Charging facility load forecasting method and system based on multi-layer optimization kernel extreme learning machine

Through multi-layer optimization of the kernel limit learning machine model, the load prediction of charging facilities is used using convolution kernel and activation functions, which solves the problems of difficulty in data collection and poor real-time performance in traditional methods, and achieves more accurate and efficient load prediction.

CN114692956BActive Publication Date: 2024-11-01GUANGZHOU POWER SUPPLY BUREAU GUANGDONG POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210241319.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-11
Publication Date
2024-11-01
Estimated Expiration
2042-03-11

AI Technical Summary

Technical Problem

The traditional charging facility load prediction method fails to effectively consider the differences between electric vehicles and fuel vehicles, and the difficulty in collecting data results in inaccurate results, poor real-time performance, large processing scale, and centralized simulation of all electric vehicles is required.

Method used

Using a multi-layer optimization nuclear limit learning machine method, a collection of factors affecting the load of charging facilities is constructed, through historical data cleaning, normalization and feature extraction, a multi-layer optimization nuclear limit learning machine load prediction model is constructed, and the convolution kernel and activation are used. The function is trained and the charging facility load prediction is carried out.

Benefits of technology

It improves the accuracy and real-timeness of charging facility load prediction, reduces the difficulty of data collection, reduces the processing scale, and enhances the reliability of predicted results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114692956B_ABST
    Figure CN114692956B_ABST
Patent Text Reader

Abstract

The present invention discloses a charging facility load prediction method and system based on a multi-layer optimized core extreme learning machine, including: constructing a set of influencing factors of electric vehicle charging load, and forming sample data based on historical data; cleaning abnormal historical data caused by electric vehicle charging station failure and charging station maintenance in the sample data, and obtaining a historical data set during the normal operation of the charging station; normalizing the acquired historical data set during the normal operation of the charging station; extracting features from the normalized historical data set during the normal operation of the charging station to obtain feature data; constructing a multi-layer optimized core extreme learning machine load prediction model; training the multi-layer optimized core extreme learning machine load prediction model through the extracted feature data; and predicting the charging facility load through the trained multi-layer optimized core extreme learning machine load prediction model. The present invention has the advantages of good real-time performance and high accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution networks, and in particular to a charging facility load prediction method, system, computer and storage medium based on a multi-layer optimization kernel extreme learning machine. Background Art

[0002] As one of the effective means to deal with energy crisis and environmental pollution, the number of electric vehicles has exploded in recent years. In recent years, the manufacturing technology of electric vehicles has been continuously improved, and the manufacturing cost has been continuously reduced. The promotion of electric vehicles has gradually received attention from the government and the public. With the continuous increase in the penetration rate of electric vehicles, their charging and operation may bring negative impacts and even challenges to the safety and economy of the power system and transportation system. Therefore, as a prerequisite for planning and scheduling charging stations, it is necessary to correctly predict the load of electric vehicle charging facilities connected to the distribution network, so as to achieve the impact evaluation and business expansion evaluation of the charging facilities connected to the distribution network.

[0003] There are several traditional methods for charging facility load forecasting: 1. Load forecasting is based on the principle of conservation of energy, the oil sales volume of gas stations is equivalently converted into the electricity sales volume of charging stations, and the equal load distance method is used for distribution; 2. The travel chain is used as the basic unit, and characteristic quantities such as time, space and electricity are used to describe the charging and operating status of electric vehicles to obtain the spatiotemporal distribution of the charging facility load; 3. Based on the historical load data of charging facilities, the probability density function of relevant characteristic information is extracted, and then the Monte Carlo method is used for random sampling for simulation.

[0004] However, among the above-mentioned traditional charging facility load forecasting methods, the charging facility load forecasting method based on the principle of conservation of energy does not take into account the differences between electric vehicles and fuel vehicles; the charging facility load forecasting method using the travel chain as the basic unit requires centralized simulation of all electric vehicles in the study area, which has a large processing scale and poor real-time performance; the probability density function parameters obtained by the statistical method fluctuate greatly, and the charging facility load forecasting results obtained are not accurate.

