Electric vehicle charging load adjustable potential prediction method and device and electronic equipment

Abstract

The invention discloses an electric vehicle charging load adjustable potential prediction method and device and electronic equipment. The method relates to the technical field of new energy and energy conservation. The method comprises the following steps: acquiring current charging load data of a target charging station in a current time period; performing data enhancement processing on the current charging load data to obtain first charging load data; based on the first charging load data, adopting a charging load adjustable potential prediction model to obtain a charging load adjustable potential prediction result of the target charging station in a prediction period, the charging load adjustable potential prediction model being based on historical charging load data of the target charging station and a corresponding historical charging load adjustable potential; and carrying out optimization training on the bidirectional improved long-short-term memory network. According to the method, the technical problem that the accuracy of electric vehicle charging load adjustable potential prediction is insufficient due to the fact that the user behavior preference, the adaptability of electricity price change and the long-term dependency relationship of the time sequence are not fully considered is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy and energy-saving technology, in particular to an electric vehicle charging load adjustable potential prediction method and device and electronic equipment. BACKGROUND

[0002] In recent years, with the increasing awareness of clean energy and environmental protection worldwide, electric vehicles as an important green transportation tool, its market share and user acceptance are rapidly growing. The popularity of electric vehicles not only helps to reduce the consumption of fossil fuels and greenhouse gas emissions, but also promotes the diversification of energy and the development of smart grids. However, the large-scale access of electric vehicles to the power grid brings new challenges to the operation and management of the power system, especially the management of electric vehicle charging load. Because the charging behavior of electric vehicle users is highly dependent on personal habits, electricity price policy, battery status and other factors, its charging load has significant randomness, uncertainty and time dependence, which directly affects the balance of supply and demand, voltage stability of the power grid, and planning and construction of charging infrastructure. The related technology has obvious deficiencies in the prediction of electric vehicle charging load adjustable potential, mainly in the following aspects:

[0003] First, most prediction models fail to fully consider the user's choice of charging time and location, especially the quantification of price sensitivity, resulting in a large deviation in the prediction results and an inability to effectively reflect the actual charging demand. Second, electric vehicle charging load is significantly affected by price changes, but related technology often ignores the relationship between electric vehicle charging load and electricity price, making the model perform poorly in the case of large price fluctuations. In addition, traditional prediction models, such as random forests or single-direction long short-term memory networks, while performing well on certain charging data sets, are limited in prediction performance due to gradient vanishing and overfitting problems when dealing with data with complex time dependencies, especially long-term dependencies. In summary, the prediction model in the related technology, when faced with large-scale electric vehicle access, complex user behavior and changing electricity price environment, fails to fully consider user behavior preferences, adaptability to price changes and long-term dependence of time series, resulting in insufficient accuracy in predicting electric vehicle charging load adjustable potential and difficulty in meeting the needs of fine management of power systems and optimized operation of power grids.

[0004] In view of the above problems, no effective solution has been proposed so far. SUMMARY

[0005] The embodiment of the present application provides a kind of electric vehicle charging load adjustable potential prediction method, device and electronic equipment, to at least solve the technical problem of insufficient precision of electric vehicle charging load adjustable potential prediction due to failure to fully consider user behavior preference, the adaptability of price variation and long-term dependence of time sequence.

[0006] According to one aspect of the embodiment of the present application, a method for predicting the adjustable potential of electric vehicle charging load is provided, comprising: obtaining current charging load data of a target charging station in a current period; performing data enhancement processing on the current charging load data to obtain first charging load data, wherein the first charging load data represents charging load data corresponding to the current period and containing user behavior preference information; and based on the first charging load data, using a charging load adjustable potential prediction model to obtain a charging load adjustable potential prediction result of the target charging station in a prediction period, wherein the charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on historical charging load data of the target charging station and corresponding historical charging load adjustable potential, and the prediction period is a period of a first predetermined length after the current period.

[0007] According to another aspect of the embodiment of the present application, a device for predicting the adjustable potential of electric vehicle charging load is also provided, comprising: a current charging load data acquisition module configured to obtain current charging load data of a target charging station, wherein the current charging load data is charging load data of the target charging station in a current period; a first charging load data acquisition module configured to perform data enhancement processing on the current charging load data to obtain first charging load data, wherein the first charging load data represents charging load data corresponding to the current period and containing user behavior preference information; and a charging load adjustable potential prediction module configured to, based on the first charging load data, use a charging load adjustable potential prediction model to obtain a charging load adjustable potential prediction result of the target charging station in a prediction period, wherein the charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on historical charging load data of the target charging station and corresponding historical charging load adjustable potential, and the prediction period is a period of a first predetermined length after the current period.

