process is expressed as: ； S450, use the validation set to evaluate the trained LSTM model, calculate the prediction error and model performance indicators, including the mean square error, and optimize the model based on the evaluation results. The mean square error loss function measures the error between the model prediction value and the actual value: , and finally obtain the charging load prediction model; S500, charging load prediction: inputting the multi-dimensional data collected in real time into an optimized charging load prediction model to obtain a charging load prediction result of the electric vehicle; S600, result output: output the prediction results in the form of charts and reports for reference in power grid scheduling and planning.
2. The method for predicting electric vehicle charging load based on data correlation according to claim 1 is characterized in that: The network structure of constructing the LSTM model in S430 specifically includes: Input layer: Set the dimension of the input layer to the length of the feature vector; LSTM layer: Add one or more LSTM layers, each containing a number of LSTM units. The number of units in the LSTM layer is adjusted according to the size and complexity of the dataset: ； Fully connected layer: The output of the LSTM layer is converted into a prediction result through a fully connected layer. The fully connected layer is used to continuously extract features and perform linear transformations, which can be expressed as: ,in, is the weight matrix, is the bias vector; Output layer: Set the output layer as a node to output the predicted charging load value.
3. The method for predicting electric vehicle charging load based on data correlation according to claim 1 is characterized in that: The S600 specifically includes: S610, visualizing the charging load prediction result to generate relevant charts, including a charging load curve chart and a load distribution chart; S620, write a forecast report, and record the forecast results in detail, including the forecast value, forecast error, and forecast time information; S630 integrates the generated charts and reports into complete output files for reference by the power grid dispatching and planning departments.
4. A data correlation-based electric vehicle charging load prediction system, used to implement the prediction method according to any one of claims 1 to 3, characterized in that: Includes the following modules: The acquisition module collects multi-dimensional data including electric vehicle charging records, charging station locations, power grid load, weather data, and traffic flow from multiple data sources; The preprocessing module is used to clean and normalize the collected multi-dimensional data; A correlation analysis module is used to analyze the correlation between multidimensional data using Pearson correlation coefficient, mutual information method and time-lag correlation analysis, and to associate and mark characteristic variables that have a strong correlation with electric vehicle charging load; The feature selection and dimension reduction module is used to select features and reduce dimensions of data based on the results of data correlation analysis using an improved principal component analysis method to construct feature vectors; The prediction model training module, based on the LSTM model, takes the constructed feature vector as input, trains historical data, and establishes a charging load prediction model; The real-time data processing and prediction module is used to input the multi-dimensional data collected in real time into the optimized charging load prediction model to generate the charging load prediction results of electric vehicles; The result output module is used to output the prediction results.