Charging station charging load prediction method, device, computer equipment and storage medium

Abstract

The present application relates to a charging load prediction method, device, computer equipment and storage medium for a charging station. The method comprises: obtaining the charging power of various electric vehicles and the number of electric vehicles charged at the charging station at each moment in a sampling time period, inputting the number of electric vehicles charged at the charging station into the charging load data based on the historical sampling time period of the charging station, training the charging load prediction model by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network, obtaining the predicted value of the number of electric vehicles charged at each moment in the next sampling time period by using the rolling correction method, and determining the predicted value of the charging load of the charging station at each moment in the next sampling time period based on the obtained predicted value of the number of electric vehicles charged and the charging power of the electric vehicles. The use of this method can obtain a more accurate charging load prediction value.

Description

Technical Field

[0001] The present application relates to the technical field of power load forecasting, and in particular to a method, device, computer equipment, storage medium and computer program product for forecasting charging load at a charging station. Background Art

[0002] As environmental and energy issues become increasingly prominent, electric vehicles, as representatives of clean energy vehicles, have developed rapidly in recent years, and charging loads have also initially shown statistical characteristics. With the continuous improvement of power batteries and vehicle technologies, the large-scale application of electric vehicles is expanding. In this context, it is of far-reaching significance to analyze and predict the characteristics of electric vehicle charging loads.

[0003] At present, the existing prediction methods for charging load of charging stations, such as prediction methods based on probability models, time series prediction methods, wavelet analysis prediction, methods based on chaos theory, and neural networks, are all passive predictions that rely on a large amount of historical data and related influencing factors. Since electric vehicles are an emerging technology industry, there is little data on their charging load, and they are highly random in time and space, so the predicted power load data has the problem of low accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a charging station charging load prediction method, device, computer equipment, computer readable storage medium and computer program product that can improve the prediction accuracy in response to the above technical problems.

[0005] In a first aspect, the present application provides a charging load prediction method for a charging station. The method comprises:

[0006] Obtain the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each time during the sampling period;

[0007] The number of electric vehicles charged at the charging station at each moment in the sampling period is input into the trained charging load prediction model according to the type of electric vehicle, and a rolling correction method is used to obtain the predicted value of the number of electric vehicles charged at each moment in the next sampling period;

[0008] Determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the predicted value of the charging quantity of each type of electric vehicles and the charging power of each type of electric vehicles at each moment in the next sampling time period;

[0009] Among them, the charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0010] In one embodiment, obtaining the number of various types of electric vehicles charged at the charging station at each time within the sampling time period includes:

[0011] Obtaining a pre-built charging characteristic behavior model of electric vehicles and the starting charging time and charging duration of various types of electric vehicles at the charging station, wherein the charging behavior characteristic model of electric vehicles is constructed based on daily behavior characteristic data of various types of electric vehicles and historical charging load data of the charging station;

[0012] According to the pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station, the charging quantity of each category of electric vehicles at the charging station at each moment in the sampling time period is determined.

[0013] In one embodiment, the charging load prediction model is trained based on the following method:

[0014] Obtaining a training sample set and an initial charging load prediction model for the charging station within the historical sampling period, wherein the training sample set includes the number of various types of electric vehicles charged at the charging station at each moment within the historical sampling period;

[0015] Inputting the charging quantity of each type of electric vehicles at the charging station at each moment into the initial charging load prediction model according to the type of electric vehicles, and obtaining the predicted value of the charging quantity of each type of electric vehicles at the time to be predicted, where the time to be predicted is the next time of each moment;

[0016] When the predicted time arrives, the actual value of the charging quantity of various electric vehicles at the predicted time is obtained;

[0017] The actual value of the number of electric vehicles charged at the time to be predicted is added to the training sample set to obtain an updated training sample set, and the residual value of the predicted value of the number of electric vehicles charged at the time to be predicted and the actual value of the number of electric vehicles charged at the time to be predicted is obtained;

[0018] Based on the updated training sample set and residual value, the network parameters of the initial charging load prediction model are rolled over and revised to obtain the charging load prediction model, taking the residual value satisfying the preset residual verification condition as the termination condition.

[0019] In one of the embodiments, based on the updated training sample set and residual value, with the residual value satisfying the preset residual verification condition as the termination condition, the network parameters of the initial charging load prediction model are rolled over to obtain the charging load prediction model, including: based on the updated training sample set and residual value, with the residual value satisfying the preset residual verification condition as the termination condition, the network parameters of the initial charging load prediction model are rolled over using a gray prediction model so that the prediction accuracy of the charging load prediction model is greater than or equal to the preset accuracy threshold, thereby obtaining the charging load prediction model.

[0020] In one embodiment, according to the predicted values ​​of the number of electric vehicles charged and the charging power of the electric vehicles at each moment in the next sampling time period, determining the predicted value of the charging load of the charging station at each moment in the next sampling time period includes:

[0021] Multiply the predicted value of the number of electric vehicles charged at each moment in the next sampling period by the charging power of each electric vehicle to obtain the charging load of each electric vehicle at each moment in the next sampling period;

[0022] The charging loads of various electric vehicles at each moment in the next sampling time period are summed up to obtain the charging load prediction value of the charging station at each moment in the next sampling time period.

[0023] In one embodiment, the charging load prediction model is a continuous wavelet neural network.

