Electric vehicle charging load prediction method and device, electronic equipment and storage medium

Abstract

The invention belongs to the technical field of load prediction, and provides an electric vehicle charging load prediction method and device, electronic equipment and a storage medium, and the method comprises the steps: collecting historical load data, historical meteorological data and historical date data of each electric vehicle charging station; generating time characteristic data and meteorological characteristic data according to the historical load data, the historical date data and the historical meteorological data; determining regional feature data based on the position of each electric vehicle charging station; based on historical load data, variational mode decomposition is utilized to obtain component data of each first IMF sequence; combining the time feature data, the regional feature data, the meteorological feature data and the first IMF sequence component data to generate second IMF component data; inputting each second IMF component data into an LSTM-Transform model to obtain a prediction result of each IMF component; and summation is carried out on the IMF component prediction results to obtain an electric vehicle charging load prediction result. According to the invention, the charging load prediction precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of load forecasting, and in particular, relates to a method, device, electronic device and storage medium for forecasting charging load of an electric vehicle. Background Art

[0002] With the rapid growth of the global electric vehicle market, the impact of electric vehicle charging load on the power grid is becoming increasingly significant. The charging demand of electric vehicles not only directly increases the load on the power grid, but also makes the load curve of the power grid more complex and dynamic. Especially during peak charging hours, a large number of charging behaviors may cause excessive pressure on the power grid and even cause power outages in local areas. Therefore, it is necessary to predict the charging load of electric vehicles to ensure the stable operation of the power grid and optimize the grid dispatch.

[0003] Traditional electric vehicle charging load forecasting methods include forecasting using time series models and forecasting based on statistical regression methods. Time series models include the Autoregressive Integrated Moving Average Model (ARIMA) and the Seasonal Autoregressive Integrated Moving Average (SARIMA). The above-mentioned electric vehicle charging load forecasting methods mainly rely on historical data and single-dimensional features. When dealing with complex nonlinear relationships and high-dimensional, multimodal data, they cannot effectively capture the dynamic characteristics of electric vehicle charging behavior over time and environmental changes, and it is difficult to integrate various factors that affect the charging load, resulting in large forecasting errors in extreme scenarios such as peak load and spike load. Therefore, how to improve the accuracy of electric vehicle charging load forecasting is an urgent problem to be solved in this field. Summary of the invention

[0004] To solve the above problems, the present disclosure provides a method, device, electronic device and storage medium for predicting charging load of an electric vehicle, the purpose of which is to improve the accuracy of prediction of charging load of an electric vehicle.

[0005] In order to achieve the above objectives, the present disclosure mainly provides the following technical solutions: In a first aspect, the present disclosure provides a method for predicting charging load of an electric vehicle, comprising: Collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station; generating time characteristic data according to the historical load data and the historical date data; generating meteorological characteristic data according to the historical load data and the historical meteorological data; Determine regional characteristic data based on the location of each electric vehicle charging station; generating a charging load time series according to the historical load data; Decomposing the charging load time series by variational mode decomposition to obtain a plurality of original IMF series component data; Processing the plurality of original IMF sequence component data using a dynamic sliding window to obtain first IMF sequence component data; Combining the time characteristic data, the regional characteristic data, the meteorological characteristic data and each of the first IMF series component data to generate each of the second IMF component data; Input the data of each second IMF component into the LSTM-Transformer model to obtain the prediction results of each IMF component; The prediction results of each IMF component are summed to obtain the prediction result of electric vehicle charging load.

[0006] In a second aspect, the present disclosure provides a device for predicting charging load of an electric vehicle, the device comprising: A collection unit, used to collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station; A first generating unit, configured to generate time characteristic data according to the historical load data and the historical date data; A second generating unit, configured to generate meteorological characteristic data according to the historical load data and the historical meteorological data; A determination unit, configured to determine the regional characteristic data based on the location of each electric vehicle charging station; A third generating unit, configured to generate a charging load time series according to the historical load data; A decomposition unit, used for decomposing the charging load time series by variational mode decomposition to obtain a plurality of original IMF series component data; A sliding unit, used for processing the plurality of original IMF sequence component data by using a dynamic sliding window to obtain each first IMF sequence component data; a combining unit, configured to combine the time characteristic data, the regional characteristic data, the meteorological characteristic data and each of the first IMF series component data to generate each of the second IMF component data; An input unit, used for inputting each second IMF component data into the LSTM-Transformer model to obtain a prediction result of each IMF component; The summing unit is used to sum the prediction results of each IMF component to obtain the prediction result of the electric vehicle charging load.

[0007] On the other hand, the present disclosure further provides a storage medium, which is used to store a computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the method of the first aspect above.

[0008] On the other hand, the present disclosure also provides an electronic device, comprising at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; and the processor is used to call program instructions in the memory to execute the method of the first aspect as described above.

[0009] Compared with the prior art, the present invention has the following advantages: The present disclosure not only collects historical load data of each electric vehicle charging station, but also collects or obtains historical meteorological data related to the charging load, historical date data and other related feature data, obtains each first IMF series component data after variational mode decomposition of the historical load data, and obtains the second IMF component data by combining each first IMF series component data and each related feature data, and uses the second IMF component data as the input data of the LSTM-Transformer model to obtain the corresponding prediction results of each IMF component. Since the second IMF component data involves data related to the charging load, such as temperature, weather, date, etc., the present disclosure effectively captures the dynamic characteristics of electric vehicle charging behavior changing with time and environment through high-dimensional and multi-modal data, and the present disclosure improves the accuracy of electric vehicle charging load prediction.