[0008] According to another aspect of the embodiment of the present application, a non-volatile storage medium is also provided, which stores a plurality of instructions adapted to be loaded and executed by a processor to perform any of the electric vehicle charging load adjustable potential prediction methods described above.

[0009] According to another aspect of the embodiments of the present application, an electronic device is provided, including one or more processors and a memory, the memory being configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any of the electric vehicle charging load adjustable potential prediction methods.

[0010] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps of any of the electric vehicle charging load adjustable potential prediction methods.

[0011] In the embodiments of the present application, the current charging load data of the target charging station in the current period is obtained; the current charging load data is subjected to data enhancement processing to obtain first charging load data, wherein the first charging load data represents the charging load data of the current period containing user behavior preference information; based on the first charging load data, a charging load adjustable potential prediction model is used to obtain the charging load adjustable potential prediction result of the target charging station in a prediction period, wherein the charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical charging load adjustable potential, and the prediction period is a period of a first predetermined length after the current period. The purpose of accurately predicting the charging load adjustable potential is achieved by obtaining the current charging load data, and the data enhancement processing is performed on the current charging load data, and the charging load adjustable potential prediction model is used, so as to realize the technical effect of improving the accuracy of the electric vehicle charging load adjustable potential prediction, and further solve the technical problem of insufficient accuracy of the electric vehicle charging load adjustable potential prediction caused by not fully considering the user behavior preference, the adaptability of the price change, and the long-term dependence of the time sequence. BRIEF DESCRIPTION OF DRAWINGS

[0012] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the present application and together with the description serve to explain the present application. In the drawings:

[0013] Figure 1 is a flowchart of an electric vehicle charging load adjustable potential prediction method according to an embodiment of the present application;

[0014] Figure 2 is a schematic diagram of an optional first bidirectional improved long short-term memory network layer according to an embodiment of the present application;

[0015] Figure 3 is a flowchart of an optional electric vehicle charging load adjustable potential prediction method according to an embodiment of the present application;

[0016] Figure 4 This is a schematic diagram of an electric vehicle charging load adjustable potential prediction device according to an embodiment of the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] A bidirectional long short-term memory (Bi-LSTM) network processes the input sequence simultaneously from two directions: forward and backward. This allows it to capture contextual information within the sequence. A Bi-LSTM network consists of two independent Bi-LSTM layers: one processes the input in forward chronological order, and the other in reverse chronological order. The outputs of these two layers are then combined, typically by simply concatenating them to form the final output.

[0020] According to an embodiment of the present invention, a method embodiment for predicting the adjustable potential of electric vehicle charging load is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0021] Figure 1This is a flowchart of a method for predicting the adjustable charging load potential of electric vehicles according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0022] Step S102: Obtain the current charging load data of the target charging station in the current time period;

[0023] Optionally, real-time collection of charging load data for all electric vehicles at the target charging station during the current time period can be achieved. This can be done by communicating with the data interface of the charging station monitoring system to extract the current charging load data from the database or real-time data stream for subsequent load forecasting and adjustment potential assessment. This ensures that the model's input data is closely related to the actual charging situation, thereby improving the accuracy and practicality of the forecast.

[0024] In an optional embodiment, before obtaining the current charging load data of the target charging station, the method further includes: obtaining historical charging load data of the target charging station during a predetermined historical period, wherein the predetermined historical period is a period of a second predetermined duration prior to the current period; performing data augmentation processing on the historical charging load data to obtain second charging load data, wherein the second charging load data represents the charging load data containing user behavior preference information corresponding to the historical period; constructing a bidirectional improved long short-term memory network; and optimizing and training the bidirectional improved long short-term memory network based on the second charging load data to obtain a charging load adjustable potential prediction model.

[0025] Optionally, before obtaining the current charging load data of the target charging station, historical charging load data for the past year is first collected and processed. By performing data augmentation on the historical data, richer and more representative second charging load data can be generated. This second charging load data not only includes the original historical charging information but also incorporates behavioral preference information such as user sensitivity to charging prices, enhancing the model's learning ability. Subsequently, a bidirectional improved long short-term memory (LSTM) network is used to optimize and train the network based on the second charging load data. This bidirectional improved LSM network, through its improved memory units and bidirectional information processing capabilities, can more effectively capture and learn the time dependence of charging load and user behavior patterns. After optimization and training, a predictive model capable of predicting the adjustable potential of electric vehicle charging load is obtained, providing powerful data support and predictive tools for charging station management and power grid planning.