[0005] Moreover, the traditional charging facility load forecasting method has certain limitations. The main reason is that it is difficult to collect the required travel data and charging preferences of electric users, the topological structure of the transportation network and the real-time traffic flow, which leads to incomplete data collection and the results obtained based on this are not reliable. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine with good real-time performance and high accuracy.

[0007] To achieve the above purpose, the technical solution provided by the present invention is:

[0008] A charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine, comprising:

[0009] Construct a set of factors affecting electric vehicle charging load and form sample data based on historical data;

[0010] Clean the abnormal historical data caused by electric vehicle charging station failures and charging station maintenance in the sample data to obtain the historical data set during the normal operation of the charging station;

[0011] Normalize the historical data set obtained during the normal operation of the charging station;

[0012] Perform feature extraction on the normalized historical data set during the normal operation of the charging station to obtain feature data;

[0013] Construct a multi-layer optimization core extreme learning machine load prediction model;

[0014] The multi-layer optimization core extreme learning machine load forecasting model is trained through the extracted feature data;

[0015] The charging facility load forecasting is carried out through the trained multi-layer optimization core extreme learning machine load forecasting model.

[0016] Furthermore, the factors affecting the electric vehicle charging load include historical charging data of electric vehicles, traffic flow at charging stations, meteorological conditions, the number of electric vehicles, and infrastructure conditions around charging stations;

[0017] The electric vehicle historical charging data includes the charging amount, charging time, charging power, and battery capacity of each electric vehicle;

[0018] The meteorological conditions include temperature, precipitation, and weather.

[0019] Furthermore, feature extraction includes: the time series variables of the factors affecting the electric vehicle charging load form a two-dimensional vector, and a convolution operation is performed on the two-dimensional vector using a convolution kernel to obtain a feature mapping vector, that is, feature data.

[0020] Furthermore, the multi-layer optimization kernel extreme learning machine load forecasting model is trained by the extracted feature data, specifically including:

[0021] A1. Set the initial value of the hidden layer n = 0;

[0022] A2, n = n + 1;

[0023] A3. Generate orthogonal random input layer weights and hidden layer biases through the encoder, including the formula:

[0024]

[0025] (1)W, B and I represent the input weight matrix, hidden layer threshold matrix and identity matrix respectively;

[0026] A4. Solve the output layer weight β through the decoder:

[0027] H=G(WX+B) (2)

[0028]

[0029] In formulas (2) and (3), H and X represent the output matrix and input matrix of the hidden layer, respectively, G(·) represents the activation function, and λ represents the regularization coefficient;

[0030] A5. Use the transpose of the output layer weights as the input layer weights of the original network structure, including the formula:

[0031] H p =G((β p ) T H p-1 ) (4)

[0032] (4) In the formula, H p represents the output matrix of the p-th hidden layer, and the output matrix H of the p-1-th hidden layer p-1 As the input of the pth layer;

[0033] A6, determine whether n is less than N-1, if so, return to step A2, otherwise, go to step A7;

[0034] A7. Training the Nth layer kernel parameters and regularization coefficients based on the optimized kernel extreme learning machine:

[0035] Ω ELM =HH T (5)

[0036] Ω ELM (i,j)=h(x i )·h(x j )=K(x i ,x j ) (6)

[0037]

[0038]

[0039] In formulas (5)-(8), Ω ELM represents the kernel matrix, Ω ELM (i,j) represents the element in the i-th row and j-th column of the kernel matrix, K(x i ,x j) represents the kernel function, γ represents the kernel parameter in the radial basis function kernel, and the optimization algorithm optimizes the parameters λ and γ;

[0040] A8. Test the multi-layer optimization kernel extreme learning machine load prediction model until the prediction error no longer decreases and the training ends.

[0041] Furthermore, the load forecasting of charging facilities through the trained multi-layer optimization core extreme learning machine load forecasting model includes:

[0042] Input historical data of factors affecting charging load prediction;

[0043] Normalize the input historical data of charging load prediction influencing factors;

[0044] The convolution operation is performed using the same convolution kernel as in training to form the input data of the multi-layer optimized kernel extreme learning machine load prediction model;

[0045] The input data after convolution operation is input into the trained multi-layer optimization kernel extreme learning machine load forecasting model to obtain the output data, which is the load forecast result of the charging facility.