[0010] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or be understood by implementing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in 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 described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 A schematic diagram of a flow chart of a method for predicting charging load of an electric vehicle according to an embodiment of the present disclosure is shown; Figure 2 A schematic diagram of a process of another electric vehicle charging load prediction method according to an embodiment of the present disclosure is shown; Figure 3A schematic diagram of a process for generating combined feature vector data according to an embodiment of the present disclosure is shown; Figure 4 A schematic diagram of a flow chart of another electric vehicle charging load prediction method according to an embodiment of the present disclosure is shown; Figure 5 The LSTM-Transformer model architecture diagram according to an embodiment of the present disclosure is shown; Figure 6 A schematic diagram of the structure of a device for predicting charging load of an electric vehicle according to an embodiment of the present disclosure is shown; Figure 7 A schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0014] Figure 1 A schematic diagram of a method for predicting charging load of an electric vehicle according to an embodiment of the present disclosure is shown. Figure 1 As shown, the electric vehicle charging load prediction method of the embodiment of the present disclosure includes: 101. Collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station.

[0015] Among them, historical load data is used to record the charging load of each electric vehicle charging station at different time points in the past. Historical meteorological data may include historical weather conditions, temperature, humidity and other data. Weather conditions may include sunny days, rainy days, snowy days, cloudy days, etc. Temperature may include actual temperature and perceived temperature. Historical date data may include date types, such as weekdays, weekends, and holidays.

[0016] In this step, the collected and acquired data can be processed using the representation of word vectors in natural language processing. For example, historical date data such as weekdays, weekends, holidays, etc. are regarded as words in natural language and converted into vector form through word vectors, so that these non-numerical data can be integrated into subsequent model calculations.

[0017] 102. Generate time characteristic data based on historical load data and historical date data.

[0018] Among them, the time characteristic data is the time data related to the charging load of the electric vehicle.

[0019] The present disclosure can collect statistics on historical load data on a daily or hourly basis, and generate all time characteristic data in combination with the date type in the historical date data.

[0020] 103. Generate meteorological characteristic data based on historical load data and historical meteorological data.

[0021] Among them, the meteorological characteristic data are meteorological data related to the charging load of electric vehicles.

[0022] In this step, various historical load data corresponding to different historical meteorological data may be statistically analyzed to determine the different effects of various meteorological factors on the charging load, thereby obtaining meteorological characteristic data.

[0023] 104. Determine regional characteristic data based on the location of each electric vehicle charging station.

[0024] Among them, the regional characteristic data is the regional data related to the charging load of electric vehicles.

[0025] In this step, the areas where the electric vehicle charging stations are located can be divided, each area can be coded, and the regional characteristic data can be determined.

[0026] 105. Generate charging load time series based on historical load data.

[0027] In the embodiment of the present disclosure, the pre-processed historical load data can be arranged in chronological order with time as the horizontal axis and the charging load value as the vertical axis to construct a charging load time series. During the construction process, a unique charging load value can be corresponding to each time point.

[0028] 106. The charging load time series is decomposed through variational mode decomposition to obtain multiple original IMF series component data.

[0029] After obtaining the charging load time series, the embodiment of the present disclosure performs variational mode decomposition (VMD) on the charging load time series to obtain multiple original intrinsic mode function (IMF) sequence component data. Before the variational mode decomposition, the variational mode decomposition hyperparameters can be globally optimized to optimize the decomposition hyperparameters, thereby improving the charging load prediction accuracy.

[0030] 107. Process the plurality of original IMF sequence component data using a dynamic sliding window to obtain first IMF sequence component data.

[0031] The first IMF sequence component data is part of the original IMF sequence component data extracted through the sliding window. This part of the data is used to represent the local dynamic data of the charging load data. The embodiment of the present disclosure uses this part of the data and the corresponding related feature data to obtain the LSTM-Transformer model input data.

[0032] 108. Combine the time characteristic data, the regional characteristic data, the meteorological characteristic data, and each first IMF series component data to generate each second IMF component data.

[0033] Among them, the second IMF component data is the input data of the LSTM-Transformer model.

[0034] In this step, each first IMF series component data can be combined with the corresponding time characteristic data, regional characteristic data, and meteorological characteristic data to obtain each second IMF component data. Alternatively, the first IMF series component data can be subjected to feature data extraction using a sliding window, and after extracting feature data such as mean value and maximum value, the data can be combined with the corresponding time characteristic data, regional characteristic data, and meteorological characteristic data to obtain the second IMF component data.

[0035] 109. Input the data of each second IMF component into the LSTM-Transformer model to obtain the prediction results of each IMF component.

[0036] In this step, the second IMF component data can be divided into training set data, test set data, and validation set data. After the LSTM-Transformer model is constructed using the training set data, the prediction results of each IMF component can be obtained based on the LSTM-Transformer model through the test set data and the validation set data.

[0037] 110. Sum the forecast results of each IMF component to obtain the forecast result of electric vehicle charging load.

[0038] The disclosed embodiment not only collects the historical load data of each electric vehicle charging station, but also collects or obtains the historical meteorological data related to the charging load, the historical date data and other related characteristic data, obtains the first IMF series component data after the variational mode decomposition of the historical load data, and obtains the second IMF component data by combining the first IMF series component data and the related characteristic data, and uses the second IMF component data as the input data of the LSTM-Transformer model to obtain the corresponding prediction results of each IMF component. Since the second IMF component data involves data related to the charging load such as temperature, weather, date, etc., the disclosed embodiment effectively captures the dynamic characteristics of the electric vehicle charging behavior changing with time and environment through high-dimensional and multi-modal data, and the disclosed embodiment improves the accuracy of the electric vehicle charging load prediction.

[0039] In order to explain in more detail the electric vehicle charging load prediction method proposed in the present disclosure, the present disclosure proposes another embodiment of the electric vehicle charging load prediction method. The specific implementation steps of the embodiment of the present disclosure are as follows: Figure 2 As shown, including: 201. Collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station, and determine the area where each electric vehicle charging station is located.

[0040] like Figure 3 As shown, after the embodiment of the present disclosure collects or obtains all data such as historical load data, historical meteorological data, historical date data, etc., the data will be cleaned and preprocessed. Based on the preprocessed data, each relevant characteristic data related to the charging load is determined respectively, and the relevant characteristic data includes time characteristic data, regional characteristic data, and meteorological characteristic data. The time characteristic data may include a daily cycle characteristic vector, the regional characteristic data may include a code, that is, a target area identifier, and the meteorological characteristic data may include a meteorological characteristic weight. In addition, the historical load data is subjected to variational mode decomposition to obtain each first IMF sequence component data. Feature vector combination is performed based on the first IMF sequence component data, the time characteristic data, the regional characteristic data, and the meteorological characteristic data.

