Charging station load prediction method and device, electronic equipment and storage medium
By improving the charging load prediction model, combining the charging timing prediction model and the encoder decoder, the problem of low accuracy in charging station load prediction is solved, and higher prediction accuracy and feature capture capabilities are achieved.
Patent Information
- Application Number
- CN202510459796.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-14
- Publication Date
- 2025-07-25
AI Technical Summary
In the prior art, the Transformer model is difficult to capture rich semantic information in charging station load prediction, resulting in low prediction accuracy and difficult to distinguish complex relationships between variables.
The initial charging load prediction model is improved as a charging load inversion prediction model, and the charging timing prediction model is added, and the charging load prediction model is built. The timing characteristics of the data and the coupling relationship between multiple variables are captured through the charging load prediction model. Feature extraction and decoding are used for encoder and decoder.
It improves the accuracy of charging station load prediction, overcomes the limitations of the traditional Transformer model in dealing with multivariable coupling relationships and long-term compatibility, and improves the feature expression ability and spatiotemporal feature learning performance of the prediction model.
Smart Images
Figure CN120373546A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of charging device management. Specifically, it relates to a method, device, electronic device and storage medium for predicting the load of a charging station. Background Art
[0002] With the popularization of electric vehicles, large-scale construction of charging piles and charging stations, and a large number of high-proportion new energy sources connected to the power grid, it has brought challenges to the safety and stability of the power grid. How to achieve orderly charging and reasonable configuration of the power grid is a technical problem that needs to be solved urgently at present. To solve these problems, power companies have begun to implement peak shaving and valley filling programs and formulate demand response strategies, and the formulation and implementation of these strategies rely on load estimation in the power consumption area.
[0003] In the prior art, the load of a charging station is usually predicted based on the Transformer model. However, in the process of load prediction, the Transformer model maps all variables to feature words at the same time point, resulting in too small a receptive field of the model itself, making it difficult to capture rich semantic information and reducing the accuracy of charging station load prediction. Summary of the Invention
[0004] In view of this, the purpose of the present application is to provide a method, device, electronic device and storage medium for predicting the load of a charging station to solve the problem of low accuracy of charging station load prediction.
[0005] In a first aspect, an embodiment of the present application provides a method for predicting the load of a charging station, including:
[0006] Improve the initial charging load prediction model to obtain a charging load inversion prediction model to capture the coupling relationship between multiple charging data variables. The initial charging load prediction model is a converter model for predicting the load of a charging station;
[0007] Add a charging time series prediction model to the charging load inversion prediction model to construct a charging load prediction model. The charging time series prediction model is a long short-term memory model for capturing the time series characteristics of charging data;
[0008] Input multi-source charging data affecting the load of the charging station into the charging load prediction model to obtain the charging station load prediction result.
[0009] In an optional embodiment, the charging load inversion prediction model includes an encoder and a decoder. Adding a charging time series prediction model to the charging load inversion prediction model to construct a charging load prediction model includes: adding a charging time series prediction model between the encoder and the decoder to construct a charging load prediction model.
[0010] In an alternative embodiment, multi-source charging data affecting the charging station load is input into a charging load prediction model to obtain a charging station load prediction result, including: using an embedding layer to map the entire sequence of the same charging data variable into a high-dimensional feature vector; using an encoder to encode the high-dimensional feature vector to obtain an initial charging feature encoding; using a charging time series prediction model to extract time series features from the initial charging feature encoding to obtain a time series charging feature encoding; using a decoder to decode the time series charging feature encoding to obtain a charging station load prediction result.
[0011] In an alternative embodiment, the encoder includes a multi-head attention layer and a feed-forward neural network layer. Using the encoder to encode multi-source charging data to obtain an initial charging feature encoding includes: using the multi-head attention layer to establish a dependency relationship between different charging data variables in the multi-source charging data to obtain multi-source relationship data; using the feed-forward neural network layer to perform layer-by-layer encoding on the historical observation features of the multi-source relationship data in the time dimension to obtain an initial charging feature encoding.
