Electric vehicle load prediction method and system

Abstract

The invention discloses an electric vehicle load prediction method and system. The method comprises the steps of obtaining a load sequence of a target home subscriber bus, and performing first preprocessing on the load sequence; presetting a first feature recognition model, and obtaining target electric vehicle features in combination with the first preprocessing result; and according to the target electric vehicle characteristics, electric vehicle load prediction is carried out in combination with the second load prediction model. The accuracy of electric vehicle charging demand prediction is improved, fluctuation of a power system can be effectively reduced, and resource configuration is optimized. Secondly, the method can process multi-source data in real time, is suitable for charging behaviors of different users, and improves the response capability of a power grid. In addition, the method is helpful for a power operator to carry out reasonable scheduling in peak hours, and the operation cost is reduced. Finally, the load prediction method provides a scientific basis for planning and construction of electric vehicle charging infrastructures, and assists the development of a future smart power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle load prediction, and in particular to an electric vehicle load prediction method and system. Background Art

[0002] Globally, electric vehicles (EVs) have attracted increasing attention due to their environmentally friendly characteristics and compatibility with renewable energy. As the penetration rate of EVs increases, new challenges are posed to the load management and optimal scheduling of power systems. Traditional load forecasting methods often rely on historical data analysis and cannot fully consider the complexity and diversity of EV usage patterns. Therefore, it is particularly important to develop an efficient and accurate load forecasting method.

[0003] The load forecasting method based on discrete cosine transform (DCT) and electric vehicle (EV) data aims to provide more accurate forecasting results through in-depth analysis of EV charging demand. DCT can effectively extract EV load characteristics and reduce redundant information, so that the model can be trained on a smaller data set, thereby improving the accuracy of the forecast. At the same time, the method can also process charging data from different users in real time, taking into account the user's behavior patterns and charging habits. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides an electric vehicle load prediction method and system, which can solve the problems mentioned in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a method for predicting electric vehicle load, comprising:

[0009] Acquire a load sequence of a target household user bus, and perform a first preprocessing on the load sequence;

[0010] Preset a first feature recognition model, and obtain the target electric vehicle feature in combination with the first preprocessing result;

[0011] The electric vehicle load is predicted based on the target electric vehicle characteristics and in combination with a second load prediction model.

[0012] As a preferred solution of the electric vehicle load prediction method of the present invention, the first preprocessing includes:

[0013] Dividing the load sequence of the target household user bus into a plurality of subsequences;

[0014] The plurality of subsequences are subjected to a first feature mapping, where the first feature mapping is used to map the plurality of subsequences from an original dimension to a target dimension.

[0015] As a preferred solution of the electric vehicle load forecasting method of the present invention, the first feature recognition model includes:

[0016] The first feature recognition model is an arbitrary model whose input is a plurality of target dimension subsequences and whose output is the target electric vehicle feature.

[0017] As a preferred solution of the electric vehicle load forecasting method of the present invention, the preset first feature recognition model and the acquisition of the target electric vehicle feature in combination with the first preprocessing result include:

[0018] Perform temporal position encoding on several target dimension subsequences;

[0019] The contextual embedding attention score of each subsequence is calculated as the first feature recognition model representation learner.

[0020] As a preferred solution of the electric vehicle load forecasting method of the present invention, the second load forecasting model includes:

[0021] The second load forecasting model is an arbitrary model whose input is the characteristics of the target electric vehicle and whose output is the predicted probability of a charging event;

[0022] The second load forecasting model uses cross entropy loss for back propagation.

[0023] As a preferred solution of the electric vehicle load forecasting method of the present invention, the second load forecasting model further includes:

[0024] The household bus load data of electric vehicle effective charging tags is used to build a balanced data set;

[0025] The number of samples containing electric vehicle charging events in the balanced data set accounts for at least 20%;

[0026] The second load forecasting model is trained and validated based on the balanced data set.

[0027] As a preferred solution of the electric vehicle load forecasting method of the present invention, the first feature mapping is used to map the plurality of subsequences from the original dimension to the target dimension and includes:

[0028] The dimension value of the target dimension is greater than the dimension value of the original dimension.

[0029] In a second aspect, the present invention provides an electric vehicle load forecasting system, comprising:

[0030] A data processing module, used for acquiring a load sequence of a target household user bus and performing a first preprocessing on the load sequence;

[0031] A feature extraction module, used to preset a first feature recognition model and obtain the target electric vehicle feature in combination with the first preprocessing result;

[0032] The load detection module is used to perform electric vehicle load prediction based on the target electric vehicle characteristics in combination with a second load prediction model.