[0026] In one optional embodiment, data augmentation processing is performed on historical charging load data to obtain second charging load data, including: processing the historical charging load data based on preset constraints to obtain constrained historical charging load data; constructing a user electricity price sensitivity vector, wherein the user electricity price sensitivity vector is obtained based on multiple electricity price sensitivity coefficients to represent the sensitivity of electric vehicle users to changes in electricity prices, and the multiple electricity price sensitivity coefficients correspond to different electricity price ranges; and obtaining the second charging load data based on the constrained historical charging load data and the user electricity price sensitivity vector.

[0027] Optionally, when processing historical charging load data, firstly, the original data is filtered and adjusted according to preset constraints to obtain constrained historical charging load data, ensuring the data's validity and consistency with reality. Next, a user electricity price sensitivity vector is constructed, using multiple electricity price sensitivity coefficients to comprehensively reflect changes in electric vehicle users' charging behavior under different electricity price levels, thereby quantifying users' charging price sensitivity. Finally, combining the constrained historical charging load data and the user electricity price sensitivity vector, a series of virtual charging behavior scenarios are generated using Monte Carlo simulation. These scenarios reflect the impact of user behavioral preferences on charging load, resulting in more comprehensive and richer second charging load data, which improves the model's accuracy in predicting the adjustable potential of charging load.

[0028] In one optional embodiment, the constraints include at least the following: the charging power of any electric vehicle charging at the target charging station does not exceed the maximum charging power of the corresponding electric vehicle; the actual stopping time of any electric vehicle at the target charging station is obtained as follows: if the interval between two adjacent charging sessions exceeds a preset stopping time, the actual stopping time is the preset stopping time; otherwise, the actual stopping time is the interval between two adjacent charging sessions, wherein the probability distribution of the preset stopping time satisfies a preset first normal distribution, which is obtained based on the time when any electric vehicle arrives at the target charging station; the probability distribution of the single charging time of any electric vehicle satisfies a preset second normal distribution; and the probability distribution of the state of charge of any electric vehicle at the start of charging satisfies a preset third normal distribution, which is obtained based on the daily mileage probability density function.

[0029] Optionally, to ensure the accuracy and rationality of charging load prediction, a series of constraints are set to standardize and refine historical charging data. First, it is stipulated that the charging power of any electric vehicle cannot exceed its maximum charging power limit, thereby ensuring data consistency and security. Next, by comparing the interval between two consecutive charging sessions of any electric vehicle with the preset parking duration, conditional judgment logic is used to calculate the actual parking duration for each session. The preset parking duration follows a preset first normal distribution, denoted by T, where... μ represents the mean of the preset parking time for any electric vehicle, and σ represents the standard deviation of the preset parking time for any electric vehicle. The preset parking time for any electric vehicle is dynamically adjusted based on the specific time the electric vehicle arrives at the charging station. For example, if the electric vehicle arrives at the charging station between 6:00 and 16:00, μ=4 and σ=1; if the electric vehicle arrives between 0:00 and 6:00 and between 16:00 and 24:00, μ=20 and σ=4. Furthermore, the single charging time of any electric vehicle is constrained to ensure that the single charging time of any electric vehicle follows a preset second normal distribution, denoted by T. s This represents the duration of a single charge for any electric vehicle, where, μ s σ represents the average charging time of any electric vehicle on a single charge. s The standard deviation of the single charging time for any electric vehicle is represented, revealing the inherent patterns and temporal distribution of charging behavior. Finally, considering the impact of the electric vehicle's state of charge upon arrival at the charging station on its charging demand, a pre-defined third normal distribution based on the daily mileage probability density function is introduced to describe the statistical characteristics of the state of charge, denoted by d, where . μ d σ represents the average daily mileage of any electric vehicle. d The standard deviation of the daily mileage of any electric vehicle can be used to further refine the user behavior model, enabling the prediction model to more comprehensively consider various factors related to the dynamics of electric vehicle charging, thereby improving prediction accuracy and practicality. Through the design of these constraints, more realistic and reliable charging load data can be obtained, laying a solid data foundation for subsequent load forecasting.

[0030] In an optional embodiment, when multiple electricity price sensitivity coefficients include low, medium, and high electricity price sensitivity coefficients, a user electricity price sensitivity vector is constructed, including obtaining the low electricity price sensitivity coefficient through the following method:

[0031] ;

[0032] The electricity price sensitivity coefficient is obtained as follows:

[0033] ;

[0034] The high electricity price sensitivity coefficient is obtained as follows:

[0035] ;

[0036] Where M1 represents the low electricity price sensitivity coefficient, M2 represents the medium electricity price sensitivity coefficient, M3 represents the high electricity price sensitivity coefficient, P represents the electricity price, A1 represents the first preset electricity price threshold, B1 represents the second preset electricity price threshold, C1 represents the third preset electricity price threshold, D1 represents the fourth preset electricity price threshold, and E1 represents the fifth preset electricity price threshold.