[0012] In an alternative embodiment, the encoder further includes a first residual connection and normalization layer, and a second residual connection and normalization layer. The first residual connection and normalization layer is used to directly add the multi-source charging data to the multi-source relationship data and perform normalization processing to obtain normalized multi-source relationship data. The second residual connection and normalization layer is used to directly add the normalized multi-source relationship data to the initial charging feature encoding and perform normalization processing to obtain normalized initial charging feature encoding.
[0013] In an alternative embodiment, the method further includes: using a standardization method to perform normalization processing on the multi-source charging data to obtain multi-source charging standard data.
[0014] In an alternative embodiment, after performing normalization processing on the multi-source charging data to obtain multi-source charging standard data, it further includes: performing a correlation analysis on the multi-source charging standard data to obtain multi-source charging standard data with correlation between different charging data variables, so as to use the multi-source charging standard data with correlation between different charging data variables for prediction.
[0015] In a second aspect, an embodiment of the present application further provides a charging station load prediction device, where the device includes:
[0016] A model improvement module, configured to improve an initial charging load prediction model to obtain a charging load inversion prediction model to capture the coupling relationship between multiple charging data variables. The initial charging load prediction model is a transformer model for predicting the charging station load;
[0017] A model construction module for adding a charging time sequence prediction model to a charging load inversion prediction model to construct a charging load prediction model. The charging time sequence prediction model is a long short-term memory model for capturing the time sequence characteristics of charging data;
[0018] A load prediction model for inputting multi-source charging data affecting the charging station load into the charging load prediction model to obtain a charging station load prediction result.
[0019] In a third aspect, an embodiment of the present application further provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus. When the machine-readable instructions are executed by the processor, the steps of the charging station load prediction method as described above are executed.
[0020] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is run by a processor, the steps of the charging station load prediction method as described above are executed.
[0021] The embodiments of the present application bring the following beneficial effects:
[0022] A charging station load prediction method, device, electronic device, and storage medium provided by an embodiment of the present application can construct a charging load prediction model based on a charging time sequence prediction model and a charging load inversion prediction model. Through the charging load prediction model, the time sequence characteristics of data and the coupling relationship between multiple variables can be captured, overcoming the limitations of traditional Transformer models in dealing with multi-variable coupling relationships and long-time sequence compatibility. Compared with the charging station load prediction methods in the prior art, the problem of low accuracy of charging station load prediction is solved.
[0023] To make the above objects, features, and advantages of the present application more obvious and understandable, the following specific embodiments are given in conjunction with the accompanying drawings and are described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present application and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.
[0025] Figure 1 Shows the flowchart of the charging station load prediction method provided by the embodiment of the present application;
[0026] Figure 2Shows a schematic structural diagram of the charging load prediction model provided by the embodiments of the present application;
[0027] Figure 3 Shows a schematic structural diagram of the charging station load prediction device provided by the embodiments of the present application;
[0028] Figure 4 Shows a schematic structural diagram of the electronic device provided by the embodiments of the present application. Detailed implementation manners
[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Usually, the components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, every other embodiment obtained by those skilled in the art without creative efforts belongs to the scope of protection of the present application.
[0030] It should be noted that before the present application was proposed, with the popularization of electric vehicles, large-scale construction of charging piles and charging stations, and a large amount of high-proportion new energy connected to the power grid, it has brought challenges to the safety and stability of the power grid. How to achieve orderly charging and reasonable configuration of the power grid is a technical problem that needs to be solved urgently at present. To solve these problems, power companies have begun to implement peak shaving and valley filling programs and formulate demand response strategies, and the formulation and implementation of these strategies rely on the load estimation of the power consumption area. In the prior art, the charging station load prediction is usually based on the Transformer model. However, in the load prediction process, the Transformer model maps all variables to the feature words at the same time point, resulting in too small receptive field of the model itself and making it difficult to capture rich semantic information. At the same time, this processing method mixes the features between variables, making it difficult for the Transformer model to distinguish the complex relationships between variables and reducing the accuracy of the charging station load prediction.