[0041] 202. Generate time characteristic data, meteorological characteristic data, and regional characteristic data based on historical load data, historical meteorological data, historical date data, and the region.

[0042] In one achievable method, the specific steps of generating time characteristic data according to historical load data and historical date data are as follows: Step 1: Divide the historical load data into daily periods to obtain daily historical load data.

[0043] Step 2: Count the daily historical load data to obtain the daily cycle feature vector.

[0044] In this step, the daily historical load data may be counted in order of time in units of hours to obtain a daily cycle feature vector.

[0045] The daily period feature vector generation method is as follows: Step 2a: Using the daily historical load data, count the average charging load, maximum charging load, minimum charging load and charging times for each hour, and calculate the proportion of the total charging load for each hour to the total daily charging load.

[0046] Step 2b: Generate a daily charging load distribution curve based on the charging load average value, charging load maximum value, charging load minimum value, charging times, and ratio.

[0047] The embodiment of the present disclosure can determine the peak period, valley period and relatively stable period of the charging load in a day based on the daily charging load distribution curve.

[0048] Step 2c: Based on the daily historical load data, calculate the rate of change of the charging load between adjacent hours to obtain supplementary characteristic data.

[0049] In the embodiment of the present disclosure, the rate of change of the charging load between adjacent hours = current hour load - previous hour load) / previous hour load, and the rate of change is used to reflect the dynamic change trend of the charging load within a day.

[0050] Step 2d: Generate a daily cycle feature vector based on the daily charging load distribution curve and the supplementary feature data.

[0051] The daily period feature vector is time feature data described with a day as the period.

[0052] In this step, the calculated rate of change of charging load between adjacent hours is used as a supplementary feature and integrated with the features contained in the daily charging load distribution curve (such as load statistics and proportion information of each time period, as well as peak, valley and stable time period information, etc.) to obtain a daily cycle feature vector. The daily cycle feature vector can reflect the static distribution of charging load within a day (daily load distribution curve characteristics) and dynamic change trend (charging load change rate characteristics).

[0053] Step 3: Generate time feature data using daily cycle feature vectors and historical date data.

[0054] The disclosed embodiment may associate the daily cycle feature vector with the date type in the corresponding historical date data. The date type may be encoded, such as using the number 1 to represent a working day, 2 to represent a weekend, and 3 to represent a holiday. The encoded date type is combined with the daily cycle feature vector to form time feature data including a time pattern (date type) and a load change feature (daily cycle feature vector).

[0055] In one achievable method, the specific steps of generating meteorological characteristic data based on historical load data and historical meteorological data are as follows: Step 1: Based on historical load data and historical meteorological data, determine the correlation between various meteorological factors and charging load through data analysis.

[0056] Among them, meteorological factors include at least temperature, humidity, and weather conditions, and weather conditions include sunny days, rainy days, snowy days, and cloudy days.

[0057] In this step, the collected historical meteorological data (such as temperature, humidity, weather conditions, etc.) and the corresponding charging load data are used to determine the correlation between each meteorological factor and the charging load through correlation analysis.

[0058] Step 2: Based on historical load data, determine the impact of different weather conditions on charging load.

[0059] In this step, the historical charging load data when different weather conditions (sunny, rainy, snowy, cloudy, etc.) occur can be counted separately, including the average value, maximum value, minimum value, and range of change of the charging load. For example, the average value of the charging load of all charging stations on sunny days and the difference with the average value under other weather conditions can be counted. Compare the statistical data of the charging load under different weather conditions to determine the laws and differences of the charging load affected by weather conditions. By calculating the change amplitude and frequency of the charging load under different weather conditions, the degree of influence of each weather condition on the charging load can be determined.

[0060] Step 3: Assign weights to each meteorological factor based on relevance and impact to obtain meteorological characteristic data.

[0061] In this step, based on the correlation between each meteorological factor and the charging load, and the degree of influence of different weather conditions on the charging load, each meteorological factor is assigned a weight, and each meteorological factor and the corresponding weight can be combined to obtain meteorological characteristic data. The embodiment of the present disclosure can also screen the meteorological factors according to their importance during the weighting process, and reduce the weight of unimportant meteorological factors.

[0062] In one achievable manner, the specific steps of determining the regional characteristic data based on the location of each electric vehicle charging station are as follows: Step 1: Divide the area according to the location of each electric vehicle charging station and the surrounding environment to obtain each target area.

[0063] Step 2: assigning area codes to each target area and generating identifications for each target area.

[0064] Step 3: Identify each area identifier as regional characteristic data.

[0065] The disclosed embodiment can divide the area according to the geographical location and surrounding environment of the charging station, for example, into commercial areas, residential areas, industrial areas, transportation hub areas (such as airports, railway stations, bus stations), public service facilities areas (such as hospitals, schools), etc. A unique code is assigned to each area to obtain the identification of each target area.

[0066] The disclosed embodiment uses time characteristic data, meteorological characteristic data, and regional characteristic data to achieve high-dimensional load forecasting, such as Figure 4 As shown, the disclosed embodiment can use the historical load data, time characteristic data, meteorological characteristic data, and regional characteristic data of electric vehicle charging stations to construct an electric vehicle charging station load prediction data set, and perform variational mode decomposition on the prediction data set to obtain each IMF series data and its residual m The charging load forecasting model is finally obtained through the dynamic sliding time window and LSTM-Transformer model. The historical load data can also be subjected to variational mode decomposition to obtain the corresponding IMF series data. After extracting the feature data through the dynamic sliding time window, this part of the feature data is combined with the time feature data, meteorological feature data, and regional feature data and then input into the LSTM-Transformer model. The implementation method of the latter is described in detail below.

[0067] 203. Generate a charging load time series according to the historical load data, and decompose the charging load time series through variational mode decomposition to obtain multiple original IMF series component data.

