Electric vehicle load probabilistic short-term prediction method based on extreme learning machine

Abstract

The invention discloses an electric vehicle load probabilistic short-term prediction method based on an extreme learning machine, and the method comprises the steps: screening factors which affect the daily charging amount of an electric vehicle load, carrying out the quantification and normalization processing of needed data, and employing a Pearson correlation coefficient to screen related factors in the data; constructing an extreme learning machine prediction model based on the number of the input features and different hidden layer neuron verification results of the extreme learning machine; training an extreme learning machine prediction model by using the normalized historical data of the selected related factors, and calculating an upper bound and a lower bound of a confidence interval according to a residual standard deviation of a predicted value and a true value of a model training set and the selected confidence level; and inputting the extreme learning machine prediction model by using the data of the selected related factors, predicting the future load of the electric vehicle, and obtaining an electric vehicle load probabilistic prediction result according to the upper bound and the lower bound of the confidence interval.

Description

Technical Field

[0001] The invention relates to the technical field of electric vehicles, and in particular to a probabilistic short-term prediction method for electric vehicle load based on an extreme learning machine. Background Art

[0002] In recent years, with the enhancement of environmental awareness and the continuous advancement of technology, electric vehicles have become the focus of favor in various markets with their many advantages such as no pollution, zero emissions, and low noise. Although it will take some time to completely replace traditional fuel vehicles, under the general trend of global "carbon neutrality", the replacement process of electric vehicles will continue to advance. However, the randomness of electric vehicle charging behavior has brought a series of challenges to the power grid. First, the randomness of charging behavior may lead to a decline in the quality of power grid power, especially during peak hours. A large number of electric vehicles charging at the same time may cause the power grid to be overloaded, affecting the stability and reliability of power supply.

[0003] Therefore, accurate prediction of electric vehicle load is the basis of orderly regulation. Traditional point prediction cannot fully reflect the dynamic changes of its load. Probabilistic prediction of future electric vehicle load can provide more uncertain information to power suppliers. The use of extreme learning machine for probabilistic prediction of electric vehicle load has the advantages of fast training speed, strong generalization ability, strong approximation ability and strong interpretability, which can effectively improve the accuracy and reliability of the prediction. Summary of the invention

[0004] The present invention provides a probabilistic short-term prediction method for electric vehicle load based on extreme learning machine, quantifies and screens factors affecting the daily charging amount of electric vehicle load, constructs an extreme learning machine prediction model, calculates the upper and lower bounds of the confidence interval according to the residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level, and performs probabilistic prediction of the short-term load of the electric vehicle.

[0005] The present invention provides a probabilistic short-term prediction method for electric vehicle load based on an extreme learning machine, which is characterized by comprising the following steps:

[0006] Screen the factors that affect the daily charging capacity of electric vehicle load, quantify and normalize the required data, and use the Pearson correlation coefficient to screen the relevant factors;

[0007] Based on the number of input features and the verification results of different hidden layer neurons of the extreme learning machine, an extreme learning machine prediction model is constructed;

[0008] The normalized historical data of the selected relevant factors are used to train the extreme learning machine prediction model, and the upper and lower bounds of the confidence interval are calculated based on the residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level;

[0009] The data of the selected relevant factors are used as input to the extreme learning machine prediction model to predict the future load of electric vehicles, and the probabilistic prediction results of the electric vehicle load are obtained according to the upper and lower bounds of the confidence interval.

[0010] Furthermore, the Pearson correlation coefficient is used to screen the relevant factors, including:

[0011] The temperature, weather and day type data collected at the same time as the historical load data of electric vehicles are extracted and quantified into quantitative values ​​of daily average temperature, weather and day type. The load of electric vehicles at the same moment a few days ago is selected, and the Pearson correlation coefficient is used as the screening basis to extract characteristic data that is correlated with the electric vehicle load as the model input.

[0012] Further, n variables x1, x2...xn, time period t0~t, data sampling period T, time lag range ΔT, time shift step Δt are selected;

[0013] Draw the time series curve of each variable, including separate drawing and centralized drawing modes, which can be switched by button;

[0014] Calculate the correlation coefficients between each variable and their corresponding time lags to form the correlation coefficient matrix Rp and the time lag matrix RΔt;

[0015] Based on data analysis and visualization tools, a correlation matrix diagram is drawn, which can be linked with the variable curve to obtain the most relevant factors affecting the daily charging amount of electric vehicle load.

[0016] Furthermore, the calculation process of the correlation coefficient between each variable is as follows: for two vectors X = (x1, ..., xn) and Y = (y1, ..., yn);

[0017]

[0018] Where n is the vector dimension, and are the means of vectors X and Y respectively.