[0031] Based on this, the embodiments of the present application provide a charging station load prediction method to improve the accuracy of the charging station load prediction.
[0032] The following explains the proprietary terms involved in the present application.
[0033] Temporal Token:
[0034] A word refers to the basic unit of text data input into the model, which can be a word, sub-word, or character. In natural language processing (NLP), words are the basic elements that make up a language, and the Transformer model understands and generates language by processing these words. Specifically, the Transformer model processes the input text through tokenization, breaking the text into a series of tokens, and each token is converted into a continuous dense vector representation, a process known as Word Embedding.
[0035] Please refer to Figure 1 , Figure 1 which is a flowchart of a charging station load prediction method provided by an embodiment of this application. As Figure 1 shown, the charging station load prediction method provided by an embodiment of this application includes:
[0036] Step S101: Improve the initial charging load prediction model to obtain a charging load inversion prediction model to capture the coupling relationship between multiple charging data variables;
[0037] Step S102: Add a charging time series prediction model to the charging load inversion prediction model to construct a charging load prediction model;
[0038] Step S103: Input multi-source charging data that affects the charging station load into the charging load prediction model to obtain the charging station load prediction result.
[0039] The charging station load prediction method provided by an embodiment of this application can construct a charging load prediction model based on the charging time series prediction model and the charging load inversion prediction model. Through the charging load prediction model, the time series characteristics of the data and the coupling relationship between multiple variables can be captured, overcoming the limitations of the traditional Transformer model in dealing with multi-variable coupling relationships and long-time series compatibility. Compared with the charging station load prediction methods in the prior art, the problem of low accuracy of charging station load prediction is solved.
[0040] To facilitate the understanding of this embodiment, the above exemplary steps provided by an embodiment of this application will be described separately below.
[0041] In step S101, the initial charging load prediction model is improved to obtain a charging load inversion prediction model to capture the coupling relationship between multiple charging data variables.
[0042] In this step, the initial charging load prediction model is a converter model for predicting the charging station load. The converter model can refer to the Transformer model, and the Transformer model is a deep neural network model based on the self-attention mechanism.
[0043] The charging load inversion prediction model is an inverted Transformer model, which can be denoted as the iTransformer (Inverted Transformer) model. The charging load inversion prediction model is used to capture the coupling relationship between multiple charging data variables.
[0044] Since when constructing words for time series data in the Transformer model, all variable values at the same time point are usually concatenated into a vector and this vector is regarded as a word (that is, each time stamp consists of multiple variable values at the same time), due to the difficulty in aligning the collection times of all variables and the overly small receptive field, and the feature embedding not considering the time offset, it is difficult to capture the correlation between multiple variables. Therefore, it is necessary to improve the initial charging load prediction model.
[0045] When improving the model, the word of the initial charging load prediction model can be changed from multiple charging data variable values collected at the same moment to the entire sequence of the same charging data variable. Here, the process of changing from concatenating all variable values at the same time point into a word to concatenating the entire sequence of the same variable into a word is called dimension inversion. Dimension inversion will be applied to each module of the initial charging load prediction model, so as to achieve model improvement and obtain the charging load inversion prediction model.
[0046] In this way, by applying dimension inversion to each module of the initial charging load prediction model, the words in the charging load inversion prediction model can take the variable as the description subject, and the entire sequence corresponding to each charging data variable can be independently embedded into the variable tag. Among them, each module of the initial charging load prediction model includes a multi-head attention module and a feed-forward neural network module, so that the multi-head attention module can depict the correlation between multiple charging data variables, and the feed-forward neural network module can encode the representation of the entire sequence corresponding to each charging data variable.