[0046] To achieve the above-mentioned object, the present invention further provides a charging facility load prediction system based on a multi-layer optimization kernel extreme learning machine, comprising a set construction module, a database module, a data cleaning module, a normalization module, a feature extraction module, a model construction module, a model training module, and a model prediction module;

[0047] The set building module is used to build a set of factors affecting the electric vehicle charging load;

[0048] The database module is used to store sample data formed by historical data;

[0049] The data cleaning module is used to clean the abnormal historical data caused by the failure of the electric vehicle charging station and the maintenance of the charging station in the sample data, and obtain the historical data set during the normal operation of the charging station;

[0050] The normalization module is used to normalize the historical data set acquired during the normal operation of the charging station;

[0051] The feature extraction module is used to extract features from the normalized historical data set during the normal operation of the charging station;

[0052] The model building module is used to build a multi-layer optimization core extreme learning machine load prediction model;

[0053] The model training module is used to train the multi-layer optimization core extreme learning machine load prediction model through the extracted feature data;

[0054] The model prediction module is used to predict the load of charging facilities through a trained multi-layer optimization core extreme learning machine load prediction model.

[0055] To achieve the above objectives, the present invention further provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the above-mentioned charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine are implemented.

[0056] To achieve the above-mentioned purpose, the present invention further provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine.

[0057] Compared with the prior art, the principles and advantages of this technical solution are as follows:

[0058] This technical solution is based on the historical load data of charging facilities and constructs a charging facility load prediction model through a multi-layer optimization kernel extreme learning machine algorithm. There is no need to collect travel data and charging preferences of electric users, the topological structure of the transportation network and real-time traffic, etc., which reduces the difficulty of data collection while enhancing the reliability of the prediction results. In addition, the model does not need to conduct centralized simulation of all electric vehicles in the study area, and the processing scale is small, so the real-time performance is better. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the services required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 This is a principle flow chart of offline training in a charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine of the present invention;

[0061] Figure 2 It is a principle flow chart of online prediction in a charging facility load prediction method based on a multi-layer optimization kernel extreme learning machine of the present invention;

[0062] Figure 3 The present invention is a connection block diagram of a charging facility load forecasting system based on a multi-layer optimization kernel extreme learning machine. DETAILED DESCRIPTION

[0063] The present invention will be further described below in conjunction with specific embodiments:

[0064] The charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine described in this embodiment includes two parts: offline training and online forecasting;

[0065] like Figure 1 As shown, offline training includes the following steps:

[0066] S1. Construct a set of factors affecting electric vehicle charging load and form sample data based on historical data;

[0067] Factors affecting electric vehicle charging load include historical electric vehicle charging data, traffic flow at charging stations, weather conditions, the number of electric vehicles in use, and infrastructure conditions around charging stations;

[0068] The electric vehicle historical charging data includes the charging amount, charging time, charging power, and battery capacity of each electric vehicle;

[0069] The meteorological conditions include temperature, precipitation, and weather.

[0070] S2. Clean the abnormal historical data caused by the failure of electric vehicle charging stations and the maintenance of charging stations in the sample data to obtain the historical data set during the normal operation of the charging stations;

[0071] S3, normalizing the historical data set obtained during the normal operation of the charging station;

[0072] S4. Extracting features from the normalized historical data set during the normal operation of the charging station to obtain feature data; specifically including:

[0073] The time series variables of the factors affecting the charging load of electric vehicles form a two-dimensional vector. The convolution kernel is used to perform convolution operation on the two-dimensional vector to obtain a feature mapping vector.