[0024] In a second aspect, the present application also provides a charging load prediction device for a charging station. The device comprises:

[0025] A data acquisition module is used to obtain the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each moment during the sampling period;

[0026] The charging quantity prediction module is used to input the charging quantity of various electric vehicles at the charging station at each moment in the sampling time period into the trained charging load prediction model according to the category of the electric vehicle, and use a rolling correction method to obtain the predicted value of the charging quantity of various electric vehicles at each moment in the next sampling time period;

[0027] A charging load determination module is used to determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the charging quantity prediction value of each type of electric vehicles and the charging power of each type of electric vehicles at each moment in the next sampling time period;

[0028] Among them, the charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0029] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0030] Obtain the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each time during the sampling period;

[0031] The number of electric vehicles charged at the charging station at each moment in the sampling period is input into the trained charging load prediction model according to the type of electric vehicle, and a rolling correction method is used to obtain the predicted value of the number of electric vehicles charged at each moment in the next sampling period;

[0032] Determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the predicted value of the charging quantity of each type of electric vehicles and the charging power of each type of electric vehicles at each moment in the next sampling time period;

[0033] Among them, the charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0034] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0035] Obtain the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each time during the sampling period;

[0036] The number of electric vehicles charged at the charging station at each moment in the sampling period is input into the trained charging load prediction model according to the type of electric vehicle, and a rolling correction method is used to obtain the predicted value of the number of electric vehicles charged at each moment in the next sampling period;

[0037] Determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the predicted value of the charging quantity of each type of electric vehicles and the charging power of each type of electric vehicles at each moment in the next sampling time period;

[0038] Among them, the charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0039] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0040] Obtain the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each time during the sampling period;

[0041] The number of electric vehicles charged at the charging station at each moment in the sampling period is input into the trained charging load prediction model according to the type of electric vehicle, and a rolling correction method is used to obtain the predicted value of the number of electric vehicles charged at each moment in the next sampling period;

[0042] Determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the predicted value of the charging quantity of each type of electric vehicles and the charging power of each type of electric vehicles at each moment in the next sampling time period;

[0043] Among them, the charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0044] The above-mentioned charging load prediction method, device, computer equipment, storage medium and computer program product for charging stations adopt rolling prediction and residual rolling correction to update the network parameters of the preset neural network, which can train a charging load prediction model with higher prediction accuracy. After the charging quantity of each type of electric vehicles at the charging station at each moment in the sampling time period is input into the trained charging load prediction model according to the category of the electric vehicles, the predicted value is determined by rolling correction of the predicted value to determine the charging quantity of each type of electric vehicles at each moment in the next sampling time period. This can make the charging load predicted based on the predicted value of the charging quantity of each type of electric vehicles at each moment in the next sampling time period and the charging power of each type of electric vehicles more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A diagram of an application environment of a charging load prediction method for a charging station in one embodiment;

[0046] Figure 2 A schematic diagram of a flow chart of a charging load prediction method for a charging station in one embodiment;

[0047] Figure 3A schematic diagram of a flow chart of the steps of training a charging load prediction model in an embodiment;

[0048] Figure 4-1 It is a schematic diagram of a prediction result of the number of online charging vehicles obtained by predicting the number of online charging vehicles at a charging station with one day as a sampling period in one embodiment;

[0049] Figure 4-2 It is a schematic diagram of the relative error result of the number of vehicles on the network obtained by predicting the number of vehicles on the network charging at a charging station with one day as a sampling period in one embodiment;

[0050] Figure 4-3 It is a schematic diagram of the prediction result of the number of vehicles on the network obtained before the rolling prediction correction and the residual rolling correction are applied for predicting the number of vehicles on the network at the charging station at intervals of 15 minutes in another embodiment;

[0051] Figure 4-4 It is a schematic diagram of the relative error result of the number of vehicles on the network before applying the rolling prediction correction and the residual rolling correction for predicting the number of vehicles on the network at the charging station at intervals of 15 minutes in another embodiment;

[0052] Figure 4-5 It is a schematic diagram of the prediction result of the number of vehicles on the network obtained after applying the rolling prediction correction and the residual rolling correction for predicting the number of vehicles on the network at the charging station at intervals of 15 minutes in another embodiment;

[0053] Figure 4-6 It is a schematic diagram of the relative error result of the number of vehicles on the network obtained after applying the rolling prediction correction and the residual rolling correction for predicting the number of vehicles on the network at intervals of 15 minutes in another embodiment;

[0054] Figure 5 It is a structural block diagram of a charging load prediction device for a charging station in one embodiment;

[0055] Figure 6 It is a structural block diagram of a charging load prediction device for a charging station in another embodiment;

[0056] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] The charging load prediction method for a charging station provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. Specifically, a trained charging load prediction model is deployed on the server. The user may upload the historical charging record data of the charging station and the charging power of various electric vehicles to the server 104 through the terminal 102, and then perform corresponding operations on the terminal 102, select the sampling time period, and send the charging load prediction message of the charging station to the server 104 through the terminal 102. The server 104 responds to the message, obtains the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each moment in the sampling time period. Then, the number of various electric vehicles charged at the charging station at each moment in the sampling time period is input into the trained charging load prediction model according to the type of electric vehicle, and a rolling correction method is used to obtain the predicted value of the number of various electric vehicles charged at each moment in the next sampling time period. Finally, according to the predicted value of the number of various electric vehicles charged at each moment in the next sampling time period and the charging power of various electric vehicles, the predicted value of the charging load of the charging station at each moment in the next sampling time period is determined. The charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by updating the network parameters of the preset neural network by rolling prediction and residual rolling correction. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0059] In one embodiment, if Figure 2 As shown, a charging load prediction method for a charging station is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0060] Step 202: Obtain the charging power of each type of electric vehicles and the number of each type of electric vehicles being charged at the charging station at each time within the sampling time period.