[0068] In this step, the historical load data are arranged in chronological order to construct a charging load time series, and the charging load time series is subjected to variational mode decomposition to obtain multiple original IMF series component data.

[0069] The formula for variational mode decomposition is: (1) (2) In the formula, For the k Sub-mode, is the charging load time series, is the impulse function, t is the time variable, is the center frequency of the kth submode, and k is the number of decomposition levels.

[0070] The disclosed embodiment utilizes variational mode decomposition to decompose the charging load time series into multiple sub-modes, processes the noise information therein, and extracts the time variation trend and frequency information of the load in order to accurately predict the load.

[0071] The disclosed embodiment can also use a secondary penalty factor α and the Lagrange multiplier λ( t ), the minimization problem is transformed into an unconstrained optimization problem, and the augmented Lagrangian function formula is: (3) In the formula, For the k Sub-mode, is the charging load time series, is the impulse function, t is the time variable, is the center frequency of the kth submode, k is the number of decomposition layers, α is the quadratic penalty factor, λ( t ) is the Lagrange multiplier.

[0072] The alternating direction multiplier method is used for iterative update. The iteration is stopped after the preset accuracy is met to obtain the K original IMF series component data after the charging load time series is decomposed. The original IMF series component data expression is: (4) (5) (6) In the formula, , and They are , and λ( t ), ω is the center frequency, n is the number of iterations, α is the quadratic penalty factor.

[0073] The disclosed embodiment effectively reduces data noise and complexity through variational modal decomposition, extracts modal components of different frequencies, makes data easier to analyze and process, and improves the prediction model's ability to process complex features of original load data.

[0074] Since the decomposition level K and the quadratic penalty factor α have an important influence on the variational mode decomposition, appropriate K and α can be selected before performing variational mode decomposition on the charging load time series to further improve the accuracy of charging load prediction. The Firefly Algorithm (FA) can be introduced to perform global optimization on the K and α hyperparameters of the variational mode decomposition.

[0075] Using the Firefly Optimization Algorithm to Optimize the Secondary Penalty Factor α And the number of decomposition layers K are globally optimized, and the specific method is as follows: Step 1: Initialize the population, set the population size q, the maximum number of iterations Tmax, and the attraction parameter β , attraction attenuation coefficient γ and other parameters. Randomly initialize q fireflies and set the parameters of the fireflies to ( ).

[0076] Step 2: Calculate the brightness, that is, calculate the objective function value. For each firefly, use the current Perform variational mode decomposition and calculate the objective function value. The formula for the objective function value is: (7) In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the charging load time series, N is the length of the charging load time series, and are the weights for minimizing the reconstruction error and the independence of the modal components, Indicates k Sub-mode.

[0077] Step 3: Update the position for each pair of fireflies i , j The Euclidean distance is calculated by adjusting the position based on the brightness comparison and the attraction formula: (8) In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the decomposition level of the jth firefly, is the quadratic penalty factor for the j-th firefly.

[0078] The attraction between two fireflies is calculated based on the distance between them: (9) In the formula, is the attraction parameter, and γ is the attraction attenuation coefficient.

[0079] According to the current position of the individual and the attraction between individuals, the new position is calculated: (10) (11) In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the decomposition level of the jth firefly, is the quadratic penalty factor for the jth firefly, is the attraction between the i-th firefly and the j-th firefly, is the step size control parameter, is a random number between [0,1].

[0080] From the perspective of the firefly optimization algorithm, , Firefly individuals i exist t The updated position parameters at each moment. , , , are the current status of the firefly individual ( t -1 moment) position parameter.

[0081] Step 4: Output the optimal solution if the number of iterations reaches the maximum value , stop searching and output the current individual position information with the highest brightness, otherwise return to step 2 and use the brightest individual information obtained by formulas (10) and (11) at the end of the search as the optimal solution for the variational mode decomposition hyperparameters.

[0082] 204. Process the plurality of original IMF sequence component data using a dynamic sliding window to obtain first IMF sequence component data.

[0083] Step 1: Initialize parameters and initialize the k The window length of the original IMF sequence component data is , the sliding step length is s.

[0084] in, You can make the windows overlap to retain more information and set the standard deviation threshold and rate of change threshold A measure of the volatility of a series.

[0085] Step 2: Calculate window statistics and set thek The starting position of the window of the original IMF series component data is , the data in the window is , calculate the standard deviation of the data within the window: (12) In the formula, For the k The mean of the data in the window, For the k The window length of the original IMF sequence component data.

[0086] Calculate the data difference in adjacent windows, and obtain the change rate sequence based on the ratio of the data difference in each adjacent window to the data in the corresponding previous window: (13) (14) In the formula, is the difference between the data in the i+1th window and the data in the i-th window, is the rate of change sequence of the data within the window.

[0087] In order to avoid the divisor being 0, the disclosed embodiment can perform a slight offset processing on the data, that is, , is a very small positive number.

[0088] Step 3: Determine and update the window size according to the set standard deviation threshold and rate of change threshold , and get the comprehensive threshold: (15) In the formula, for The weight coefficient of for The weight coefficient of .

[0089] Define the composite indicator: (16) In the formula, for The weight coefficient of for The weight coefficient of is the standard deviation of the data in the kth window, is the rate of change of data in the kth window.

[0090] when < θ When the data fluctuation in the window is small, the window can be appropriately reduced. > θ The window can be appropriately enlarged.

[0091] Updated window formula: (17) Where w is the window length, α<1, β>1, both are window scaling factors.

[0092] In the whole processing process, the steps of calculation, judgment and update are repeated until all the original IMF series component data are traversed, so as to dynamically adjust the window size to adapt to the volatility characteristics of the IMF series in different periods. That is, the optimal time window is obtained based on the updated window, and the data in the optimal time window is determined as the first IMF series component data, so as to use the first IMF series component data to obtain the model input data.

[0093] 205. Combine the time characteristic data, the regional characteristic data, the meteorological characteristic data, and each first IMF series component data to generate each second IMF component data.