[0074] S5. Construct a multi-layer optimization core extreme learning machine load prediction model;

[0075] S6. Training the multi-layer optimization core extreme learning machine load prediction model through the extracted feature data, specifically including:

[0076] A1. Set the initial value of the hidden layer n = 0;

[0077] A2, n = n + 1;

[0078] A3. Generate orthogonal random input layer weights and hidden layer biases through the encoder, including the formula:

[0079]

[0080] (1)W, B and I represent the input weight matrix, hidden layer threshold matrix and identity matrix respectively;

[0081] A4. Solve the output layer weight β through the decoder:

[0082] H=G(WX+B) (2)

[0083]

[0084] , (3)In the formula, H and X represent the output matrix and input matrix of the hidden layer respectively, G(·) represents the activation function, and λ represents the regularization coefficient;

[0085] A5. Use the transpose of the output layer weights as the input layer weights of the original network structure, including the formula:

[0086] H p =G((β p ) T H p-1 ) (4)

[0087] (4) In the formula, H p represents the output matrix of the p-th hidden layer, and the output matrix H of the p-1-th hidden layer p-1 As the input of the pth layer;

[0088] A6, determine whether n is less than N-1, if so, return to step A2, otherwise, go to step A7;

[0089] A7. Training the Nth layer kernel parameters and regularization coefficients based on the optimized kernel extreme learning machine:

[0090] Ω ELM =HH T (5)

[0091] Ω ELM (i,j)=h(x i )·h(x j )=K(x i ,x j ) (6)

[0092]

[0093]

[0094] In formulas (5)-(8), Ω ELM represents the kernel matrix, Ω ELM (i,j) represents the element in the i-th row and j-th column of the kernel matrix, K(x i ,x j ) represents the kernel function, γ represents the kernel parameter in the radial basis function kernel, and the optimization algorithm optimizes the parameters λ and γ;

[0095] A8. Test the multi-layer optimization kernel extreme learning machine load prediction model until the prediction error no longer decreases and the training ends.

[0096] like Figure 2 As shown, the online prediction is:

[0097] The charging facility load forecasting is performed through the trained multi-layer optimization core extreme learning machine load forecasting model, including:

[0098] Input historical data of factors affecting charging load prediction;

[0099] Normalize the input historical data of charging load prediction influencing factors;

[0100] The convolution operation is performed using the same convolution kernel as in training to form the input data of the multi-layer optimized kernel extreme learning machine load prediction model;

[0101] The input data after convolution operation is input into the trained multi-layer optimization kernel extreme learning machine load forecasting model to obtain the output data, which is the load forecast result of the charging facility.

[0102] The present invention also provides a charging facility load forecasting system based on a multi-layer optimization kernel extreme learning machine, such as Figure 3 As shown, it includes a collection construction module, a database module, a data cleaning module, a normalization module, a feature extraction module, a model construction module, a model training module, and a model prediction module;

[0103] The set building module is used to build a set of factors affecting the electric vehicle charging load;

[0104] The database module is used to store sample data formed by historical data;

[0105] The data cleaning module is used to clean the abnormal historical data caused by the failure of the electric vehicle charging station and the maintenance of the charging station in the sample data, and obtain the historical data set during the normal operation of the charging station;

[0106] The normalization module is used to normalize the historical data set acquired during the normal operation of the charging station;

[0107] The feature extraction module is used to extract features from the normalized historical data set during the normal operation of the charging station;

[0108] The model building module is used to build a multi-layer optimization core extreme learning machine load prediction model;

[0109] The model training module is used to train the multi-layer optimization core extreme learning machine load prediction model through the extracted feature data;

[0110] The model prediction module is used to predict the load of charging facilities through a trained multi-layer optimization core extreme learning machine load prediction model.

[0111] The present invention further provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine when executing the computer program.

[0112] The present invention further provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine.