[0037] Optionally, when considering the price sensitivity of electric vehicle users, a piecewise function can be used to divide users' price sensitivity into three levels, each corresponding to a different sensitivity coefficient. This allows for a detailed depiction of users' charging behavior tendencies under different electricity price levels. This hierarchical quantification method can more accurately capture the price factor in users' charging decisions, distinguish and respond to the differences in charging behavior among different user groups under specific electricity price environments, thereby making more accurate load forecasts.

[0038] In one optional embodiment, constructing a bidirectional improved long short-term memory (LSTM) network includes: optimizing bidirectional LSM units to obtain optimized bidirectional LSM units; constructing a first bidirectional improved LSM network layer based on the optimized bidirectional LSM units; performing a first linear transformation and a nonlinear transformation on the output vector of the first bidirectional improved LSM network layer to construct a second bidirectional improved LSM network layer, wherein the first linear transformation is used to adjust the dimension of the output vector of the first bidirectional improved LSM network layer; performing a second linear transformation on the output vector of the second bidirectional improved LSM network layer to construct a third bidirectional improved LSM network layer, wherein the second linear transformation is used to adjust the dimension of the output vector of the second bidirectional improved LSM network layer; and obtaining a bidirectional improved LSM network based on the first bidirectional improved LSM network layer, the second bidirectional improved LSM network layer, and the third bidirectional improved LSM network layer.

[0039] Optionally, a bidirectional improved long short-term memory network architecture is designed for the electric vehicle charging load prediction problem. Figure 2This is a schematic diagram of an optional first bidirectional improved long short-term memory network layer according to an embodiment of the present invention. The diagram shows the direction of data flow and the internal state update process. By introducing a specific function (hereinafter referred to as the HardSigmoid function) to update the bidirectional long short-term memory unit parameters, the updated bidirectional long short-term memory unit parameters are obtained, thus resulting in an optimized bidirectional long short-term memory unit. The bidirectional long short-term memory unit parameters include an input gate, a forget gate, candidate memory units, an output gate, memory units, and hidden states. For example, t represents the current time, a time node in the time series, used for time series prediction and sequence modeling; h represents the number of hidden units, the number of units used for state transfer within the bidirectional improved long short-term memory network; n represents the batch size, the number of sequence samples processed simultaneously in one training session; d represents the input dimension, i.e., the data dimension input to the network at each time step. In electric vehicle charging load prediction, this could be the charging load, electricity price, user behavior characteristics, etc., at a specific time point; W represents the weight matrix, used to calculate the output of the gating mechanism and the hidden state; b represents the bias vector, a constant bias term used when calculating the output of the gating mechanism and the hidden state, helping the network learn more complex feature mappings. This represents the Hard Sigmoid function. The parameters of the bidirectional long short-term memory unit are updated as follows: , X represents the output of the input gate at the current moment. t W represents the input information at the current moment. xi H represents the weight matrix from the current input information to the current input gate. t-1 W represents the hidden state at the previous moment. hi Let b represent the weight matrix from the hidden state of the previous time step to the input gate of the current time step. i This represents the bias vector of the input gate; , X represents the output of the forget gate at the current moment. t W represents the input information at the current moment. xf H represents the weight matrix from the current input information to the current forget gate. t-1 W represents the hidden state at the previous moment. hf b represents the weight matrix from the hidden state of the previous time step to the forget gate of the current time step. f The bias vector representing the forget gate; , X represents the output of the output gate at the current moment. t W represents the input information at the current moment. xo H represents the weight matrix from the input information at the current time step to the output gate at the current time step. t-1 W represents the hidden state at the previous moment.ho b represents the weight matrix from the hidden state of the previous time step to the output gate of the current time step. o This represents the bias vector of the output gate; , X represents the output of the candidate memory unit at the current moment. t W represents the input information at the current moment. xc H represents the weight matrix from the input information at the current time step to the candidate memory units at the current time step. t-1 W represents the hidden state at the previous moment. hc b represents the weight matrix from the hidden state of the previous time step to the candidate memory unit of the current time step. c The bias vector representing the candidate memory cell; C t C represents the output of the memory unit at the current moment. t-1 This represents the output of the memory unit from the previous time step. This indicates element-wise multiplication (Hadamard product). H t Let tanh represent the hidden state at the current time step, and let tanh represent the hyperbolic tangent function. .