[0033] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method when executing the computer program.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a method and system for predicting electric vehicle load, obtains the load sequence of the target household user bus, and performs a first preprocessing on the load sequence; presets a first feature recognition model, and obtains the target electric vehicle characteristics in combination with the first preprocessing result; and predicts the electric vehicle load in combination with the second load prediction model according to the target electric vehicle characteristics. The accuracy of electric vehicle charging demand prediction is improved, which can effectively reduce the fluctuation of the power system and optimize resource allocation. Secondly, the method can process multi-source data in real time, adapt to the charging behavior of different users, and improve the responsiveness of the power grid. In addition, it helps power operators to make reasonable dispatches during peak hours, reduce operating costs, and promote the widespread application of renewable energy, environmental protection and sustainable development. Finally, this load forecasting method provides a scientific basis for the planning and construction of electric vehicle charging infrastructure, and helps the development of future smart grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0037] Figure 1 A method flow chart of an electric vehicle load prediction method and system provided by an embodiment of the present invention;

[0038] Figure 2 An internal structure diagram of a computer device of an electric vehicle load forecasting method and system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0040] Example 1

[0041] Reference Figure 1-Figure 2 , which is the first embodiment of the present invention, provides an electric vehicle load forecasting method and system, including:

[0042] There are some problems in the existing related technologies, such as low prediction accuracy, poor real-time performance, and difficulty in adapting to the diversity of user charging behaviors.

[0043] The present application provides a method that can effectively solve the above-mentioned problems. Next, multiple embodiments will be combined to explain in detail how to implement the electric vehicle load prediction method;

[0044] Figure 1 A method flow chart of an electric vehicle load prediction method and system is shown, including:

[0045] S101, obtaining a load sequence of a target household user bus, and performing a first preprocessing on the load sequence;

[0046] In an optional embodiment, the first preprocessing may include denoising the load sequence to reduce noise interference in the data and improve the accuracy of subsequent processing. It may also include normalizing the load sequence to ensure that the data is on a uniform scale to facilitate model learning. In addition, the first preprocessing may also include detecting and correcting outliers to avoid the impact of extreme data on the prediction results.

[0047] In the embodiment of the present application, the first preprocessing includes:

[0048] Dividing the load sequence of the target household user bus into a plurality of subsequences;

[0049] A first feature mapping is performed on the plurality of subsequences, where the first feature mapping is used to map the plurality of subsequences from an original dimension to a target dimension.

[0050] In an optional embodiment, the load sequence of the target household user bus is divided into several subsequences by presetting the subsequence size, step size, etc.

[0051] In an optional embodiment, the load sequence of the target household user bus can be divided into several subsequences by a preset sliding window method. For example, a sliding window can be set with a window size of 1 hour and a step length of 15 minutes, so that a continuous load subsequence can be obtained. In this way, the temporal variation trend of the load data can be captured, providing richer information for subsequent feature extraction and load forecasting.

[0052] In an optional embodiment, the load sequence of the target household user bus is divided into several subsequences and can also be divided by a preset clustering algorithm to identify load subsequences with similar characteristics. For example, the K-means clustering algorithm can be used to divide the load sequence into several groups according to their characteristic similarity, each group representing a typical load mode. Such processing helps to improve the adaptability of the model to different charging behaviors, thereby improving the accuracy of the prediction.

[0053] In the embodiment of the present application, in order to obtain the abstract characteristics of the load sequence of the household user bus, the load sequence p is first divided into segments, where T is the length of the load sequence p, L is the length of the subsequence, and s is the stride size. Then, the load sequence p is divided into load subsequence groups in is the nth load subsequence, where is the set of real numbers.

[0054] In the embodiment of the present application, the first feature mapping is used to map a plurality of subsequences from the original dimension to the target dimension, including:

[0055] The dimension value of the target dimension is greater than the dimension value of the source dimension.

[0056] In an optional embodiment, the first feature map can use a deep learning model, such as a convolutional neural network (CNN) or a recurrent neural network (RNN), to capture the temporal dependency and spatial characteristics of the load sequence. In this way, the model can learn the complex patterns of the load data and make accurate predictions.