[0047] In step S102, a charging time series prediction model is added to the charging load inversion prediction model to construct a charging load prediction model.
[0048] In this step, the charging time series prediction model refers to the LSTM (Long Short-Term Memory Network) network model. The charging time series prediction model is a long short-term memory model used to capture the time series characteristics of charging data. Using the charging time series prediction model, the time series change law of the charging station load data can be determined.
[0049] In the embodiments of the present application, the charging load inversion prediction model includes an encoder and a decoder. When constructing the charging load prediction model, a charging time sequence prediction model can be added between the encoder and the decoder to construct the charging load prediction model.
[0050] In step S103, multi-source charging data affecting the charging station load is input into the charging load prediction model to obtain the charging station load prediction result.
[0051] In this step, the multi-source charging data refers to multi-type charging data used to characterize the recent charging situation of the charging station. The multi-source charging data includes the recent charging station load, the occupancy of charging piles, the charging duration, meteorological data, and the passenger flow.
[0052] Among them, the multi-source charging data includes multiple charging data variables. As an example, the charging data variables can be the charging duration of the vehicle, the daily passenger flow of the charging station, or the temperature during charging. The recent charging situation can refer to the charging situation within a preset time range before the current moment. As an example, the recent charging situation can be the charging situation within the past 1 month.
[0053] The charging station load prediction result can refer to the future charging station load prediction result. As an example, the charging station load prediction result can be the charging station load prediction result for the next 1 day.
[0054] Next, refer to Figure 2 to introduce the determination process of the charging station load prediction result.
[0055] Figure 2 shows a schematic structural diagram of the charging load prediction model provided by the embodiments of the present application. As Figure 2 shown, the charging load inversion prediction model includes an embedding layer 210, an encoder 220, an LSTM layer 230, and a decoder 240. When determining the charging station load prediction result, the multi-source charging data is first input into the embedding layer 210. The embedding layer 210 is used to independently aggregate the global features of each charging data variable, map the entire sequence of each charging data variable into a high-dimensional feature vector, and the high-dimensional feature vector includes all the temporal changes of the charging data variable in the past time. Among them, the high-dimensional feature vector is a mathematical representation of the word. In addition, positional encoding is also required to save the relative or absolute position of the word in the entire sequence.
[0056] Then, the high-dimensional feature vectors corresponding to all the obtained charging data variables are input into the encoder 220, and the encoder 220 is used to encode the high-dimensional feature vectors to obtain the initial charging feature encoding.
[0057] Here, the encoder 220 includes a multi-head attention layer 221, Add&Norm 222, a feed-forward neural network layer 223, and Add&Norm 224. When using the encoder 220 to encode high-dimensional feature vectors, the multi-head attention layer 221 can be utilized to establish the dependency relationships between different charging data variables in the multi-source charging data. The multi-head attention layer 221 will model the correlations between different high-dimensional feature vectors, helping the charging load prediction model capture the dependency relationships between different positions in the initial charging feature encoding. Multiple parallel attention heads can respectively learn different feature representations of the input data, thereby obtaining information in different subspaces, and finally splicing these feature representations together to obtain multi-source relationship data, enhancing the feature expression ability of the charging load prediction model.
[0058] Input the multi-source relationship data and the high-dimensional feature vectors into Add&Norm 222, and Add&Norm 222 outputs the normalized multi-source relationship data. Add&Norm 222 includes a residual connection Add and layer normalization Norm. The Add in Add&Norm 222 is used to directly add the high-dimensional feature vectors to the output of the multi-head attention layer 221, alleviating the problem of gradient disappearance or gradient explosion in the training of deep networks, so as to help information bypass the intermediate layers and ensure that the information can be transmitted smoothly. The Norm in Add&Norm 222 is used to perform normalization processing on the output of Add to maintain the stability of the data distribution.