[0061] Electric vehicles may include electric vehicles, electric bicycles, electric motorcycles, electric trucks, and the like. In this embodiment, electric vehicles are described by taking electric vehicles as an example. According to the types of electric vehicles charged by the charging station, the electric vehicles in this embodiment may include: pure electric buses, pure electric taxis, new energy logistics vehicles, pure electric sanitation vehicles, new energy private cars, new energy online car-hailing rental vehicles, and pure electric mud trucks. In actual applications, users will select the historical charging record data of the charging station within the sampling time period (such as one day) and use it to predict the charging load of the charging station in the next sampling time period (such as the next day). Among them, the historical charging record data of the charging station includes the number of various types of electric vehicles charged at the charging station at each moment during the sampling time period. In specific implementation, the user can select the charging record data of the charging station in the corresponding sampling time period, and upload the charging power of various types of electric vehicles to the server, send the charging load prediction message to the server through the terminal, and the server responds to the message to obtain the charging power of various types of electric vehicles and the number of various types of electric vehicles charged at the charging station at each moment during the sampling time period.

[0062] In step 204, the number of electric vehicles charged at the charging station at each moment in the sampling time period is input into the trained charging load prediction model according to the category of the electric vehicles, and a rolling correction method is used to obtain a predicted value of the number of electric vehicles charged at each moment in the next sampling time period. The charging load prediction model is based on the charging load data of the charging station in the historical sampling time period, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0063] The charging load prediction model can be a continuous wavelet neural network or a discrete wavelet neural network that uses a fusion of wavelet analysis and neural networks. In this embodiment, a continuous wavelet neural network can be selected as the original charging load prediction model, and then, based on the charging load data of the historical sampling time period of the charging station, the charging load prediction model is trained by updating the network parameters of the continuous wavelet neural network or the discrete wavelet neural network through rolling prediction and residual rolling correction. The trained charging load prediction model can obtain the predicted value of the charging number of various electric vehicles in the next sampling time period based on the number of electric vehicles charged in the input sampling time period (also known as the number of electric vehicles charged on the network) through rolling correction.

[0064] Specifically, based on the existing N-time load history data of a node, each time of the historical data can be taken as the “current time τ”, and the next time of each time can be taken as the “time to be predicted τ+1”, and then the corresponding sampling interval D can be defined. cOn this basis, the number of online charging vehicles of various types of electric vehicles at each moment in the sampling interval is collected as input, and the number of online charging vehicles at the "predicted moment" is used as the true value of the output, so NT can be formed as {{I t |t∈D τ},I τ+1}, thus forming a training data set χ for training the network. When the initial charging load forecasting model is fully trained on the training set consisting of historical data, it can be used for real-time intraday rolling forecasting and online update. For the current time τ0, the charging load power model can predict the number of charging vehicles of various types of electric vehicles on the grid at time τ0+1. When the time τ0+1 comes, the real value of the number of vehicles charging on the grid at the time τ0+1 can be obtained. Thus obtaining a new tuple This tuple is added to the training sample set to form a new training sample set, and the network weight is updated with the new training sample set. In this way, the predicted value of the number of electric vehicles on the grid for each type of electric vehicle at each moment in the next sampling time period is obtained. In this embodiment, considering the computational cost and efficiency, the training data tuple corresponding to the farthest time node in the historical data can be deleted from the training sample.

[0065] Step 206, determining the charging load prediction value of the charging station at each moment in the next sampling time period according to the charging quantity prediction value of each type of electric vehicle and the charging power of each type of electric vehicle at each moment in the next sampling time period.

[0066] After obtaining the predicted value of the charging number of each type of electric vehicles at each moment in the next sampling time period, the charging load of each type of electric vehicle can be obtained based on the predicted value of the charging number of each type of electric vehicle and the charging power of each type of electric vehicle according to the type of electric vehicle, and then the charging load prediction value of the charging station can be obtained based on the charging load of each type of electric vehicle.

[0067] In the above-mentioned charging load prediction method for charging stations, the network parameters of the preset neural network are updated by rolling prediction and residual rolling correction, so that a charging load prediction model with higher prediction accuracy can be trained. After the charging quantity of each type of electric vehicles at the charging station at each moment in the sampling time period is input into the trained charging load prediction model according to the category of the electric vehicles, the predicted value is determined by rolling correction of the predicted value to determine the charging quantity of each type of electric vehicles at each moment in the next sampling time period. This can make the charging load predicted based on the predicted value of the charging quantity of each type of electric vehicles at each moment in the next sampling time period and the charging power of each type of electric vehicles more accurate.

[0068] In one embodiment, the charging load prediction model is a continuous wavelet neural network.

[0069] In this embodiment, the charging load prediction model is still a continuous wavelet neural network (hereinafter referred to as wavelet network), which includes: input layer, hidden layer, and output layer. The error generated during the training and learning process of the network is back-propagated, and the weight w of the network connection and the characteristic parameters a and b of the wavelet analysis are adjusted in the direction of gradient descent. The specific implementation process of the wavelet network is as follows: Assume: x k : Input pattern vector, y i : Output mode vector, w ij : The weight from the middle hidden layer to the output layer, w jk : The weight from the input layer to the middle hidden layer, a j , b j : scaling and translation parameters of the intermediate hidden layer, d i : The expected output corresponding to the output vector mode, m, n, N are the number of neurons in the input layer, hidden layer and output layer respectively, so at time point t:

[0070]

[0071] The error function E is:

[0072]

[0073] If you Then there is a is the scale parameter and b is the translation parameter.