[0094] The embodiment of the present disclosure can extract the first IMF sequence component data from the dynamic sliding window, calculate the first IMF sequence component data, obtain the mean, maximum, standard deviation and other characteristic data of the first IMF sequence component data, and splice this part of the characteristic data with the temperature characteristic data, date characteristic data, and regional characteristic characteristic data to obtain the second IMF component data. Assume that the window feature vector obtained after extraction from the dynamic sliding window ( is the mean, is the maximum value, min is the minimum value, is the standard deviation, r is the rate of change sequence), after temperature normalization, , the date is coded as , after regional coding: , combined feature vector Then , F is the second IMF component data. The disclosed embodiment retains the dynamic local characteristics of the IMF sequence, and at the same time combines the actual meteorological physical quantity, time information and regional characteristics to achieve high-dimensional charging load prediction.

[0095] The specific method of generating each second IMF component data is as follows: Step 1: Determine the mean, maximum, and standard deviation of the first IMF sequence component data.

[0096] Step 2: Concatenate the mean, maximum, standard deviation, and corresponding time characteristic data, regional characteristic data, and meteorological characteristic data to obtain a combined feature vector.

[0097] Step 3: Determine the combined eigenvector as the second IMF component data.

[0098] 206. Input each second IMF component data into the LSTM-Transformer model to obtain the prediction results of each IMF component, and sum the prediction results of each IMF component to obtain the electric vehicle charging load prediction result.

[0099] The LSTM-Transformer model architecture diagram of the disclosed embodiment is as follows Figure 5 As shown in the figure, the input data is first normalized and mapped to a suitable range, and then enters the LSTM layer. LSTM can effectively process long-term dependencies in long sequence data and capture the temporal characteristics of the data. The encoder part adds the data output from the LSTM to the position code. Then the data enters the encoder structure composed of N identical modules, where N is a positive integer. Each module contains a multi-head attention mechanism, summing and normalization, a feedforward neural network, and a summing and normalization operation again. The multi-head attention mechanism allows the model to capture different information in different representation subspaces and pay attention to the association between different parts of the input sequence. The decoder structure of the decoder is similar to that of the encoder. The multi-head attention mechanism of the decoder will pay attention to the output of the encoder and its own intermediate results at the same time to generate a suitable output. The output of the output layer decoder is first transformed by the linear layer for dimension transformation, and then output by the Softmax layer to obtain the prediction results of each IMF component. The prediction results of the electric vehicle charging load are obtained by summing the prediction results of each IMF component.

[0100] The specific steps of inputting each second IMF component data into the LSTM-Transformer model and obtaining the prediction results of each IMF component are as follows: Step 1: Adjust LSTM and Transformer parameters according to the optimal time window.

[0101] The specific steps to adjust LSTM and Transformer parameters are as follows: Step 1a: Send the first IMF sequence component data to the embedding layer, and the embedding layer adaptively allocates the embedding dimension according to the data length to ensure that data of different lengths can be reasonably encoded.

[0102] Step 1b: Add an adaptive layer and adjust the LSTM initial hidden state dimension and the number of heads and layers of the Transformer according to the optimal time window length, so that the model architecture fits the characteristics of input data of different lengths.

[0103] Construct a mapping function to characterize the relationship between the key parameters of LSTM and Transformer and the optimal time window length: (18) (19) In the formula, , , , , is a hyperparameter, is the time window length, is the optimal time window length, is the hidden state dimension, are the number of heads and layers of Transformer respectively.

[0104] The above hyperparameters can be determined by using validation set data, testing different values, and observing the model performance on the validation set.

[0105] Step 1c: The LSTM and Transformer key parameters calculated in the adaptive layer according to the mapping rules are transmitted to the LSTM and Transformer layers respectively.

[0106] Step 1d: Introduce an improved attention mechanism. The scaling factor scales the attention score according to the optimal time window length to balance the calculation effects under windows of different lengths. The calculation formula is: (20) Where Q is the query vector, K is the key vector, and V is the value vector. is the optimal time window length.

[0107] Step 2: Divide the second IMF component data into training set data and test set data.

[0108] The embodiment of the present disclosure may also divide the second IMF component data into training set data, test set data and validation set data.

[0109] Step 3: Input the training set data into the LSTM layer for processing to obtain the time series characteristics of each IMF component.

[0110] Step 4: Use the time series features of each IMF component as the input of the Transformer model, model the global dependency of the time series features of each IMF component by improving the attention mechanism, and construct the target LSTM-Transformer model.

[0111] The embodiment of the present disclosure may use a loss function (such as mean square error) and an Adam optimizer to perform parameter optimization for model training, and may also use an early stopping method to prevent overfitting during the training process.

[0112] Step 5: Input the test set data into the target LSTM-Transformer model to obtain the prediction results of each IMF component.

[0113] The disclosed embodiment utilizes dynamic sliding window technology to automatically adjust the window size according to the characteristics of each modal component, thereby improving the accuracy and robustness of the LSTM-Transformer model in dealing with dependencies of different time scales. The LSTM-Transformer model combines the LSTM's ability to capture short-term dynamic characteristics with the Transformer's ability to model global dependencies, effectively solving the long-term dependency problem and significantly improving the prediction accuracy. Through experimental comparison with actual electric vehicle load data, the disclosed embodiment is superior to traditional LSTM, LSTM-Transformer, VMD-LSTM and other models in terms of evaluation indicators such as mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE), providing more accurate results for electric vehicle charging load prediction, which is helpful for the optimized scheduling of smart grids and the development of the electric vehicle charging industry.