[0040] Based on the optimized bidirectional long short-term memory (LSTM) units, a first bidirectional improved LSM network layer is constructed, capable of simultaneously analyzing time-series data from front to back and from back to front, capturing the bidirectional time dependence of charging load. Subsequently, the output vector of the first bidirectional improved LSM network layer is subjected to dimensionality adjustment and feature extraction through a first linear transformation and a nonlinear transformation, forming a second bidirectional improved LSM network layer. The first linear transformation adjusts the dimension of the output vector of the first bidirectional improved LSM network layer, while the nonlinear transformation is implemented using the GELU function. The complex time-series features contained in the vectors are further extracted. Finally, after a second linear transformation, the output vector of the second network layer is further adjusted to a dimension suitable for the model output, constructing a third bidirectional improved long short-term memory network layer. Through the entire construction process of the bidirectional improved long short-term memory network, the ideas of bidirectional information processing and multi-layer feature extraction are incorporated, which can effectively integrate the time-series features and user behavior preferences in historical charging data, significantly improving the accuracy and flexibility of electric vehicle charging load prediction.

[0041] In one optional embodiment, based on the second charging load data, a bidirectional improved long short-term memory network is optimized and trained to obtain a charging load adjustable potential prediction model. This includes: randomly dividing the second charging load data according to a daily cycle to obtain training samples for the bidirectional improved long short-term memory network, where the daily cycle represents a time period of one day; optimizing the bidirectional improved long short-term memory network based on the training samples by: performing multiple random augmentation processes on the training samples to obtain multiple output samples for the bidirectional improved long short-term memory network; smoothing the label weights in the bidirectional improved long short-term memory network to obtain updated label weights, where the label weights represent parameters used to adjust the difference between the predicted label value and the true label value, and the predicted label value is obtained based on the output samples; optimizing the learning rate in the bidirectional improved long short-term memory network; repeating the above operations until a predetermined termination condition is reached; and obtaining the charging load adjustable potential prediction model based on the optimized bidirectional improved long short-term memory network obtained when the predetermined termination condition is reached.

[0042] Optionally, using the second charging load data as a basis, a bidirectional improved long short-term memory network can be deeply optimized and trained on a daily time period to construct a predictive model for the adjustable potential of charging load. Specifically, the second charging load data is randomly divided, for example, 80% of the second charging load data is selected as the training set and 20% as the test set. The training samples are subjected to multiple random augmentation processes using a multi-sample Dropout algorithm. In the training iterations, multiple Dropout layers with shared parameters are used, each layer using a different mask, thereby generating multiple different output samples. These output samples are calculated using the same activation function and loss function to obtain multiple loss values. The average of these loss values ​​is then calculated as the final loss, used for gradient updates. This further increases data diversity and improves the model's predictive ability when faced with unseen data. Simultaneously, label smoothing technology is introduced to optimize the bidirectional improved long short-term memory network, enabling it to more comprehensively understand and predict changes in charging load. Label smoothing introduces a certain amount of random error into the true labels during training, thereby preventing the model from becoming overconfident and overfitting. Under normal circumstances, if the predicted label value output by the bidirectional improved long short-term memory network is equal to the true label value, the label weight is usually set to 1; otherwise, it is 0. This results in a very high certainty score when the prediction is correct, but a sharp drop in certainty score when the prediction is incorrect. By introducing label smoothing, even when the prediction is correct, the weight of the true label is slightly reduced, while the weight of the incorrectly predicted label is slightly increased, thus obtaining the updated label weight p. i , Here, θ represents a small hyperparameter that introduces error, i represents the true label value, and y represents the predicted label value output by the bidirectional improved long short-term memory network. Furthermore, the learning rate in the bidirectional improved long short-term memory network is optimized during training using the AdaMod optimizer, enabling the network to converge to the optimal solution more efficiently. Optimization is achieved by setting parameters such as the learning rate, decay rate, learning rate smoothing coefficient, and stability term in the AdaMod optimizer. Specifically, the learning rate controls the step size of model parameter updates, determining the speed at which model weights change during training iterations; the decay rate controls the exponential decay rate of the first and second gradient moments, updating the running average and squared average of the model weights to adapt to different model states; the learning rate smoothing coefficient smooths the learning rate, making it more stable during training and avoiding drastic fluctuations due to local noise; and the stability term is added to the denominator during bidirectional improved long short-term memory network update calculations to prevent zero or very small gradients, ensuring the stability of model updates. The entire training process is executed cyclically until a predetermined termination condition is met, such as reaching a certain number of iterations or the prediction error being lower than a set threshold. The final optimized network is the prediction model for the adjustable charging load potential. Through this optimization training process, it can be ensured that the prediction model for the adjustable charging load potential has high accuracy and adaptability when facing complex and variable charging demands.