[0074] The activation value of the i-th neuron is:

[0075] The output of the i-th neuron is: Y i =f(net i ) (i＝1,2...m)

[0076] ψ(t) is the wavelet mother function, so:

[0077] According to the idea of ​​gradient descent, the adjustment process of network weights and wavelet feature parameters is as follows:

[0078]

[0079]

[0080]

[0081]

[0082] The η in the formula is the set network learning coefficient. After arbitrarily specifying the weights and thresholds of the network, first forward infer the model network to be predicted, then perform deviation calculation and reverse correction. In this process, formulas (1-1) to (1-6) are required. During the training process of the network, it is continuously self-learning and adjusting the various parameters of the network until the set conditions are met and the training is terminated.

[0083] Regarding the training sample set, in order to ensure a certain degree of accuracy in network mapping, the number of samples required will increase accordingly, but at the same time, the scale of the network will also increase, and the prediction efficiency will decrease. Therefore, the following experience is used: according to the total number of network connection weights, take 5 to 10 times of it as the number of training samples.

[0084] Since the various input factors of the neural network have different physical meanings, different physical dimensions will appear. At the same time, since the neurons of the wavelet neural network use Meyer wavelets, in order to make each input component have the same influence on the network at the beginning of network training, and to avoid the saturation of neuron output due to the absolute value of the net input exceeding the range after transformation, the input data must be scaled and the size of the network's input and output data must be limited between plus and minus 1, that is, data normalization.

[0085] The following transformations are commonly used to convert input and output data to positive values ​​between zero and one:

[0086]

[0087] In the formula, is the normalized data, x i is the input or output data, x min is the minimum value of the data, x max is the maximum value of the data.

[0088] Regarding the number of hidden neurons, in this embodiment, the trial and error method is used to determine the optimal number of hidden neurons. When the trial and error method is used, the approximate number of hidden neurons can be calculated based on some commonly used empirical formulas for determining the number of hidden neurons, and this can be used as the initial value of the trial and error method. The following are several commonly used empirical formulas:

[0089]

[0090] m=2n+1 (1-9)

[0091] m=lg2 n (1-10)

[0092] Among them, m, n, N are the number of neurons in the input layer, hidden layer and output layer respectively, and a is the scale parameter.

[0093] Regarding the initial learning rate, in order to ensure that the convergence of the model is relatively stable and fast, in this embodiment, the learning rate is selected in the range of 0.01 to 0.8, and the learning rate can be increased or decreased according to the actual situation. In this embodiment, by selecting the continuous wavelet neural network as the charging load prediction model, the law of the BP (Back Propagation) neural network is followed, and the organic integration of wavelet analysis and neural network is completed, thereby ensuring the prediction accuracy.

[0094] like Figure 3 As shown, in one embodiment, the charging load prediction model is trained based on the following method:

[0095] Step 100, obtaining a training sample set and an initial charging load prediction model of a charging station within a historical sampling period, wherein the training sample set includes the number of various types of electric vehicles charged at the charging station at each moment within the historical sampling period;

[0096] Step 102, inputting the number of electric vehicles charged at each charging station at each moment into the initial charging load prediction model according to the type of electric vehicles, and obtaining the predicted value of the number of electric vehicles charged at the time to be predicted, where the time to be predicted is the next time of each moment;

[0097] Step 104, when the time to be predicted arrives, obtain the actual value of the charging quantity of each type of electric vehicle at the time to be predicted;

[0098] Step 106, adding the actual value of the number of electric vehicles charged at the time to be predicted to the training sample set to obtain an updated training sample set, and obtaining the residual value between the predicted value of the number of electric vehicles charged at the time to be predicted and the actual value of the number of electric vehicles charged at the time to be predicted;

[0099] Step 108, based on the updated training sample set and residual value, with the residual value satisfying a preset residual check condition as a termination condition, the network parameters of the initial charging load prediction model are subjected to rolling correction to obtain a charging load prediction model.

[0100] In order to improve the prediction accuracy of the model and realize the rolling update of the intraday network weights, in this embodiment, intraday rolling correction and residual rolling correction are adopted, and the accuracy of the model is continuously updated and improved through online learning. Residual correction is a posterior difference test of the prediction accuracy of the model. It is a method of testing based on the statistical situation between the actual value and the model prediction value. Specifically, it can be achieved by dynamically adjusting the time base point to ensure that the model always receives the latest load data, that is, the number of electric vehicles charged on the grid at each moment. If the current moment is defined as τ, then all time nodes contained in the interval [τ-T,τ] can be defined as a set of sampling time intervals D c, where T is the number of sampling time points. By introducing the sampling interval D at the time node τ c To predict the load data of a node at time τ+1, the model uses the load I of the node at time t c is the input, where t∈D c , considering temperature and weather at the same time, the neural network contains T+2 input neurons and 1 output neuron, and the output is the predicted load situation at time τ+1. In the initial stage, the training sample set can be designed for training through the existing N-time load history data of a certain node. Specifically, each moment of the historical data can be taken as the "current moment τ", and the next moment of each moment can be taken as the "time to be predicted τ+1", and then the corresponding sampling interval D can be defined. c On this basis, the load data in the sampling interval (the number of online charging vehicles at each moment, i.e., the charging load of electric vehicles) is collected as input, and the number of online charging vehicles at the "predicted moment" is taken as the true value of the output, so NT can be formed, as shown in {{I t |t∈D τ},I τ+1}, thus forming a training sample set χ for training the model. That is, for the current time τ0, the model can predict the load situation at time τ0+1 When the time τ0+1 comes, the real value of the number of vehicles charging on the grid at the time τ0+1 can be obtained. Thus obtaining a new tuple This tuple is added to the training set to form a new training sample set. At the same time, the residual of the predicted value of the number of vehicles charging on the grid and the actual value of the number of vehicles charging on the grid are obtained. Based on the residual ε between the actual value and the model prediction value, the probability of the occurrence of the residual smaller point and the size of the index related to the variance of the prediction error are determined by analyzing the absolute value of the residual. The specific method of residual correction is as follows:

[0101] Let x (0) is the actual value sequence, x (0) ={x (0) (1),x (0) (2),...,x (0) (n)}. is the predicted value sequence,

[0102] Then the actual value x at time k is (0) (k) and predicted value The difference is ε(k), which is called the residual at time k, that is:

[0103]

[0104] Actual value x (0)The average value of (k)(k=1,2,...,n) is Right now:

[0105]

[0106] The average value of the residual ε(k) (k=1,2,...,n) is but:

[0107]

[0108] In the above formula, k is the number of prediction residual data, usually m≤n.

[0109] The actual value variance is but:

[0110]

[0111] The residual variance is but:

[0112]

[0113] From the above, we can get two very important data indicators of the posterior difference test, namely the small error probability p and the posterior difference ratio C, whose values ​​are:

[0114]

[0115]

[0116] In the formula, C and p are data indicators, and the value range of indicators C and p is C>0, 0≤p≤1. p represents the probability that the absolute value of the difference between the residual and the residual mean is less than a given threshold, and C represents the degree of dispersion between the actual value and the predicted value. Among them, the smaller the indicator C, the better. Since the size of S1 represents the size of the actual data variance, that is, the size of the actual data dispersion; the size of S2 indicates the size of the residual variance. If it is larger, the residual dispersion is greater. Whether the accuracy of the model meets the test requirements is judged by whether the small error probability P and the posterior difference ratio C meet the preset threshold. In this embodiment, by performing a posterior difference test on the prediction results, the parameters of the charging load prediction model are rolled according to the residual values ​​obtained, so that the entire network can be continuously updated with the update of time to ensure the accuracy of its intraday rolling prediction.

[0117] In one of the embodiments, based on the updated training sample set and residual value, with the residual value satisfying the preset residual verification condition as the termination condition, the network parameters of the initial charging load prediction model are rolled over to obtain the charging load prediction model, including: based on the updated training sample set and residual value, with the residual value satisfying the preset residual verification condition as the termination condition, the network parameters of the initial charging load prediction model are rolled over using a gray prediction model so that the prediction accuracy of the charging load prediction model is greater than or equal to the preset accuracy threshold, thereby obtaining the charging load prediction model.

[0118] Normally, the charging load prediction model must be tested and qualified before it can be used for prediction. If the charging load prediction model fails the test, the residual gray prediction model can be used for correction. Specifically, the residual rolling correction process can be based on the residual of the predicted value and the true value of the number of charging vehicles on the grid, perform a posteriori difference test, calculate the small error probability P and the posteriori difference ratio C, and judge whether P and C meet the preset small error probability threshold and the posteriori difference ratio threshold. If not, the residual test fails, and returns to the step of rolling prediction to obtain the predicted value of the number of charging vehicles on the grid until the residual test passes. If the residual test passes, the network parameters of the model are updated using the small error probability P and the posteriori difference ratio C, so that the prediction accuracy of the charging load prediction model is greater than or equal to the preset accuracy threshold, and the charging load prediction model is obtained to improve the prediction accuracy of the model.

[0119] In one embodiment, obtaining the number of electric vehicles of each category charged at the charging station at each moment in the sampling time period includes: obtaining a pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station, the electric vehicle charging behavior characteristic model being constructed based on daily behavior characteristic data of each category of electric vehicles and historical charging load data of the charging station; determining the number of electric vehicles of each category charged at the charging station at each moment in the sampling time period based on the pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station.

[0120] In practical applications, the charging characteristics of electric vehicles are affected by a variety of factors, which are related to the type of electric vehicle, charging power, charging time, and charge state of the electric vehicle. Overall, they present a relatively rich set of characteristics, which also makes it difficult to model the charging characteristics of electric vehicles. This application obtains the relevant characteristics of electric vehicles and their charging model parameters through the analysis of survey data, and establishes a charging characteristic behavior model for electric vehicles.

[0121] Specifically, the survey data uses the electric vehicle related data of XX District, XX City as the research basis. By processing and analyzing the daily behavior characteristics data of electric vehicles and the historical data of the electric vehicle station in XX District, the electric vehicle related information shown in Table 1 and Table 2 is obtained. As shown in Table 1, electric vehicle vehicle types are mainly divided into 7 types of electric vehicles. Among them, the operating characteristics of new energy private cars are not fixed compared with other types of electric vehicles, so the centralized charging time is also not fixed. For other types of electric vehicles, they all have fixed operating characteristics, resulting in a relatively concentrated charging period. In addition, for pure electric buses, pure electric taxis and new energy logistics vehicles, they all have specific operating periods and centralized charging periods, which makes it easier to establish their charging behavior models.

[0122] As can be seen from Table 2, the probability distribution functions of various factors affecting the charging characteristics of electric vehicles are as follows:

[0123] a) Types of electric vehicles

[0124] The initial charging time period, daily mileage, and charging load power of electric vehicles vary greatly depending on the type of vehicle. Therefore, when predicting the load power of a charging station for charging vehicles, it is first necessary to classify the types of vehicles entering the station.

[0125] As shown in Table 2, pure electric buses, pure electric taxis, and new energy logistics vehicles are characterized by relatively fixed driving trajectories and relatively concentrated charging times, so the load position generated by them on charging stations is relatively stable. However, the use of new energy private cars is more flexible, and their driving trajectories and charging time requirements show a random distribution probability. Therefore, the impact of different types of electric vehicles on the load of charging stations needs to be discussed in categories.