[0059] Among them, in the initial charging load prediction model, layer normalization normalizes multiple charging data variables at the same moment, making each charging data variable mixed together and indistinguishable. However, in the improved charging load inversion prediction model, layer normalization acts on the inside of the words, allowing the features of all charging data variables to be in a relatively unified distribution, weakening the differences in measurement units, and effectively dealing with the non-stationarity problem of time series.
[0060] Input the normalized multi-source relationship data into the feed-forward neural network layer 223. Among them, the feed-forward neural network layer 223 is composed of two fully connected layers. The first fully connected layer maps the original dimension of the multi-source relationship data to a larger dimension, and the activation function of the first layer is Relu. The mapped multi-source relationship data is non-linearly transformed through the activation function Relu. The second fully connected layer maps the non-linearly transformed multi-source relationship data back to the original dimension, and the second fully connected layer does not use an activation function. Utilize the feed-forward neural network layer 223 to perform layer-by-layer encoding on the historical observation features of the multi-source relationship data in the time dimension to obtain the initial charging feature encoding.
[0061] Among them, in the initial charging load prediction model, the feedforward neural network layer acts on multiple charging data variables collected at the same time, and the generation times of multiple charging data variables may not be consistent. However, in the improved charging load inversion prediction model, the feedforward neural network layer acts on the entire sequence corresponding to the same charging data variable, which has a large enough model capacity to extract the temporal features shared in historical observations and future predictions, and obtains the prediction result through feature extrapolation.
[0062] Input the initial charging feature encoding into Add&Norm224. Add&Norm224 outputs the normalized initial charging feature encoding. Add&Norm224 also includes a residual connection Add and layer normalization Norm. The Add in Add&Norm224 is used to directly add the normalized multi-source relationship data to the output of the feedforward neural network layer 223 to ensure the integrity of information transmission. The Norm in Add&Norm224 is used to normalize the output of Add to maintain the stability of the data distribution.
[0063] Input the normalized initial charging feature encoding into the LSTM layer 230. The LSTM layer 230 can refer to a trained charging time series prediction model, and the LSTM layer 230 is used for time series feature extraction. Use the charging time series prediction model to extract time series features from the normalized initial charging feature encoding to obtain time series charging feature encoding. Among them, the LSTM layer 230 includes multiple stacked LSTM cells, and the multiple stacked LSTM cells include gated attention units.
[0064] In this way, combining the dynamic weight allocation of the multi-head attention layer 221 with the gated state of the LSTM layer 230 enables the gated attention units in the LSTM layer 230 to achieve cross-scale adaptive adjustment to accurately reflect the importance of features.
[0065] In addition, the combination of the charging load inversion prediction model and the charging time series prediction model forms a two-way interactive hierarchical feature extraction process. On the one hand, the charging time series prediction model can capture the dynamic evolution law in the time dimension; on the other hand, the charging load inversion prediction model is responsible for spatial modeling across charging data variables. These two complement each other and jointly improve the performance of the charging load prediction model in spatio-temporal feature learning.
[0066] Input the time series charging feature encoding into the decoder 240. Use the decoder 240 to decode the time series charging feature encoding to obtain the charging station load prediction result. The decoder 240 includes a multi-head attention mechanism, a feedforward neural network, and Add&Norm. The decoder 240 can use the attention mechanism to dynamically adjust the weights in different time steps, thereby effectively capturing the complex dependencies in time series data.
[0067] It should be noted that the charging load prediction model includes N modules composed of an encoder 220, an LSTM layer 230, and a decoder 240. The specific number of N can be adjusted according to task requirements. Increasing the number of N can improve the capacity and complexity of the charging load prediction model, thereby enhancing the performance of the model.