[0019] Furthermore, the time lag is obtained as follows:

[0020] For every vector of two variables, move Δt each time within the time lag range ΔT and calculate the Pearson correlation coefficient between the corresponding two vectors. A total of ΔT / Δt moves are made. The one with the largest absolute value is selected as the correlation coefficient considering the time lag factor, and the moving time is taken as the time lag.

[0021] Furthermore, the verification results of different hidden layer neurons include:

[0022] Based on the influence of the number of hidden layer nodes on the prediction accuracy of the model, a range traversal is performed on the number of hidden layer nodes. When the number of hidden layer nodes exceeds a certain threshold, the extreme learning machine prediction model will have stable generalization performance. The number of hidden layer nodes slightly larger than the threshold is selected to ensure the generalization ability of the model and the low complexity of the model.

[0023] Furthermore, the residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level include:

[0024] The extreme learning machine prediction model obtained by the training is used, and the training set data is used as the input of the prediction model to obtain the prediction result of the training set data of the prediction model, and the residual between the training set target data and the training set target data is obtained, and the standard deviation is calculated. The residual formula is as follows:

[0025] R d =Y1-Y0

[0026] Where Y1 is the real recorded data in the training set; Y0 is the predicted value obtained by the prediction model based on the training set input; R d is the residual between the real recorded data in the training set and the output data predicted by the model.

[0027] The standard deviation of the residuals is processed as follows:

[0028]

[0029] Wherein, N is the number of samples in the training set; μ is the mean of the residual; and S is the standard deviation of the residual.

[0030] Furthermore, the upper and lower bounds of the confidence interval are calculated using the formula:

[0031] U α (x i )=Y0(x i )+nor(1-α / 2)·S

[0032] L α (x i )=Y0(x i )-nor(1-α / 2)·S

[0033] Among them, U α is the upper bound of the confidence interval of the probability prediction; L α is the lower bound of the confidence interval for the probability prediction; x i is the i-th sample of the input prediction model; nor is the quantile return value of the standard normal distribution; α is the selected confidence level.

[0034] Furthermore, the prediction of the future load of electric vehicles includes:

[0035] The relevant data of the selected features are used to input the prediction model to obtain the corresponding prediction value, and based on the upper and lower bounds of the constructed confidence interval, the probabilistic prediction result of electric vehicle load based on extreme learning machine is obtained.

[0036] The present invention provides a probabilistic short-term prediction method for electric vehicle load based on an extreme learning machine. First, factors affecting the daily charging amount of the electric vehicle load are quantified and screened. Then, an extreme learning machine prediction model is constructed based on the number of input features and verification results of neurons in different hidden layers of the extreme learning machine. The upper and lower bounds of the confidence interval are calculated according to the residual standard deviation of the predicted value and the true value of the model in a training set and the selected confidence level. The future short-term load change of the electric vehicle is probabilistically predicted, which provides a basis for orderly regulation of the electric vehicle load and is conducive to ensuring the operation and maintenance management on the power grid side. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of a method for probabilistic short-term prediction of electric vehicle load based on an extreme learning machine provided in an embodiment of the present invention;

[0038] Figure 2 is a schematic diagram of verification results of neurons in different hidden layers provided by an embodiment of the present invention;

[0039] Figure 3 is a schematic diagram comparing a probability prediction curve of an electric vehicle cluster load and an actual electric vehicle load curve provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0041] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0042] In order to solve the problem that some electric energy meters are disconnected in residential areas and the load of electric vehicle cluster cannot be obtained synchronously in real time, an embodiment of the present invention provides an orderly charging control method and device for electric vehicles.

[0043] See also Figure 1 An embodiment of the present invention provides a probabilistic short-term forecasting method for electric vehicle load based on an extreme learning machine, comprising:

[0044] S1. Screen the factors that affect the daily charging capacity of electric vehicle load, quantify and normalize the required data, and use the Pearson correlation coefficient to screen the relevant factors;

[0045] S2. Based on the number of input features and the verification results of different hidden layer neurons of the extreme learning machine, an extreme learning machine prediction model is constructed;

[0046] S3, using the normalized historical data of the selected relevant factors to train the extreme learning machine prediction model, and calculating the upper and lower bounds of the confidence interval based on the residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level;

[0047] S4. Use the data of the selected relevant factors to input the extreme learning machine prediction model, predict the future load of electric vehicles, and obtain the probabilistic prediction result of the electric vehicle load according to the upper and lower bounds of the confidence interval.

[0048] S4. Use the electric vehicle load power collection values ​​and user behavior classification data from some connected electric energy meters to input the electric vehicle load real-time estimation model to achieve real-time estimation of the electric vehicle cluster load under all electric energy meters.