[0126] b) Charging power of electric vehicles

[0127] When a battery is charged and discharged at a certain current, a curve is used to represent the change of the battery terminal voltage over time. These curves are called the characteristic curves of the battery and are used to represent the various characteristics of the battery.

[0128] When charging, the terminal voltage V of the battery consists of two parts: a slow rise and a sharp rise. At the beginning of charging, the terminal voltage rises rapidly. As time goes on, the voltage rise slows down. The middle section (SOC is between 20% and 80%) can even be abstracted as a horizontal straight line, that is, constant power charging. It can be seen that the charging characteristics can be approximated by a three-segment function. Generally, in the early stage of charging, the charging power gradually increases, reaches a peak, and then gradually decreases at a large amplitude. In order to simplify the model, the constant current stage is set as a linear function, and the constant voltage stage is set as an exponential function. The model of power changing with time is as follows:

[0129]

[0130] c) State of charge of electric vehicles

[0131] In practical applications, the charging time on the same charging equipment varies according to the different remaining power of the battery of the electric vehicle entering the station, that is, the state of charge of the electric vehicle. The lower the state of charge, the more charging load is required and the longer the charging time is. Using the law of large numbers and the central limit theorem, the historical data of the state of charge of electric vehicles are analyzed, and the probability statistics results are approximately normal distribution, as shown in Table 2.

[0132] Table 1 Operational characteristics and charging characteristics of different types of electric vehicles

[0133]

[0134]

[0135] As can be seen from the above, the charging characteristic behavior model of electric vehicles can obtain the charging characteristic data of electric vehicles, including the operating characteristics and charging characteristics of different types of electric vehicles and the charging model parameters of different types of electric vehicles, and combined with the starting charging time and charging time of each type of electric vehicle, the number of electric vehicles charging on the grid at each moment can be obtained, that is, the charging load of electric vehicles. For example, if there are 10 electric vehicles charging at irregular intervals every day, assuming that they are all fully charged in 1 hour, the number of vehicles being charged at any time of the day can be obtained by looking up the table.

[0136] Table 2 Charging model parameters for different types of electric vehicles

[0137]

[0138] In this embodiment, by establishing a charging characteristic behavior model of electric vehicles, the charging time of various types of electric vehicles can be quickly obtained, and then the starting charging time and charging time of each type of electric vehicle can be converted into the number of online charging vehicles at each moment.

[0139] In one embodiment, step 206 includes: multiplying the predicted value of the number of electric vehicles charged at each moment in the next sampling time period by the charging power of each electric vehicle to obtain the charging load of each electric vehicle at each moment in the next sampling time period, and summing the charging load of each electric vehicle at each moment in the next sampling time period to obtain the predicted value of the charging load of the charging station at each moment in the next sampling time period.

[0140] In specific implementation, the charging load of the charging station can be determined by: for each type of electric vehicle, multiplying the charging power of the electric vehicle under this type by the number of online charging vehicles at each moment to obtain the charging load of each type of electric vehicle, and then summing the charging loads of various types of electric vehicles at each moment in the next sampling time period to obtain the charging load prediction value of the charging station at each moment in the next sampling time period. In this embodiment, the category of electric vehicles is a statistical unit, and the charging power of electric vehicles under this category is multiplied by the number of online charging vehicles at each moment to obtain the charging load of each type of electric vehicle, and then the charging load prediction value of the charging station can be obtained in an orderly manner.

[0141] In order to clearly illustrate the charging load prediction method for a charging station provided by the present application, an example is provided below:

[0142] Specifically, the historical data of electric vehicle stations in XX District, XX City is used as an example for simulation analysis. In order to reduce the influence of other factors, the method of predicting the next day by the previous day is used to predict the charging load at each point in 24 hours of a day. First, the historical charging records of 42kw charging piles on any day are selected to classify electric vehicles with charging history records. Electric vehicles of specific categories are selected, and the charging power of each type of electric vehicle is obtained through the charging characteristic behavior model of electric vehicle. Combined with the starting charging time and charging time of the electric vehicles of this category, the number of online charging vehicles at each moment of the day is obtained. Then, the number of online charging electric vehicles at each moment of the day is input into the charging load prediction model, and the rolling correction method is used to obtain the predicted value of the charging number of various types of electric vehicles every hour in the next day. Then, for each type of electric vehicle, the charging power of the electric vehicle under this category is multiplied by the number of online charging vehicles at each moment to obtain the charging load of each category of electric vehicles. Then, the charging load of each type of electric vehicle is summed up every hour in the next day to obtain the charging load prediction value of the charging station every hour in the next day. Specifically, the prediction results can be obtained as follows Figure 4-1 and Figure 4-2 As shown. Figure 4-1 and Figure 4-2It can be seen that the charging load prediction method for charging stations provided by the present application has a high accuracy in the number of electric vehicles charging online, and the error at each point in 24 hours a day is less than 8%, which can accurately describe the charging load characteristics of electric vehicle charging stations. Compared with the previous probabilistic modeling method, this method is faster, simpler and more accurate. In order to fully demonstrate the effect of rolling prediction and residual rolling correction, the number of online charging vehicles at charging stations is predicted at intervals of 15 minutes, and the two prediction results before and after the application of rolling prediction correction and residual rolling correction are compared, as shown in the figure below. Figure 4-3 to Figure 4-6 As shown, Figure 4-3 and Figure 4-4 The forecast result obtained before the rolling forecast correction and residual rolling correction is adopted. Figure 4-5 and Figure 4-6 This is the prediction result obtained after rolling prediction correction and residual rolling correction. Through the comparison of the above results, it is known that due to the small base number of electric vehicles on the grid at intervals of 15 minutes, the relative error is large. The number of vehicles on the grid predicted without rolling correction has roughly the same trend, but the relative error is around 20%, and the prediction effect is poor. However, the relative error of the number of electric vehicles on the grid after rolling correction is significantly reduced compared with that before rolling correction, both below 8%, which fully demonstrates that the use of neural networks combined with rolling correction to predict electric vehicle charging load is very practical. At the same time, it reflects that the rolling correction method has practical application value in reducing errors.