[0114] Based on the above method, the embodiment of the present disclosure provides an electric vehicle charging load prediction device, which is used to improve the accuracy of charging load prediction. The embodiment of the device corresponds to the aforementioned method embodiment. For ease of reading, this embodiment will no longer repeat the details of the aforementioned method embodiment one by one, but it should be clear that the device in this embodiment can correspond to all the contents of the aforementioned method embodiment. Specifically, Figure 6 As shown, the device comprises: The collection unit 301 is used to collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station; A first generating unit 302, configured to generate time characteristic data according to the historical load data and the historical date data; A second generating unit 303, configured to generate meteorological characteristic data according to the historical load data and the historical meteorological data; A determination unit 304, configured to determine regional characteristic data based on the location of each electric vehicle charging station; A third generating unit 305 is used to generate a charging load time series according to the historical load data; A decomposition unit 306 is used to decompose the charging load time series by variational mode decomposition to obtain a plurality of original IMF series component data; A sliding unit 307 is used to process the plurality of original IMF sequence component data using a dynamic sliding window to obtain each first IMF sequence component data; A combining unit 308, configured to combine the time characteristic data, the regional characteristic data, the meteorological characteristic data and each of the first IMF series component data to generate each of the second IMF component data; An input unit 309 is used to input each second IMF component data into the LSTM-Transformer model to obtain a prediction result of each IMF component; The summing unit 310 is used to sum the prediction results of each IMF component to obtain the electric vehicle charging load prediction result.

[0115] Furthermore, the first generating unit is specifically used for: Dividing the historical load data based on a day as a cycle to obtain daily historical load data; Statistics are performed on daily historical load data to obtain daily cycle characteristic vectors; The daily cycle feature vector and the historical date data are used to generate time feature data.

[0116] Furthermore, the first generating unit is further specifically configured to: Using daily historical load data, count the average charging load, maximum charging load, minimum charging load and charging times for each hour, and calculate the proportion of the total charging load for each hour to the total daily charging load; Generate a daily charging load distribution curve according to the charging load average value, the charging load maximum value, the charging load minimum value, the charging times, and the ratio; Based on daily historical load data, the change rate of charging load between adjacent hours is calculated to obtain supplementary characteristic data; A daily cycle characteristic vector is generated according to the daily charging load distribution curve and the supplementary characteristic data.

[0117] Furthermore, the second generating unit is specifically used for: Based on the historical load data and the historical meteorological data, determining the correlation between each meteorological factor and the charging load through data analysis, wherein the meteorological factors include at least temperature, humidity, and weather conditions, and the weather conditions include at least sunny days, rainy days, snowy days, and cloudy days; Based on the historical load data, determining the degree of influence of different weather conditions on the charging load; According to the correlation and the degree of influence, the weight of each meteorological factor is allocated to obtain meteorological characteristic data.

[0118] Furthermore, the determining unit is specifically used for: According to the location of each electric vehicle charging station and the surrounding environment, the area is divided to obtain each target area; Allocating area codes to the target areas to generate identifications of the target areas; Each of the area identifiers is determined as regional feature data.

[0119] Furthermore, the decomposition unit is also specifically used for: The formula for variational mode decomposition is:

[0120]

[0121] In the formula, For the k Sub-mode, is the charging load time series, is the impulse function, t is the time variable, is the center frequency of the kth submode, k is the number of decomposition levels; Using quadratic penalty factors α and the Lagrange multiplier λ( t ), the minimization problem is transformed into an unconstrained optimization problem, and the augmented Lagrangian function formula is: In the formula, For the k Sub-mode, is the charging load time series, is the impulse function, t is the time variable, is the center frequency of the kth submode, k is the number of decomposition layers, α is the quadratic penalty factor, λ( t ) is the Lagrange multiplier; The alternating direction multiplier method is used for iterative update, and the iteration is stopped after the preset accuracy is met to obtain the decomposed K original IMF sequence component data, and the expression of the original IMF sequence component data is:

[0122]

[0123]

[0124] In the formula, , and They are , and λ( t ), ω is the center frequency, n is the number of iterations, α is the quadratic penalty factor.

[0125] Furthermore, the decomposition unit is also specifically used for: The firefly optimization algorithm is used to optimize the secondary penalty factor α and the decomposition layer number K to perform global optimization.

[0126] Furthermore, the decomposition unit is also specifically used for: Random Initialization q fireflies, and set the parameters of the fireflies to ( ); Calculate the objective function value, the formula of the objective function value is:

[0127] In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the charging load time series, N is the length of the charging load time series, and are the weights for minimizing the reconstruction error and the independence of the modal components, Indicates k sub-mode; For each pair of fireflies i , j The Euclidean distance is calculated by adjusting the position based on the brightness comparison and the attraction formula:

[0128] In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the decomposition level of the jth firefly, is the quadratic penalty factor of the j-th firefly; The attraction between two fireflies is calculated based on the distance between them:

[0129] In the formula, is the attraction parameter, γ is the attraction attenuation coefficient; According to the current position of the individual and the attraction between individuals, the new position is calculated:

[0130]

[0131] In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the decomposition level of the jth firefly, is the quadratic penalty factor for the jth firefly, is the attraction between the i-th firefly and the j-th firefly, is the step size control parameter, is a random number between [0,1].

[0132] Furthermore, the decomposition unit is also specifically used for: Initialize k The window length of the original IMF sequence component data is , the sliding step length is s; Set k The starting position of the window of the original IMF series component data is , the data in the window is , calculate the standard deviation of the data within the window:

[0133] In the formula, For the k The mean of the data in the window, For the k The window length of the original IMF sequence component data; Calculate the data difference in adjacent windows to get the change rate sequence:

[0134]

[0135] In the formula, is the difference between the data in the i+1th window and the data in the i-th window, is the change rate sequence of the data in the window; Set the standard deviation threshold to and rate of change threshold , and get the comprehensive threshold:

[0136] In the formula, for The weight coefficient of for The weight coefficient of Define the composite indicator:

[0137] In the formula, for The weight coefficient of for The weight coefficient of is the standard deviation of the data in the kth window, is the rate of change of data in the kth window; when < θ When the window is reduced, > θ When expanding the window; Updated window formula:

[0138] Where w is the window length, α<1, β>1, both are window scaling coefficients; An optimal time window is obtained based on the updated window, and the data in the optimal time window is determined as the first IMF sequence component data.

[0139] Furthermore, the combination unit is specifically used for: Determine the mean, maximum and standard deviation of the first IMF sequence component data; splicing the mean, the maximum, the standard deviation, and the corresponding time characteristic data, the regional characteristic data, and the meteorological characteristic data to obtain a combined characteristic vector; The combined eigenvector is determined as second IMF component data.