[0057] In an optional embodiment, the first feature mapping can also be combined with other machine learning techniques, such as support vector machine (SVM) or random forest, to enhance the generalization ability of the prediction model. In practical applications, a suitable feature mapping method can be selected according to specific needs to achieve the best prediction effect.

[0058] In an optional embodiment, the first feature mapping can also be designed by presetting a mapping relationship matrix to achieve feature extraction of the load sequence. The mapping relationship matrix can be pre-trained or dynamically adjusted according to historical data. Through the mapping relationship matrix, the low-dimensional features of the load sequence can be converted into high-dimensional features, so as to better capture the inherent laws of the load data. In actual operation, the design of the mapping relationship matrix needs to consider the characteristics of the data and the requirements of the prediction model to ensure the effectiveness of feature extraction.

[0059] In the embodiment of the present application, the load subsequence group is mapped to a high-dimensional abstract feature:

[0060] P n =p n κ

[0061] Among them, P n is the abstract feature after the nth mapping, κ is the learnable mapping relationship matrix, D is the dimension of the abstract feature.

[0062] It should be noted that obtaining the load sequence of the target household user bus and performing the first preprocessing on the load sequence can provide more accurate and reliable input data for subsequent load forecasting. Through the above preprocessing steps, noise can be effectively removed, data can be standardized, and outliers can be processed, thereby ensuring that the load forecasting model can be trained and predicted on a clean and standardized data set. In addition, by dividing the load sequence into subsequences and performing feature mapping, the changing characteristics of the load data at different time scales can be captured, which is crucial for understanding the user's charging behavior and predicting future load changes. Ultimately, these preprocessing and feature extraction steps provide high-quality input features for the load forecasting model, thereby improving the accuracy and reliability of the prediction.

[0063] S102, presetting a first feature recognition model, and obtaining the target electric vehicle feature in combination with the first preprocessing result;

[0064] In the embodiment of the present application, the first feature recognition model includes:

[0065] The first feature recognition model is an arbitrary model whose input is a plurality of target dimension subsequences and whose output is the target electric vehicle feature.

[0066] In an optional embodiment, the first feature recognition model may use a convolutional neural network (CNN). If the input data has local correlation (for example, battery voltage changes in a short period of time), then using CNN to capture these local features will be very effective.

[0067] In an optional embodiment, the first feature recognition model can also use a recurrent neural network (RNN) or a long short-term memory network (LSTM). When it is necessary to process sequence data with time dependency, RNN or its variant LSTM is a very suitable choice. They can remember past input information, which is particularly useful for analyzing time series data of electric vehicles.

[0068] In an optional embodiment, the first feature recognition model can use an attention mechanism, and when processing long sequence data, the attention mechanism can be combined to improve the effect of the model. This allows the model to focus on certain parts of the input sequence when processing each output, thereby better capturing important information.

[0069] In an optional embodiment, the first feature recognition model can use an ensemble learning method, such as random forest or gradient boosting machine, especially when the input data is not highly time-correlated sequence data. These methods can improve prediction accuracy and model stability by combining multiple weak learners.

[0070] In the embodiment of the present application, the first feature recognition model is preset, and the target electric vehicle feature is acquired in combination with the first preprocessing result, including:

[0071] Perform temporal position encoding on several target dimension subsequences;

[0072] The contextual embedding attention score of each subsequence is calculated as the first feature recognition model representation learner.

[0073] In an optional embodiment, temporal position encoding of several target dimension subsequences may be performed using a preset temporal bias matrix, or position encoding technology may be used to incorporate time information into the subsequences so that the model can understand the relative or absolute position of each element in the sequence.

[0074] In an optional embodiment, the temporal position encoding may be fixed or dynamically learned according to the data set. In this way, the model can take into account the impact of time factors on the electric vehicle load when processing sequence data.

[0075] In the embodiment of the present application, in order to capture the time dependency contained in the time series data, time series position encoding is performed, and a preset time series offset matrix is ​​selected for encoding:

[0076] P pos =P+b pos

[0077] Among them, b pos is the learnable temporal bias matrix, P pos It is the abstract feature after temporal position offset.