[0143] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0144] Based on the same inventive concept, the embodiment of the present application also provides a charging station charging load prediction device for implementing the charging station charging load prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more charging station charging load prediction device embodiments provided below can refer to the limitations of the charging station charging load prediction method above, and will not be repeated here.

[0145] In one embodiment, if Figure 5 As shown, a charging load prediction device for a charging station is provided, comprising: a data acquisition module 510, a charging quantity prediction module 520 and a charging load determination module 530, wherein:

[0146] The data acquisition module 510 is used to acquire the charging power of various electric vehicles and the number of various electric vehicles charged at the charging station at each time within the sampling time period.

[0147] The charging quantity prediction module 520 is used to input the charging quantity of each type of electric vehicle at the charging station at each moment in the sampling time period into the trained charging load prediction model according to the category of the electric vehicle, and use a rolling correction method to obtain the predicted value of the charging quantity of each type of electric vehicle at each moment in the next sampling time period.

[0148] The charging load determination module 530 is used to determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the charging quantity prediction value of each type of electric vehicle and the charging power of each type of electric vehicle at each moment in the next sampling time period;

[0149] Among them, the charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network.

[0150] In the above-mentioned charging load prediction device of the charging station, the method of updating the network parameters of the preset neural network by rolling prediction and residual rolling correction is adopted, so that a charging load prediction model with higher prediction accuracy can be trained, and after the charging quantity of each type of electric vehicles at the charging station at each moment in the sampling time period is input into the trained charging load prediction model according to the category of the electric vehicles, the predicted value is determined by rolling correction of the predicted value to determine the charging quantity of each type of electric vehicles at each moment in the next sampling time period, which can make the charging load predicted according to the predicted value of the charging quantity of each type of electric vehicles at each moment in the next sampling time period and the charging power of each type of electric vehicles more accurate.

[0151] In one of the embodiments, the data acquisition module 510 is also used to obtain a pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station, and determine the charging quantity of each category of electric vehicles at the charging station at each moment during the sampling time period based on the pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station. The electric vehicle charging behavior characteristic model is constructed based on the daily behavior characteristic data of each category of electric vehicles and the historical charging load data of the charging station.

[0152] like Figure 6 As shown, in one embodiment, the device also includes a model training module 540, which is used to obtain a training sample set and an initial charging load prediction model of the charging station within a historical sampling time period, the training sample set including the number of each type of electric vehicle charged at the charging station at each moment within the historical sampling time period; the number of each type of electric vehicle charged at the charging station at each moment is input into the initial charging load prediction model according to the category of the electric vehicle, and the predicted value of the number of each type of electric vehicle charged at the time to be predicted is obtained, and the time to be predicted is the next moment of each moment; when the time to be predicted is reached, the actual value of the number of each type of electric vehicle charged at the time to be predicted is obtained; the actual value of the number of each type of electric vehicle charged at the time to be predicted is added to the training sample set to obtain an updated training sample set, and the residual value of the predicted value of the number of each type of electric vehicle charged at the time to be predicted and the actual value of the number of each type of electric vehicle charged at the time to be predicted is obtained; based on the updated training sample set and the residual value, the network parameters of the initial charging load prediction model are rolled corrected with the residual value satisfying the preset residual verification condition as the termination condition to obtain the charging load prediction model.

[0153] In one embodiment, the model training module 540 is also used to perform rolling corrections on the network parameters of the initial charging load prediction model using a gray prediction model based on the updated training sample set and residual value, with the residual value satisfying a preset residual verification condition as a termination condition, so that the prediction accuracy of the charging load prediction model is greater than or equal to a preset accuracy threshold, thereby obtaining a charging load prediction model.

[0154] In one of the embodiments, the data acquisition module 510 is also used to multiply the predicted value of the charging quantity of each type of electric vehicle at each moment in the next sampling time period by the charging power of each type of electric vehicle to obtain the charging load of each type of electric vehicle at each moment in the next sampling time period, and sum the charging load of each type of electric vehicle at each moment in the next sampling time period to obtain the predicted value of the charging load of the charging station at each moment in the next sampling time period.

[0155] Each module in the above-mentioned charging station charging load prediction device can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0156] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory and a network interface 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as historical charging records of charging stations and charging characteristic behavior models of electric vehicles. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a charging load prediction method for a charging station is implemented.

[0157] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In one of the embodiments, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned charging station charging load prediction method when executing the computer program.

[0159] In one of the embodiments, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned charging station charging load prediction method are implemented.