[0049] In one embodiment, the Pearson correlation coefficient is used to screen the relevant factors, including:

[0050] The temperature, weather and day type data collected at the same time as the historical load data of electric vehicles are extracted and quantified into quantitative values ​​of daily average temperature, weather and day type. The load of electric vehicles at the same moment a few days ago is selected, and the Pearson correlation coefficient is used as the screening basis to extract characteristic data that is correlated with the electric vehicle load as the model input.

[0051] Select n variables x1, x2...xn, time period t0~t, data sampling period T, time lag range ΔT, time shift step Δt;

[0052] Draw the time series curve of each variable, including separate drawing and centralized drawing modes, which can be switched by button;

[0053] Calculate the correlation coefficients between each variable and their corresponding time lags to form the correlation coefficient matrix Rp and the time lag matrix RΔt;

[0054] Based on data analysis and visualization tools, a correlation matrix diagram is drawn, which can be linked with the variable curve to obtain the most relevant factors affecting the daily charging amount of electric vehicle load.

[0055] The calculation process of the correlation coefficient between each variable is as follows: for two vectors X = (x1, ..., xn) and Y = (y1, ..., yn);

[0056]

[0057] Where n is the vector dimension, and are the means of vectors X and Y respectively.

[0058] The time lag is obtained by:

[0059] For every vector of two variables, move Δt each time within the time lag range ΔT and calculate the Pearson correlation coefficient between the corresponding two vectors. A total of ΔT / Δt moves are made. The one with the largest absolute value is selected as the correlation coefficient considering the time lag factor, and the moving time is taken as the time lag.

[0060] In this embodiment, one month of historical load data of electric vehicles in a residential area of ​​a certain city and the temperature, weather and day type at the corresponding time are used, and the time interval is 15 minutes. The Pearson analysis results are shown in Table 1. The daily average values ​​of temperature and day type are selected as factors affecting the load of electric vehicles.

[0061] Table 1 Analysis of Pearson correlation coefficients between daily charging capacity and influencing factors

[0062] Influencing factors temperature weather Day Type Evaluation results -0.2743 0.0103 0.3777

[0063] In this embodiment, the daily temperature and daily type average value and the electric vehicle load at the same time of the previous day, the previous time of the previous day, and the next time of the previous day are used as neural network inputs, and the data are divided into a training set and a test set at a ratio of 9:1.

[0064] In one embodiment, different hidden layer neuron verification results include:

[0065] Based on the influence of the number of hidden layer nodes on the prediction accuracy of the model, a range traversal is performed on the number of hidden layer nodes. When the number of hidden layer nodes exceeds a certain threshold, the extreme learning machine prediction model will have stable generalization performance. The number of hidden layer nodes slightly larger than the threshold is selected to ensure the generalization ability of the model and the low complexity of the model.

[0066] In this embodiment, the verification results of different hidden layer neurons are as follows: Figure 2 As shown in the figure, when the number of hidden layer nodes exceeds 40, the extreme learning machine prediction model has stable generalization performance. In this real-time example, the number of hidden layers is selected as 40.

[0067] In one embodiment, based on the residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level, it includes:

[0068] The extreme learning machine prediction model obtained by the training is used, and the training set data is used as the input of the prediction model to obtain the prediction result of the training set data of the prediction model, and the residual between the training set target data and the training set target data is obtained, and the standard deviation is calculated. The residual formula is as follows:

[0069] R d =Y1-Y0

[0070] Where Y1 is the real recorded data in the training set; Y0 is the predicted value obtained by the prediction model based on the training set input; R d is the residual between the real recorded data in the training set and the output data predicted by the model.

[0071] The standard deviation of the residuals is processed as follows:

[0072]

[0073] Wherein, N is the number of samples in the training set; μ is the mean of the residual; and S is the standard deviation of the residual.

[0074] In one embodiment, the upper and lower bounds of the confidence interval are calculated using the formula:

[0075] U α (x i )=Y0(x i )+nor(1-α / 2)·S

[0076] L α (x i )=Y0(x i )-nor(1-α / 2)·S

[0077] Among them, U α is the upper bound of the confidence interval of the probability prediction; L α is the lower bound of the confidence interval for the probability prediction; x i is the i-th sample of the input prediction model; nor is the quantile return value of the standard normal distribution; α is the selected confidence level.

[0078] In one embodiment, predicting the future load of electric vehicles includes:

[0079] The relevant data of the selected features are used to input the prediction model to obtain the corresponding prediction value, and based on the upper and lower bounds of the constructed confidence interval, the probabilistic prediction results of electric vehicle load based on extreme learning machine are obtained. The results are as follows: Figure 3 The corresponding model evaluation indicators are shown in Table 2, which demonstrates the good performance of the model and accurately predicts the changes in short-term electric vehicle load.