[0078] In the embodiment of the present application, the multi-head attention network calculates the contextual embedding attention score of each subsequence, thereby serving as a representation learner for the subsequent electric vehicle charging event prediction module. Each subsequence is mapped to an attention head h, and the mapping relationship is expressed as:

[0079] Q h =P pos W h,Q

[0080] K h =P pos W h,K

[0081] V h =P pos W h,V

[0082] Among them, W h,Q , W h,K , W h,V is the learnable mapping matrix of the hth attention head, and the feature dimension after mapping is D m , and then calculate the abstracted attention head score α h and the hidden representation ψ h ;

[0083]

[0084] ψ h =α h ×V h

[0085] Where S() is the Softmax function, that is

[0086]

[0087] Kh T K h The transpose of V h are learnable parameters.

[0088] In an optional embodiment, the target electric vehicle characteristics may include battery status information, vehicle location information, user driving habits, etc. These characteristics can provide a more comprehensive perspective for electric vehicle load forecasting and help the model more accurately predict charging demand and load changes.

[0089] In the embodiments of the present application, the target electric vehicle characteristics are not limited. The present application uses battery status information and vehicle location information as target electric vehicle characteristics. Other relevant technical personnel can make other detailed selections according to actual needs, such as battery temperature, battery health status, battery charge and discharge cycle times in battery status information. Vehicle location information may include the current area of ​​the vehicle, driving route, destination prediction, etc. User driving habit information can cover driving behavior data such as acceleration mode, braking frequency, and average speed.

[0090] It should be noted that the first feature recognition model may be designed accordingly according to the features of the target electric vehicle actually selected.

[0091] It should be noted that the preset first feature recognition model and the target electric vehicle features obtained in combination with the first preprocessing result can more accurately reflect the real-time status and potential demand of the electric vehicle. By integrating these features, the model can learn the complex pattern of electric vehicle load changes and predict the charging demand in the future. In addition, the design of this feature recognition model allows the system to flexibly adapt to different types of input data, thereby improving the accuracy and reliability of the prediction. In practical applications, this prediction method and system can help power suppliers optimize grid load distribution, reduce energy waste, and provide more convenient charging services for electric vehicle users.

[0092] S103, performing electric vehicle load forecasting based on the target electric vehicle characteristics and in combination with a second load forecasting model.

[0093] In the embodiment of the present application, the second load forecasting model includes:

[0094] The second load forecasting model is an arbitrary model whose input is the characteristics of the target electric vehicle and whose output is the predicted probability of the charging event;

[0095] The second load forecasting model uses cross entropy loss for back propagation.

[0096] In an optional embodiment, the second load forecasting model can be trained by a training data set to improve the accuracy of the forecast. The model can be a deep learning model, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a long short-term memory network (LSTM), or a traditional machine learning model, such as a support vector machine (SVM) or a random forest. When selecting a model, it is necessary to consider the characteristics of the data and the specific requirements of the forecasting task.

[0097] In the embodiment of the present application, the second load forecasting model further includes:

[0098] The household bus load data of electric vehicle effective charging tags is used to build a balanced data set;

[0099] The number of samples containing electric vehicle charging events in the balanced data set accounts for at least 20%;

[0100] A second load forecasting model is trained and validated based on the balanced data set.

[0101] In the embodiment of the present application, the training process of the second load forecasting model includes the following steps:

[0102] Collect and preprocess historical EV load data, including target EV characteristics and corresponding charging event history records;

[0103] Furthermore, the preprocessed data is used to train the second load forecasting model, and the model parameters are adjusted through an optimization algorithm to minimize the cross entropy loss function;

[0104] Furthermore, during the training process, the validation set is used to evaluate the model performance to prevent overfitting, and the generalization ability of the model is improved by adjusting the model structure or parameters;

[0105] Furthermore, after training is completed, the model is finally evaluated using the test set to ensure that the model has good predictive performance on unseen data.

[0106] It should be noted that through the above steps, the second load forecasting model can learn the law of electric vehicle load changes and predict the probability of charging events in the future. Combined with the real-time electric vehicle features obtained by the first feature recognition model, the second load forecasting model can provide more accurate load forecasting results, bringing practical application value to power suppliers and electric vehicle users.

[0107] In the embodiment of the present application, all hidden representations ψ h The hidden representation features ψ are concatenated row by row and input into the feedforward neural network FFN for processing to obtain the representation features processed by the feature representation learning module, that is,

[0108] Z=ReLu(ψφ 1 +b 1 )φ 2 +b 2

[0109] Among them, φ 1 ,φ 2 is the learnable weight parameter, b 1 , b 2 is a learnable bias parameter and ReLu() is an activation function.