[0140] Furthermore, the input unit is specifically used for: Adjust LSTM and Transformer parameters according to the optimal time window; The second IMF component data is divided into training set data and test set data; The training set data is input into the LSTM layer for processing to obtain the time series characteristics of each IMF component; Taking the time series features of each IMF component as the input of the Transformer model, modeling the global dependency relationship of the time series features of each IMF component by improving the attention mechanism, and constructing a target LSTM-Transformer model; The test set data is input into the target LSTM-Transformer model to obtain the prediction results of each IMF component.

[0141] Furthermore, the input unit is also specifically used for: The first IMF sequence component data is sent to an embedding layer, and the embedding layer adaptively allocates embedding dimensions according to the data length; Add an adaptive layer to adjust the LSTM initial hidden state dimension and the number of heads and layers of the Transformer according to the optimal time window length; Construct a mapping function to characterize the relationship between the key parameters of LSTM and Transformer and the optimal time window length:

[0142]

[0143] In the formula, , , , , is a hyperparameter, is the optimal time window length, is the hidden state dimension, are the number of heads and layers of Transformer respectively; The LSTM and Transformer key parameters calculated in the adaptive layer according to the mapping rules are transmitted to the LSTM and Transformer layers respectively; An improved attention mechanism is introduced. The scaling factor scales the attention score according to the optimal time window length to balance the calculation effects under windows of different lengths. The calculation formula is:

[0144] Where Q is the query vector, K is the key vector, and V is the value vector. is the optimal time window length.

[0145] Furthermore, the embodiment of the present disclosure also provides a processor, the processor is used to run a program, wherein the program executes the above Figure 1-2 The method described in .

[0146] Furthermore, the embodiment of the present disclosure also provides a storage medium, wherein the storage medium is used to store a computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the above Figure 1-2 The method described in .

[0147] Furthermore, the present disclosure provides an electronic device 4, such as Figure 7 As shown, the device includes at least one processor 41, at least one memory 42 and a bus 43 connected to the processor 41; the processor 41 and the memory 42 communicate with each other through the bus 43; the processor 41 is used to call the program instructions in the memory 42 to execute the above-mentioned electric vehicle charging load prediction method. The device in this article can be a server, PC, PAD, mobile phone, etc.

[0148] Furthermore, the present disclosure also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the steps of the inspection method for network devices as described above.

[0149] Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A method for predicting electric vehicle charging load, characterized in that: The method comprises: Collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station; generating time characteristic data according to the historical load data and the historical date data; generating meteorological characteristic data according to the historical load data and the historical meteorological data; Determine regional characteristic data based on the location of each electric vehicle charging station; generating a charging load time series according to the historical load data; Decomposing the charging load time series by variational mode decomposition to obtain a plurality of original IMF series component data; Processing the plurality of original IMF sequence component data using a dynamic sliding window to obtain first IMF sequence component data; Combining the time characteristic data, the regional characteristic data, the meteorological characteristic data and each of the first IMF series component data to generate each of the second IMF component data; Input the data of each second IMF component into the LSTM-Transformer model to obtain the prediction results of each IMF component; The prediction results of each IMF component are summed to obtain the prediction result of electric vehicle charging load.

2. The method according to claim 1, characterized in that Generating time characteristic data according to the historical load data and the historical date data includes: Dividing the historical load data based on a day as a cycle to obtain daily historical load data; Statistics are performed on daily historical load data to obtain daily cycle characteristic vectors; The daily cycle feature vector and the historical date data are used to generate time feature data.

3. The method according to claim 2, characterized in that The daily historical load data is statistically analyzed to obtain the daily cycle feature vector, including: Using daily historical load data, count the average charging load, maximum charging load, minimum charging load and charging times for each hour, and calculate the proportion of the total charging load for each hour to the total daily charging load; Generate a daily charging load distribution curve according to the charging load average value, the charging load maximum value, the charging load minimum value, the charging times, and the ratio; Based on daily historical load data, the change rate of charging load between adjacent hours is calculated to obtain supplementary characteristic data; A daily cycle characteristic vector is generated according to the daily charging load distribution curve and the supplementary characteristic data.

4. The method according to claim 1, characterized in that: Generate meteorological characteristic data according to the historical load data and the historical meteorological data, including: Based on the historical load data and the historical meteorological data, determining the correlation between each meteorological factor and the charging load through data analysis, wherein the meteorological factors include at least temperature, humidity, and weather conditions, and the weather conditions include at least sunny days, rainy days, snowy days, and cloudy days; Based on the historical load data, determining the degree of influence of different weather conditions on the charging load; According to the correlation and the degree of influence, the weight of each meteorological factor is allocated to obtain meteorological characteristic data.

5. The method according to claim 1, characterized in that: Determine regional characteristic data based on the location of each electric vehicle charging station, including: According to the location of each electric vehicle charging station and the surrounding environment, the area is divided to obtain each target area; Allocating area codes to the target areas to generate identifications of the target areas; Each of the area identifiers is determined as regional feature data.

6. The method according to claim 1, characterized in that The charging load time series is decomposed by variational mode decomposition to obtain multiple original IMF series component data, including: The formula for variational mode decomposition is: In the formula, For the k Sub-mode, is the charging load time series, is the impulse function, t is the time variable, is the center frequency of the kth submode, k is the number of decomposition levels; Using quadratic penalty factors α and the Lagrange multiplier λ( t ), the minimization problem is transformed into an unconstrained optimization problem, and the augmented Lagrangian function formula is: In the formula, For the k Sub-mode, is the charging load time series, is the impulse function, t is the time variable, is the center frequency of the kth submode, k is the number of decomposition layers, α is the quadratic penalty factor, λ( t ) is the Lagrange multiplier; The alternating direction multiplier method is used for iterative update, and the iteration is stopped after the preset accuracy is met to obtain the decomposed K original IMF sequence component data, and the expression of the original IMF sequence component data is: In the formula, , and They are , and λ( t ), ω is the center frequency, n is the number of iterations, α is the quadratic penalty factor.