[0043] Step S104: Perform data augmentation processing on the current charging load data to obtain the first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period.

[0044] Optionally, by implementing data augmentation techniques on the real-time collected current charging load data, richer and more diverse first charging load data can be generated. This first charging load data not only includes the original charging load information but also incorporates behavioral preference characteristics such as user sensitivity to charging prices. This allows the charging load adjustable potential prediction model to more accurately consider the impact of user behavior on load changes when predicting the adjustable potential of the charging load.

[0045] Step S106: Based on the first charging load data, the adjustable charging load potential prediction model is used to obtain the prediction result of the adjustable charging load potential of the target charging station in the predicted period. The adjustable charging load potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical adjustable charging load potential. The predicted period is the first predetermined period after the current period.

[0046] Optionally, by utilizing first-generation charging load data that has been data-augmented and integrated with user behavior preferences, an optimized and trained charging load adjustability potential prediction model can be used to accurately predict the future charging load adjustability potential of a target charging station. This prediction model, based on the target charging station's past charging load data and corresponding historical records of load adjustability potential, is trained on a bidirectional improved long short-term memory network, resulting in more accurate and reliable predictions. This promotes balanced management of electric vehicle charging load, reduces grid load fluctuations, and improves power system stability.

[0047] Through the above steps S102 to S106, the goal of accurately predicting the adjustable potential of electric vehicle charging load can be achieved by acquiring the current charging load data, performing data augmentation processing on the current charging load data, and using the adjustable potential prediction model of charging load. This achieves the technical effect of improving the accuracy of the adjustable potential prediction of electric vehicle charging load, and solves the technical problem of insufficient accuracy in predicting the adjustable potential of electric vehicle charging load caused by failing to fully consider user behavior preferences, the adaptability of electricity price changes, and the long-term dependence of time series.

[0048] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation method. Figure 3 This is a flowchart of an optional electric vehicle charging load adjustable potential prediction method according to an embodiment of the present invention, such as... Figure 3 As shown, the method includes:

[0049] S1: Obtain historical charging load data of the target charging station. The specific implementation process is the same as in the aforementioned embodiments, and will not be repeated here.

[0050] S2: Construct constraints and user electricity price sensitivity vectors, and perform data augmentation processing on historical charging load data. The specific implementation process is the same as the aforementioned embodiments, and will not be repeated here.

[0051] S3: Construct a bidirectional improved long short-term memory network. The specific implementation process is the same as the previous embodiments, and will not be repeated here.

[0052] S4: By using the multi-sample Dropout algorithm, label smoothing technology and AdaMod optimizer, the bidirectional improved long short-term memory network is trained to obtain a charging load adjustable potential prediction model. The specific implementation process is the same as the previous embodiment, and will not be repeated here.

[0053] S5: Obtain the current charging load data of the target charging station, perform data augmentation processing, and use the charging load adjustable potential prediction model to obtain the charging load adjustable potential prediction result. The specific implementation process is the same as the aforementioned embodiment, and will not be repeated here.

[0054] This embodiment also provides an electric vehicle charging load adjustable potential prediction device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the terms "module" and "device" can refer to a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0055] According to an embodiment of the present invention, an apparatus embodiment for implementing the above-described method for predicting the adjustable potential of electric vehicle charging load is also provided. Figure 4 This is a schematic diagram of the structure of an electric vehicle charging load adjustable potential prediction device according to an embodiment of the present invention, as shown below. Figure 4 As shown, the above-mentioned electric vehicle charging load adjustable potential prediction device includes: a current charging load data acquisition module 400, a first charging load data acquisition module 402, and a charging load adjustable potential prediction module 404, wherein:

[0056] The current charging load data acquisition module 400 is used to acquire the current charging load data of the target charging station, wherein the current charging load data is the charging load data of the target charging station in the current time period;

[0057] The first charging load data acquisition module 402 is connected to the current charging load data acquisition module 400 and is used to perform data enhancement processing on the current charging load data to obtain the first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period.

[0058] The charging load adjustable potential prediction module 404 is connected to the first charging load data acquisition module 402. It is used to obtain the charging load adjustable potential prediction result of the target charging station in the prediction period based on the first charging load data and the charging load adjustable potential prediction model. The charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical charging load adjustable potential. The prediction period is the first predetermined time period after the current time period.

[0059] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0060] It should be noted that the aforementioned current charging load data acquisition module 400, first charging load data acquisition module 402, and charging load adjustable potential prediction module 404 correspond to steps S102 to S106 in the embodiments. The instances and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run in a computer terminal.