[0160] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above-mentioned charging station charging load prediction method when executed by a processor.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0162] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A charging load prediction method for a charging station, characterized in that: The method comprises: Obtaining a pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station, and determining the number of each category of electric vehicles charged at the charging station at each time within a sampling time period according to the pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station; The number of electric vehicles charged at the charging station at each moment in the sampling time period is input into the trained charging load prediction model according to the type of electric vehicle, and a predicted value of the number of electric vehicles charged at each moment in the next sampling time period is obtained by rolling correction; Determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the predicted value of the number of electric vehicles charged at each moment in the next sampling time period and the charging power of the electric vehicles; Among them, the electric vehicle charging behavior characteristic model is constructed based on the daily behavior characteristic data of various types of electric vehicles and the historical charging load data of charging stations. The charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network. The charging load prediction model is a continuous wavelet neural network.

2. The charging load prediction method for a charging station according to claim 1, characterized in that: The charging load prediction model is trained based on the following method: Obtaining a training sample set and an initial charging load prediction model for the charging station within a historical sampling period, wherein the training sample set includes the number of various types of electric vehicles charged at the charging station at each moment within the historical sampling period; Inputting the number of electric vehicles charged at the charging station at each moment into the initial charging load prediction model according to the category of the electric vehicles, and obtaining a predicted value of the number of electric vehicles charged at a time to be predicted, where the time to be predicted is the next time of each moment; When the predicted time is reached, obtaining the actual value of the number of electric vehicles charged at the predicted time; Adding the actual value of the number of electric vehicles charged at the time to be predicted to the training sample set to obtain an updated training sample set, and obtaining the residual value between the predicted value of the number of electric vehicles charged at the time to be predicted and the actual value of the number of electric vehicles charged at the time to be predicted; Based on the updated training sample set and the residual value, the network parameters of the initial charging load prediction model are rolled over and corrected with the residual value satisfying a preset residual verification condition as a termination condition to obtain a charging load prediction model.

3. The charging load prediction method for a charging station according to claim 2, characterized in that: Based on the updated training sample set and the residual value, taking the residual value satisfying a preset residual check condition as a termination condition, rolling correction is performed on the network parameters of the initial charging load prediction model to obtain the charging load prediction model, including: Based on the updated training sample set and the residual value, with the residual value satisfying a preset residual verification condition as the termination condition, a grey prediction model is used to perform rolling correction on the network parameters of the initial charging load prediction model so that the prediction accuracy of the charging load prediction model is greater than or equal to the preset accuracy threshold, thereby obtaining a charging load prediction model.

4. The charging load prediction method for a charging station according to any one of claims 1 to 3, characterized in that: Determining the charging load prediction value of the charging station at each moment in the next sampling time period according to the charging quantity prediction value of each type of electric vehicles and the charging power of each type of electric vehicles at each moment in the next sampling time period includes: Multiply the predicted value of the number of electric vehicles charged at each moment in the next sampling time period by the charging power of each electric vehicle to obtain the charging load of each electric vehicle at each moment in the next sampling time period; The charging loads of various electric vehicles at each moment in the next sampling time period are summed to obtain the charging load prediction value of the charging station at each moment in the next sampling time period.

5. A charging load prediction device for a charging station, characterized in that: The device comprises: A data acquisition module, used to obtain a pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station, and determine the number of each category of electric vehicles charged at the charging station at each time within a sampling time period according to the pre-constructed electric vehicle charging characteristic behavior model and the starting charging time and charging duration of each category of electric vehicles at the charging station; A charging quantity prediction module is used to input the charging quantity of each type of electric vehicle at the charging station at each moment in the sampling time period into the trained charging load prediction model according to the type of electric vehicle, and use a rolling correction method to obtain the predicted value of the charging quantity of each type of electric vehicle at each moment in the next sampling time period; A charging load determination module, used to determine the charging load prediction value of the charging station at each moment in the next sampling time period according to the charging quantity prediction value of each type of electric vehicles at each moment in the next sampling time period and the charging power of the various types of electric vehicles; Among them, the electric vehicle charging behavior characteristic model is constructed based on the daily behavior characteristic data of various types of electric vehicles and the historical charging load data of charging stations. The charging load prediction model is based on the charging load data of the historical sampling time period of the charging station, and is trained by using rolling prediction and residual rolling correction to update the network parameters of the preset neural network. The charging load prediction model is a continuous wavelet neural network.

6. The device according to claim 5, characterized in that The device also includes a model training module, which is used to obtain a training sample set and an initial charging load prediction model of the charging station within a historical sampling time period, wherein the training sample set includes the number of each type of electric vehicles charged at the charging station at each moment within the historical sampling time period, and the number of each type of electric vehicles charged at the charging station at each moment is input into the initial charging load prediction model according to the category of the electric vehicles to obtain a predicted value of the number of each type of electric vehicles charged at a time to be predicted, wherein the time to be predicted is the next moment of each moment, and when the time to be predicted is reached, the actual value of the number of each type of electric vehicles charged at the time to be predicted is obtained, and the actual value of the number of each type of electric vehicles charged at the time to be predicted is added to the training sample set to obtain an updated training sample set, and the residual value of the predicted value of the number of each type of electric vehicles charged at the time to be predicted and the actual value of the number of each type of electric vehicles charged at the time to be predicted is obtained, and based on the updated training sample set and the residual value, the network parameters of the initial charging load prediction model are rolled corrected with the residual value satisfying a preset residual check condition as a termination condition to obtain a charging load prediction model.

7. The device according to claim 6, characterized in that The model training module is also used to perform rolling correction on the network parameters of the initial charging load prediction model using a grey prediction model based on the updated training sample set and the residual value, with the residual value satisfying a preset residual verification condition as a termination condition, so that the prediction accuracy of the charging load prediction model is greater than or equal to the preset accuracy threshold, thereby obtaining a charging load prediction model.

8. 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 4 are implemented.

9. 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 4 are implemented.

10. A computer program product, comprising a computer program, 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 6 are implemented.

Images