[0110] Among them, the electric vehicle features obtained by representation are input into the charging event module containing a fully connected network to predict charging events, that is,

[0111] y m =σ(W m T Z)

[0112] Among them, W m T is the transposed learnable prediction weight matrix for the mth period, σ() is the sigmoid activation function, and y m is the predicted probability of the charging event in the mth period.

[0113] In an optional embodiment, in order to train the proposed model, according to the true label y of the dataset m , calculate the cross entropy loss for back propagation, that is

[0114]

[0115] Among them, y i,m is the actual charging event of the i-th user in the m-th period, y i,m is the predicted charging event probability of the i-th user in the m-th time period.

[0116] In an optional embodiment, a balanced data set is constructed using household user bus load data containing valid electric vehicle charging tags. Since the proportion of household user electric vehicle charging events is small in a given time period, the data set needs to be balanced, and the number of samples containing electric vehicle charging events accounts for at least 20% to ensure that the first feature recognition model can learn the target feature.

[0117] In an optional embodiment, the present application designs a DCT-EV model including a load embedding module, an electric vehicle feature representation learning module, and a charging event prediction module. Specifically:

[0118] Obtaining a load sequence of a household user bus, and dividing the obtained load sequence into a plurality of subsequence groups;

[0119] Furthermore, the subsequence features are mapped to high-dimensional abstract features and input into the electric vehicle feature representation learning module containing a multi-head attention network for feature representation;

[0120] Furthermore, the obtained electric vehicle features are input into a charging event module containing a fully connected network to predict charging events and output the probability of charging.

[0121] Furthermore, the high-dimensional abstract features are input into the electric vehicle feature representation learning module containing a multi-head attention network for feature representation.

[0122] Furthermore, the high-dimensional abstract features are input into the electric vehicle feature representation learning module containing a multi-head attention network for feature representation. The multi-head attention network calculates the contextual embedding attention score of each subsequence, thereby serving as a representation learner for the subsequent electric vehicle charging event prediction module.

[0123] Furthermore, all hidden representations ψ of the output h The hidden representation features ψ are concatenated row by row and input into the feed-forward neural network FFN for processing to obtain the representation features processed by the feature representation learning module.

[0124] Among them, the electric vehicle features obtained by representation are input into the charging event module containing a fully connected network to predict charging events. The bus load data of household users containing effective charging tags of electric vehicles is used to construct a balanced data set, of which 80% is a training set and 20% is a test set. A loss function is constructed to train the DCT-EV model. Based on the balanced data set, the performance of the proposed DCT-EV model is evaluated by comparing the existing deep neural network models.

[0125] In summary, the present invention proposes a method for predicting electric vehicle load, which obtains the load sequence of the target household user bus and performs a first preprocessing on the load sequence; presets a first feature recognition model, and obtains the target electric vehicle characteristics in combination with the first preprocessing result; and predicts the electric vehicle load in combination with the second load prediction model according to the target electric vehicle characteristics. The accuracy of electric vehicle charging demand prediction is improved, which can effectively reduce the fluctuation of the power system and optimize resource allocation. Secondly, the method can process multi-source data in real time, adapt to the charging behavior of different users, and improve the responsiveness of the power grid. In addition, it helps power operators to make reasonable dispatches during peak hours, reduce operating costs, and promote the widespread application of renewable energy, environmental protection and sustainable development. Finally, this load forecasting method provides a scientific basis for the planning and construction of electric vehicle charging infrastructure, and helps the development of future smart grids.

[0126] This embodiment also provides an electric vehicle load prediction system, including:

[0127] A data processing module, used for acquiring a load sequence of a target household user bus and performing a first preprocessing on the load sequence;

[0128] A feature extraction module, used to preset a first feature recognition model and obtain the target electric vehicle feature in combination with the first preprocessing result;

[0129] The load detection module is used to perform electric vehicle load prediction based on the characteristics of the target electric vehicle in combination with the second load prediction model.

[0130] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0131] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 2 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for predicting the load of an electric vehicle is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0132] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0133] Acquire a load sequence of a target household user bus, and perform a first preprocessing on the load sequence;

[0134] Preset a first feature recognition model, and obtain the target electric vehicle feature in combination with the first preprocessing result;

[0135] According to the characteristics of the target electric vehicle, the electric vehicle load is predicted in combination with the second load prediction model.