[0068] In an optional embodiment, before inputting multi-source charging data into the charging load inversion prediction model, a normalization method can be used to normalize the multi-source charging data to obtain multi-source charging station normalized data. Then, a correlation analysis is performed on the multi-source charging station normalized data to obtain multi-source charging station normalized data with correlations between different charging data variables, so as to perform predictions using the multi-source charging station normalized data with correlations between different charging data variables.
[0069] Among them, the normalization method can refer to the Min-Max normalization method. The Min-Max normalization method can map multi-source charging data to a specified range (such as within the range of [0, 1]) through linear transformation, so that features of different scales are mapped to the same range, and the values of each feature are at the same order of magnitude, to eliminate the influence of dimensions, thereby enhancing the stability of the model and the reliability of the prediction results. At the same time, since the Min-Max normalization method uses linear transformation, it can retain the relative relationship between different scale features in the multi-source charging data, and the normalized data values can accurately reflect the proportional relationship between different scale features in the multi-source charging data.
[0070] In addition, a Generalized Additive Models (GAM) can be used to perform a correlation analysis on the multi-source charging station normalized data to explore its internal correlation. Through the correlation analysis, data with significant correlations can be retained, and irrelevant data can be excluded, which helps to reduce the interference of noise on the charging load prediction model, thereby effectively improving the prediction accuracy and reliability of the model.
[0071] In one example, before inputting multi-source charging data into the charging load prediction model, the charging load prediction model needs to be trained and evaluated. At this time, Dynamic Time Warping (DTW), Mean Absolute Error (MAE), and Mean Squared Error (MSE) can be used as model evaluation metrics to evaluate the performance of the model in the load prediction task. Among them, Dynamic Time Warping (DTW) is used to quantify the similarity between time series and solve the problem of inconsistent time steps through elastic matching; Mean Absolute Error (MAE) measures the average absolute difference between the predicted value and the true value; Mean Squared Error (MSE) is used to measure the average squared error between the predicted value and the true value, giving higher weights to larger errors. When the model evaluation metrics do not meet the requirements, the charging load prediction model needs to be retrained.
[0072] Based on the same inventive concept, an embodiment of the present application also provides a charging station load prediction device corresponding to the charging station load prediction method. Since the principle of solving problems by the device in the embodiment of the present application is similar to that of the above-mentioned charging station load prediction method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.
[0073] Please refer to Figure 3 , Figure 3 which is a schematic structural diagram of a charging station load prediction device provided by an embodiment of the present application. As Figure 3 shown in
[0074] The model improvement module 301 is used to improve the initial charging load prediction model to obtain a charging load inversion prediction model to capture the coupling relationship between multiple charging data variables. The initial charging load prediction model is a converter model for predicting the charging station load;
[0075] The model construction module 302 is used to construct a charging load prediction model based on the charging load inversion prediction model and the charging time series prediction model. The charging time series prediction model is used to capture the time series characteristics of the charging data;
[0076] The load prediction model 303 is used to input multi-source charging data affecting the charging station load into the charging load prediction model to obtain a charging station load prediction result.
[0077] Please refer to Figure 4 , Figure 4 which is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 4 shown in
[0078] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 runs, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, the steps of the charging station load prediction method in the method embodiment as described above can be executed. For the specific implementation manner, reference can be made to the method embodiment, which will not be elaborated here. Figure 1 shown in the method embodiment, and for the specific implementation manner, reference can be made to the method embodiment, which will not be elaborated here.
[0079] The embodiment of the present application also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is run by a processor, the steps of the charging station load prediction method in the method embodiment as described above can be executed. For the specific implementation manner, reference can be made to the method embodiment, which will not be elaborated here. Figure 1 shown in the method embodiment, and for the specific implementation manner, reference can be made to the method embodiment, which will not be elaborated here.
[0080] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be elaborated here.
[0081] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the couplings or direct couplings or communication connections shown or discussed with each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.
[0082] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0083] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.
[0084] When the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.