[0080] Table 2 Evaluation table of probabilistic short-term forecasting methods for electric vehicle load based on extreme learning machine

[0081]

[0082] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned probabilistic short-term prediction method for electric vehicle load based on extreme learning machine.

[0083] The above are preferred embodiments of the present invention. It should be noted that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A probabilistic short-term forecasting method for electric vehicle load based on extreme learning machine, characterized in that: The steps include: S1. Screen the factors that affect the daily charging capacity of electric vehicle load, quantify and normalize the required data, and use the Pearson correlation coefficient to screen the relevant factors; S2. Based on the number of input features and the verification results of different hidden layer neurons of the extreme learning machine, an extreme learning machine prediction model is constructed; S3, using the normalized historical data of the selected relevant factors to train the extreme learning machine prediction model, and calculating the upper and lower bounds of the confidence interval based on the residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level; S4. Use the data of the selected relevant factors to input the extreme learning machine prediction model, predict the future load of electric vehicles, and obtain the probabilistic prediction result of the electric vehicle load according to the upper and lower bounds of the confidence interval.

2. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 1, characterized in that: The Pearson correlation coefficient is used to screen the relevant factors, including: The temperature, weather and day type data collected at the same time as the historical load data of electric vehicles are extracted and quantified into quantitative values ​​of daily average temperature, weather and day type. The load of electric vehicles at the same moment a few days ago is selected, and the Pearson correlation coefficient is used as the screening basis to extract characteristic data that is correlated with the electric vehicle load as the model input.

3. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 2, characterized in that: Select n variables x1, x2...xn, time period t0~t, data sampling period T, time lag range ΔT, time shift step Δt; Draw the time series curve of each variable, including separate drawing and centralized drawing modes, which can be switched by button; Calculate the correlation coefficients between each variable and their corresponding time lags to form the correlation coefficient matrix Rp and the time lag matrix RΔt; Based on data analysis and visualization tools, a correlation matrix diagram is drawn, which can be linked with the variable curve to obtain the most relevant factors affecting the daily charging amount of electric vehicle load.

4. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine as claimed in claim 3, characterized in that: The calculation process of the correlation coefficient between each variable is: For two vectors X = (x1, ..., xn) and Y = (y1, ..., yn); Where n is the vector dimension, and are the means of vectors X and Y respectively.

5. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 4, characterized in that: The time lag is obtained by: For every vector of two variables, move Δt each time within the time lag range ΔT and calculate the Pearson correlation coefficient between the corresponding two vectors. A total of ΔT / Δt moves are made. The one with the largest absolute value is selected as the correlation coefficient considering the time lag factor, and the moving time is taken as the time lag.

6. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 1, characterized in that: Verification results of different hidden layer neurons, including: Based on the influence of the number of hidden layer nodes on the prediction accuracy of the model, a range traversal is performed on the number of hidden layer nodes. When the number of hidden layer nodes exceeds a certain threshold, the extreme learning machine prediction model will have stable generalization performance. The number of hidden layer nodes slightly larger than the threshold is selected to ensure the generalization ability of the model and the low complexity of the model.

7. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 1, characterized in that: The residual standard deviation of the predicted value and the true value of the model training set and the selected confidence level include: The extreme learning machine prediction model obtained by the training is used, and the training set data is used as the input of the prediction model to obtain the prediction result of the training set data of the prediction model, and the residual between the training set target data and the training set target data is obtained, and the standard deviation is calculated. The residual formula is as follows: R d =Y1-Y0 Where Y1 is the real recorded data in the training set; Y0 is the predicted value obtained by the prediction model based on the training set input; R d is the residual between the real recorded data in the training set and the output data predicted by the model. The standard deviation of the residuals is processed as follows: Wherein, N is the number of samples in the training set; μ is the mean of the residual; and S is the standard deviation of the residual.

8. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 7, characterized in that: The upper and lower bounds of the confidence interval are calculated as follows: U α (x i )=Y0(x i )+nor(1-α / 2)·S L α (x i )=Y0(x i )-nor(1-α / 2)·S Among them, U α is the upper bound of the confidence interval of the probability prediction; L α is the lower bound of the confidence interval for the probability prediction; x i is the i-th sample of the input prediction model; nor is the quantile return value of the standard normal distribution; α is the selected confidence level.

9. The electric vehicle load probabilistic short-term forecasting method based on extreme learning machine according to claim 1, characterized in that: The prediction of future load of electric vehicles includes: The relevant data of the selected features are used to input the prediction model to obtain the corresponding prediction value, and based on the upper and lower bounds of the constructed confidence interval, the probabilistic prediction result of electric vehicle load based on extreme learning machine is obtained.

10. A computer-readable storage medium, characterized in that: The storage medium includes at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the probabilistic short-term prediction method for electric vehicle load based on the extreme learning machine as described in any one of claims 1-6.

Images