7. The method according to claim 6, characterized in that The method further comprises: The firefly optimization algorithm is used to optimize the secondary penalty factor α and the decomposition level number K to perform global optimization.

8. The method according to claim 7, characterized in that The firefly optimization algorithm is used to optimize the secondary penalty factor α and the decomposition layer number K to perform global optimization, including: Random Initialization q fireflies, and set the parameters of the fireflies to ( ); Calculate the objective function value, the formula of the objective function value is: In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the charging load time series, N is the length of the charging load time series, and are the weights for minimizing the reconstruction error and the independence of the modal components, Indicates k sub-mode; For each pair of fireflies i , j The Euclidean distance is calculated by adjusting the position based on the brightness comparison and the attraction formula: In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the decomposition level of the jth firefly, is the quadratic penalty factor of the j-th firefly; The attraction between two fireflies is calculated based on the distance between them: In the formula, is the attraction parameter, γ is the attraction attenuation coefficient; According to the current position of the individual and the attraction between individuals, the new position is calculated: In the formula, is the decomposition level of the i-th firefly, is the quadratic penalty factor for the i-th firefly, is the decomposition level of the jth firefly, is the quadratic penalty factor for the jth firefly, is the attraction between the i-th firefly and the j-th firefly, is the step size control parameter, is a random number between [0,1].

9. The method according to claim 1, characterized in that: The plurality of original IMF sequence component data are processed by using a dynamic sliding window to obtain each first IMF sequence component data, including: Initialize k The window length of the original IMF sequence component data is , the sliding step length is s; Set k The starting position of the window of the original IMF series component data is , the data in the window is , calculate the standard deviation of the data within the window: In the formula, For the k The mean of the data in the window, For the k The window length of the original IMF sequence component data; Calculate the data difference in adjacent windows to get the change rate sequence: In the formula, is the difference between the data in the i+1th window and the data in the i-th window, is the change rate sequence of the data in the window; Set the standard deviation threshold to and rate of change threshold , and get the comprehensive threshold: In the formula, for The weight coefficient of for The weight coefficient of Define the composite indicator: In the formula, for The weight coefficient of for The weight coefficient of is the standard deviation of the data in the kth window, is the rate of change of data in the kth window; when < θ When the window is reduced, > θ When expanding the window; Updated window formula: Where w is the window length, α<1, β>1, both are window scaling coefficients; An optimal time window is obtained based on the updated window, and the data in the optimal time window is determined as the first IMF sequence component data.

10. The method according to claim 9, characterized in that Combining the time characteristic data, the regional characteristic data, the meteorological characteristic data, and each of the first IMF series component data to generate each of the second IMF component data, including: Determine the mean, maximum, and standard deviation of the first IMF sequence component data; splicing the mean, the maximum, the standard deviation, and the corresponding time characteristic data, the regional characteristic data, and the meteorological characteristic data to obtain a combined characteristic vector; The combined eigenvector is determined as second IMF component data.

11. The method according to claim 10, characterized in that Input the data of each second IMF component into the LSTM-Transformer model to obtain the prediction results of each IMF component, including: Adjust LSTM and Transformer parameters according to the optimal time window; The second IMF component data is divided into training set data and test set data; The training set data is input into the LSTM layer for processing to obtain the time series characteristics of each IMF component; Taking the time series features of each IMF component as the input of the Transformer model, modeling the global dependency relationship of the time series features of each IMF component by improving the attention mechanism, and constructing a target LSTM-Transformer model; The test set data is input into the target LSTM-Transformer model to obtain the prediction results of each IMF component.

12. The method according to claim 11, characterized in that Adjust LSTM and Transformer parameters according to the optimal time window, including: The first IMF sequence component data is sent to an embedding layer, and the embedding layer adaptively allocates embedding dimensions according to the data length; Add an adaptive layer to adjust the LSTM initial hidden state dimension and the number of heads and layers of the Transformer according to the optimal time window length; Construct a mapping function to characterize the relationship between the key parameters of LSTM and Transformer and the optimal time window length: In the formula, , , , , is a hyperparameter, is the optimal time window length, is the hidden state dimension, are the number of heads and layers of Transformer respectively; The LSTM and Transformer key parameters calculated in the adaptive layer according to the mapping rules are transmitted to the LSTM and Transformer layers respectively; An improved attention mechanism is introduced. The scaling factor scales the attention score according to the optimal time window length to balance the calculation effects under windows of different lengths. The calculation formula is: Where Q is the query vector, K is the key vector, and V is the value vector. is the optimal time window length.

13. An electric vehicle charging load prediction device, characterized in that: The device comprises: A collection unit, used to collect historical load data, historical meteorological data, and historical date data of each electric vehicle charging station; A first generating unit, configured to generate time characteristic data according to the historical load data and the historical date data; A second generating unit, configured to generate meteorological characteristic data according to the historical load data and the historical meteorological data; A determination unit, configured to determine the regional characteristic data based on the location of each electric vehicle charging station; A third generating unit, configured to generate a charging load time series according to the historical load data; A decomposition unit, used for decomposing the charging load time series by variational mode decomposition to obtain a plurality of original IMF series component data; A sliding unit, used for processing the plurality of original IMF sequence component data by using a dynamic sliding window to obtain each first IMF sequence component data; a combining unit, configured to combine the time characteristic data, the regional characteristic data, the meteorological characteristic data and each of the first IMF series component data to generate each of the second IMF component data; An input unit, used for inputting each second IMF component data into the LSTM-Transformer model to obtain the prediction results of each IMF component; The summing unit is used to sum the prediction results of each IMF component to obtain the prediction result of the electric vehicle charging load.

14. The device according to claim 13, characterized in that The first generating unit is specifically used for: Dividing the historical load data based on a day as a cycle to obtain daily historical load data; Perform statistics on daily historical load data to obtain daily cycle characteristic vectors; The daily cycle feature vector and the historical date data are used to generate time feature data.

15. An electronic device, characterized in that: The device includes at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; and the processor is used to call program instructions in the memory to execute the method described in any one of claims 1-12.

16. A computer storage medium, characterized in that: The storage medium is used to store a computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 12.

Images