[0085] Finally, it should be noted that the above-mentioned embodiments are only specific implementation manners of this application, used to illustrate the technical solution of this application, rather than limiting it. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the art within the technical scope disclosed in this application can still modify the technical solutions recorded in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, 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 this application, and should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.
Claims
1. A method for predicting the load of a charging station, characterized in that the load Including: Improve the initial charging load prediction model to obtain a charging load inversion prediction model for capturing the coupling relationship between multiple charging data variables. The initial charging load prediction model is a converter model for predicting the load of a charging station. Add a charging time series prediction model to the charging load inversion prediction model to construct a charging load prediction model. The charging time series prediction model is a long short-term memory model for capturing the time series characteristics of charging data. Input multi-source charging data affecting the load of the charging station into the charging load prediction model to obtain a charging station load prediction result.
2. The method according to claim 1, wherein The charging load inversion prediction model includes an encoder and a decoder. Adding a charging time series prediction model to the charging load inversion prediction model to construct a charging load prediction model includes: Add the charging time series prediction model between the encoder and the decoder to construct a charging load prediction model.
3. The method according to claim 2, wherein The charging load inversion prediction model further includes an embedding layer. Inputting multi-source charging data affecting the load of the charging station into the charging load prediction model to obtain a charging station load prediction result includes: Use the embedding layer to map the entire sequence of the same charging data variable into a high-dimensional feature vector. Use the encoder to encode the high-dimensional feature vector to obtain an initial charging feature encoding. Use the charging time series prediction model to extract time series features from the initial charging feature encoding to obtain a time series charging feature encoding. Use the decoder to decode the time series charging feature encoding to obtain a charging station load prediction result.
4. The method according to claim 3, wherein The encoder includes a multi-head attention layer and a feed-forward neural network layer. Using the encoder to encode the multi-source charging data to obtain an initial charging feature encoding includes: Use the multi-head attention layer to establish the dependency relationship between different charging data variables in the multi-source charging data to obtain multi-source relationship data. Use the feed-forward neural network layer to encode the historical observation features of the multi-source relationship data layer by layer in the time dimension to obtain an initial charging feature encoding.
5. The method according to claim 4, wherein The encoder further includes a first residual connection and normalization layer, and a second residual connection and normalization layer. The first residual connection and normalization layer is used to directly add the multi-source charging data to the multi-source relationship data and perform normalization processing to obtain normalized multi-source relationship data. The second residual connection and normalization layer is used to directly add the normalized multi-source relationship data to the initial charging feature encoding and perform normalization processing to obtain normalized initial charging feature encoding.
6. The method according to claim 1, wherein Improving the initial charging load prediction model to obtain a charging load inversion prediction model includes: Change the words of the initial charging load prediction model from multiple charging data variables collected at the same time to the entire sequence of the same charging data variable.
7. The method according to claim 1, wherein The method further includes: Adopt a standardization method to normalize the multi-source charging data to obtain multi-source charging standard data. Perform a correlation analysis on the multi-source charging standard data to obtain the multi-source charging standard data with correlations between different charging data variables, so as to use the multi-source charging standard data with correlations between different charging data variables for prediction.
8. A charging station load prediction device, characterized in that, It includes: A model improvement module, configured to improve an initial charging load prediction model to obtain a charging load inversion prediction model for capturing the coupling relationship between multiple charging data variables, where the initial charging load prediction model is a converter model for predicting the load of a charging station; A model construction module, configured to add a charging time series prediction model to the charging load inversion prediction model to construct a charging load prediction model, where the charging time series prediction model is a long short-term memory model for capturing the time series characteristics of charging station data; A load prediction model, configured to input the multi-source charging data affecting the load of a charging station into the charging load prediction model to obtain a charging station load prediction result.
9. An electronic device, characterized in that, It includes: A processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus, and the processor executes the machine-readable instructions to perform the steps of the charging station load prediction method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is run by the processor, it performs the steps of the charging station load prediction method according to any one of claims 1 to 7.