[0113] The embodiments described above are only preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. Therefore, all changes made according to the shape and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine, characterized in that: include: Construct a set of factors affecting electric vehicle charging load and form sample data based on historical data; Clean the abnormal historical data caused by electric vehicle charging station failures and charging station maintenance in the sample data to obtain the historical data set during the normal operation of the charging station; Normalize the historical data set obtained during the normal operation of the charging station; Perform feature extraction on the normalized historical data set during the normal operation of the charging station to obtain feature data; Construct a multi-layer optimization core extreme learning machine load prediction model; The multi-layer optimization core extreme learning machine load forecasting model is trained through the extracted feature data; The charging facility load is predicted through the trained multi-layer optimization core extreme learning machine load prediction model; The training of the multi-layer optimization kernel extreme learning machine load forecasting model through the extracted feature data specifically includes: A1. Set the initial value of the hidden layer n = 0; A2, n = n + 1; A3. Generate orthogonal random input layer weights and hidden layer biases through the encoder, including the formula: (1)W, B and I represent the input weight matrix, hidden layer threshold matrix and identity matrix respectively; A4. Solve the output layer weight β through the decoder: H=G(WX+B)(2) In formulas (2) and (3), H and X represent the output matrix and input matrix of the hidden layer, respectively, G(·) represents the activation function, and λ represents the regularization coefficient; A5. Use the transpose of the output layer weights as the input layer weights of the original network structure, including the formula: H p =G((β p ) T H p-1 ) (4) (4) In the formula, H p represents the output matrix of the p-th hidden layer, and the output matrix H of the p-1-th hidden layer p-1 As the input of the pth layer; A6, determine whether n is less than N-1, if so, return to step A2, otherwise, go to step A7; A7. Training the Nth layer kernel parameters and regularization coefficients based on the optimized kernel extreme learning machine: Oh ELM =HH T (5) Ω ELM (i,j)=h(x i )·h(x j )=K(x i ,x j ) (6) In formulas (5)-(8), Ω ELM represents the kernel matrix, Ω ELM (i,j) represents the element in the i-th row and j-th column of the kernel matrix, K(x i , x j ) represents the kernel function, γ represents the kernel parameter in the radial basis function kernel, and the optimization algorithm optimizes the parameters λ and γ; A8. Test the multi-layer optimization core extreme learning machine load prediction model until the prediction error no longer decreases and the training ends; The load forecasting model of multi-layer optimization kernel extreme learning machine is used to forecast the load of charging facilities, including: Input historical data of factors affecting charging load prediction; Normalize the input historical data of charging load prediction influencing factors; The convolution operation is performed using the same convolution kernel as in training to form the input data of the multi-layer optimized kernel extreme learning machine load prediction model; The input data after convolution operation is input into the trained multi-layer optimization kernel extreme learning machine load forecasting model to obtain the output data, which is the load forecast result of the charging facility.

2. The charging facility load forecasting method based on multi-layer optimization kernel extreme learning machine according to claim 1 is characterized in that: The factors affecting the electric vehicle charging load include historical charging data of electric vehicles, traffic flow at charging stations, weather conditions, the number of electric vehicles in use, and infrastructure conditions around charging stations; The electric vehicle historical charging data includes the charging amount, charging time, charging power, and battery capacity of each electric vehicle; The meteorological conditions include temperature, precipitation, and weather.

3. The charging facility load forecasting method based on multi-layer optimization kernel extreme learning machine according to claim 1 is characterized in that: Feature extraction includes: the time series variables of the factors affecting the electric vehicle charging load form a two-dimensional vector, and a convolution kernel is used to perform a convolution operation on the two-dimensional vector to obtain a feature mapping vector, that is, feature data.

4. A charging facility load forecasting system based on a multi-layer optimization kernel extreme learning machine, used to implement the charging facility load forecasting method based on a multi-layer optimization kernel extreme learning machine according to claim 1, characterized in that: It includes a collection construction module, a database module, a data cleaning module, a normalization module, a feature extraction module, a model construction module, a model training module, and a model prediction module; The set building module is used to build a set of factors affecting the electric vehicle charging load; The database module is used to store sample data formed by historical data; The data cleaning module is used to clean the abnormal historical data caused by the failure of the electric vehicle charging station and the maintenance of the charging station in the sample data, and obtain the historical data set during the normal operation of the charging station; The normalization module is used to normalize the historical data set acquired during the normal operation of the charging station; The feature extraction module is used to extract features from the normalized historical data set during the normal operation of the charging station; The model building module is used to build a multi-layer optimization core extreme learning machine load prediction model; The model training module is used to train the multi-layer optimization core extreme learning machine load prediction model through the extracted feature data; The model prediction module is used to predict the load of charging facilities through a trained multi-layer optimization core extreme learning machine load prediction model.

5. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.