[0061] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0062] The aforementioned electric vehicle charging load adjustable potential prediction device may further include a processor and a memory. The aforementioned current charging load data acquisition module 400, first charging load data acquisition module 402, and charging load adjustable potential prediction module 404 are all stored in the memory as program modules, and the processor executes the aforementioned program modules stored in the memory to realize the corresponding functions.

[0063] The processor contains a core that retrieves the corresponding program modules from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0064] According to an embodiment of this application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein, when the program runs, it controls the device where the non-volatile storage medium is located to execute any of the electric vehicle charging load adjustable potential prediction methods.

[0065] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals, and the non-volatile storage medium includes stored programs.

[0066] Optionally, during program execution, the device containing the non-volatile storage medium may be controlled to perform the following functions: acquire the current charging load data of the target charging station in the current time period; perform data augmentation processing on the current charging load data to obtain first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period; based on the first charging load data, adopt the charging load adjustable potential prediction model to obtain the charging load adjustable potential prediction result of the target charging station in the predicted time period, wherein the charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical charging load adjustable potential, and the predicted time period is the first predetermined time period after the current time period.

[0067] According to an embodiment of this application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any of the above-described methods for predicting the adjustable potential of electric vehicle charging load.

[0068] According to an embodiment of this application, an embodiment of a computer program product is also provided. Optionally, in this embodiment, the computer program product includes a computer program that, when executed by a processor, implements the steps of any of the above-described electric vehicle charging load adjustable potential prediction methods.

[0069] Optionally, when the above-mentioned computer program product is executed on a data processing device, it is suitable to execute an initialization program with the following method steps: acquiring the current charging load data of the target charging station in the current time period; performing data augmentation processing on the current charging load data to obtain first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period; based on the first charging load data, using a charging load adjustable potential prediction model, obtaining the predicted result of the charging load adjustable potential of the target charging station in the predicted time period, wherein the charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical charging load adjustable potential, and the predicted time period is the first predetermined time period after the current time period.

[0070] This invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: acquiring current charging load data of a target charging station in the current time period; performing data augmentation processing on the current charging load data to obtain first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period; and based on the first charging load data, using a charging load adjustable potential prediction model to obtain a prediction result of the charging load adjustable potential of the target charging station in a predicted time period, wherein the charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical charging load adjustable potential, and the predicted time period is a time period of a first predetermined duration after the current time period.

[0071] The order of the above embodiments of the present invention is merely for description and does not represent the superiority or inferiority of the embodiments.

[0072] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0073] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of modules described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between modules, and may be electrical or other forms.

[0074] The modules described above as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0075] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0076] If the aforementioned integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0077] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting the adjustable charging load potential of electric vehicles, characterized in that, include: Obtain the current charging load data of the target charging station in the current time period; The current charging load data is subjected to data augmentation processing to obtain first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period; Based on the first charging load data, a charging load adjustable potential prediction model is used to obtain the predicted result of the charging load adjustable potential of the target charging station during the predicted period. The charging load adjustable potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data of the target charging station and the corresponding historical charging load adjustable potential. The predicted period is the first predetermined duration after the current period.

2. The method according to claim 1, characterized in that, Before acquiring the current charging load data of the target charging station, the method further includes: Obtain the historical charging load data of the target charging station during a predetermined historical period, wherein the predetermined historical period is a period of a second predetermined duration prior to the current period; The historical charging load data is subjected to data augmentation processing to obtain second charging load data, wherein the second charging load data represents the charging load data containing user behavior preference information corresponding to the historical time period; Construct a bidirectional improved long short-term memory network; Based on the second charging load data, the bidirectional improved long short-term memory network is optimized and trained to obtain the charging load adjustable potential prediction model.

3. The method according to claim 2, characterized in that, The process of performing data augmentation on the historical charging load data to obtain second charging load data includes: Based on preset constraints, the historical charging load data is processed to obtain constrained historical charging load data. A user electricity price sensitivity vector is constructed, wherein the user electricity price sensitivity vector is obtained based on multiple electricity price sensitivity coefficients, which are used to represent the sensitivity of electric vehicle users to changes in electricity prices, and the multiple electricity price sensitivity coefficients correspond to different electricity price ranges; The second charging load data is obtained based on the constrained historical charging load data and the user electricity price sensitivity vector.