[0136] Example 2

[0137] Reference Figure 1-Figure 2 , which is an embodiment of the present invention, provides an electric vehicle load prediction method and system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0138] Classical deep learning was used for testing and compared with the method of the present invention. The test results were compared by scientific demonstration to verify the real effect of the method.

[0139] The classic deep learning model is deployed on the same type of test equipment, and the algorithms used are DNN and LSTM. Table 1 is the test environment table, and Table 2 is the experimental results comparison table.

[0140] Test environment: As shown in Table 1, test device 1 deploys the classic deep learning model, and test device 2 deploys the method of the present invention. The statistical recognition error MSE is calculated as follows:

[0141]

[0142] Among them, the parameter x t is the predicted charging event probability of the target working device at time t, x t It is the actual charging event of the target working device at time t. It is 1 if it is charging and 0 if it is not charging.

[0143] Table 1 Test environment

[0144]

[0145] Table 2 Test results

[0146] Test Group Test equipment 1 Test equipment 2 Test equipment 3 MSE 0.029 0.011 0.009

[0147] It should be noted that

[0148] The experimental results show that the method of the present invention has higher accuracy and efficiency in predicting electric vehicle load. By comparing the MSE values ​​of test equipment 1 and test equipment 2, it can be seen that the MSE value of the method of the present invention is significantly lower than that of test equipment 1 using the classic deep learning model. This shows that the method of the present invention has a smaller error and a more reliable prediction result when predicting the probability of electric vehicle charging events. In addition, the MSE value of test equipment 3 further verifies the superiority of the method of the present invention. Its MSE value is 0.009, which is the lowest among the three test equipments, indicating that the method of the present invention has good performance in practical applications.

[0149] In summary, the present invention provides an effective electric vehicle load forecasting method and system, which can significantly improve the accuracy of electric vehicle load forecasting by adopting specific algorithms and data processing processes, and provide strong technical support for the optimization and management of electric vehicle charging networks.

[0150] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0151] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0152] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0153] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0154] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0155] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0156] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for predicting electric vehicle load, characterized in that: include: Acquire a load sequence of a target household user bus, and perform a first preprocessing on the load sequence; Preset a first feature recognition model, and obtain the target electric vehicle feature in combination with the first preprocessing result; The electric vehicle load is predicted based on the target electric vehicle characteristics and in combination with a second load prediction model.

2. The electric vehicle load forecasting method according to claim 1, characterized in that: The first preprocessing comprises: Dividing the load sequence of the target household user bus into a plurality of subsequences; The plurality of subsequences are subjected to a first feature mapping, where the first feature mapping is used to map the plurality of subsequences from an original dimension to a target dimension.

3. The electric vehicle load forecasting method according to claim 2, characterized in that: The first feature recognition model includes: The first feature recognition model is an arbitrary model whose input is a plurality of target dimension subsequences and whose output is the target electric vehicle feature.

4. The electric vehicle load forecasting method according to claim 3, characterized in that: The preset first feature recognition model and the first preprocessing result are combined to obtain the target electric vehicle feature, including: Perform temporal position encoding on several target dimension subsequences; The contextual embedding attention score of each subsequence is calculated as the first feature recognition model representation learner.

5. The electric vehicle load prediction method according to claim 4, characterized in that: The second load forecasting model includes: The second load forecasting model is an arbitrary model whose input is the characteristics of the target electric vehicle and whose output is the predicted probability of a charging event; The second load forecasting model uses cross entropy loss for back propagation.

6. The electric vehicle load forecasting method according to claim 5, characterized in that: The second load forecasting model also includes: The household bus load data of electric vehicle effective charging tags is used to build a balanced data set; The number of samples containing electric vehicle charging events in the balanced data set accounts for at least 20%; The second load forecasting model is trained and validated based on the balanced data set.

7. The electric vehicle load forecasting method according to claim 6, characterized in that: The first feature mapping is used to map the plurality of subsequences from the original dimension to the target dimension, and includes: The dimension value of the target dimension is greater than the dimension value of the original dimension.

8. An electric vehicle load forecasting system, characterized in that: include: A data processing module, used for acquiring a load sequence of a target household user bus and performing a first preprocessing on the load sequence; A feature extraction module, used to preset a first feature recognition model and obtain the target electric vehicle feature in combination with the first preprocessing result; The load detection module is used to perform electric vehicle load prediction based on the target electric vehicle characteristics in combination with a second load prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images