4. The method according to claim 3, characterized in that, The constraints include at least the following: The charging power of any electric vehicle charging at the target charging station shall not exceed the maximum charging power of the corresponding electric vehicle. The actual stopping time of any electric vehicle at the target charging station is obtained in the following way: if the interval between two adjacent charging sessions exceeds a preset stopping time, then the actual stopping time is the preset stopping time; otherwise, the actual stopping time is the interval between two adjacent charging sessions. The probability distribution of the preset stopping time satisfies a preset first normal distribution, which is obtained based on the time when any electric vehicle arrives at the target charging station. The probability distribution of the single charging time of any electric vehicle satisfies a preset second normal distribution; The probability distribution of the state of charge of any electric vehicle at the start of charging satisfies a preset third normal distribution, wherein the preset third normal distribution is obtained based on the probability density function of daily mileage.

5. The method according to claim 3, characterized in that, When the plurality of electricity price sensitivity coefficients include low electricity price sensitivity coefficients, medium electricity price sensitivity coefficients, and high electricity price sensitivity coefficients, the construction of the user electricity price sensitivity vector includes: The low electricity price sensitivity coefficient is obtained as follows: ； The electricity price sensitivity coefficient is obtained as follows: ； The high electricity price sensitivity coefficient is obtained as follows: ； Wherein, M1 represents the low electricity price sensitivity coefficient, M2 represents the medium electricity price sensitivity coefficient, M3 represents the high electricity price sensitivity coefficient, P represents the electricity price, A1 represents the first preset electricity price threshold, B1 represents the second preset electricity price threshold, C1 represents the third preset electricity price threshold, D1 represents the fourth preset electricity price threshold, and E1 represents the fifth preset electricity price threshold.

6. The method according to claim 2, characterized in that, The construction of the bidirectional improved long short-term memory network includes: The bidirectional long short-term memory unit was optimized to obtain the optimized bidirectional long short-term memory unit; Based on the optimized bidirectional long short-term memory unit, a first bidirectional improved long short-term memory network layer is constructed. A first linear transformation and a nonlinear transformation are performed on the output vector of the first bidirectional improved long short-term memory network layer to construct a second bidirectional improved long short-term memory network layer, wherein the first linear transformation is used to adjust the dimension of the output vector of the first bidirectional improved long short-term memory network layer. A second linear transformation is performed on the output vector of the second bidirectional improved long short-term memory network layer to construct a third bidirectional improved long short-term memory network layer, wherein the second linear transformation is used to adjust the dimension of the output vector of the second bidirectional improved long short-term memory network layer; The bidirectional improved long short-term memory network is obtained based on the first bidirectional improved long short-term memory network layer, the second bidirectional improved long short-term memory network layer, and the third bidirectional improved long short-term memory network layer.

7. The method according to claim 2, characterized in that, The step of optimizing and training the bidirectional improved long short-term memory network based on the second charging load data to obtain the charging load adjustable potential prediction model includes: The second charging load data is randomly divided according to the daily cycle to obtain the training samples of the bidirectional improved long short-term memory network, wherein the daily cycle represents a time period of one day. Based on the training samples, the bidirectional improved long short-term memory network is optimized and trained in the following manner: The training samples are subjected to multiple random augmentation processes to obtain multiple output samples of the bidirectional improved long short-term memory network; The label weights in the bidirectional improved long short-term memory network are smoothed to obtain updated label weights, wherein the label weights represent parameters used to adjust the difference between the predicted label value and the true label value, and the predicted label value is obtained based on the output sample; Optimize the learning rate in the bidirectional improved long short-term memory network; Repeat the above operation until the predetermined termination condition is met; Based on the optimized bidirectional improved long short-term memory network obtained when the predetermined termination condition is met, the predictive model for the adjustable charging load potential is obtained.

8. A device for predicting the adjustable charging load potential of electric vehicles, characterized in that, include: The current charging load data acquisition module is used to acquire the current charging load data of the target charging station, wherein the current charging load data is the charging load data of the target charging station in the current time period; The first charging load data acquisition module is used to perform data augmentation processing on the current charging load data to obtain first charging load data, wherein the first charging load data represents the charging load data containing user behavior preference information corresponding to the current time period. The adjustable charging load potential prediction module is used to obtain the adjustable charging load potential prediction result of the target charging station during the prediction period based on the first charging load data and the adjustable charging load potential prediction model. The adjustable charging load potential prediction model is obtained by optimizing and training a bidirectional improved long short-term memory network based on the historical charging load data and corresponding historical adjustable charging load potential of the target charging station. The prediction period is the first predetermined duration after the current period.

9. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores multiple instructions adapted for loading and execution by a processor of the electric vehicle charging load adjustable potential prediction method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the electric vehicle charging load adjustable potential prediction method according to any one of claims 1 to 7.

Images