Charging quantity prediction method and device

Abstract

The invention relates to the technical field of charge quantity prediction, and provides a charge quantity prediction method and device. The method comprises the following steps: acquiring a plurality of characteristics which influence the charging amount per hour after the charging cost in a target charging station is adjusted; inputting the historical data of the plurality of features into a gradient boosting tree model, and based on an output result of the gradient boosting tree model, screening out a plurality of target features with relatively large influence degrees; inputting the current data of the plurality of target features into a charging amount prediction model to obtain an average charging amount of a target hour in a week after current price adjustment; the charging quantity prediction model is obtained through training by using historical data of a plurality of target features and the average charging quantity of a target hour in a week after historical price adjustment on the basis of a gradient boosting tree model. On one hand, the contribution of historical data rules and target features to the prediction result can be fully mined, so that the prediction accuracy is improved, and on the other hand, a mechanical means can be adopted to replace artificial experience prediction, so that the prediction efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of charge capacity prediction, and in particular to a charge capacity prediction method and device. Background Art

[0002] With the rapid development of the electric vehicle industry, consumers' demand for electric vehicle charging is also growing. For service providers that provide charging services, the charging fees that consumers need to pay when charging at the charging stations they cover include the basic electricity price and the service fee of the service provider. In order to maximize profits, service providers can adjust the price of their controllable service fees, thereby adjusting the price of charging fees. However, the adjustment of charging fees will affect the amount of charging by consumers at charging stations, which will further affect the profits of service providers.

[0003] Since the service provider's revenue is closely related to the charging costs and charging amount of consumers at charging stations, and there is also a mutual influence between the charging costs and the charging amount, the service provider needs to predict the charging amount after the price adjustment before adjusting the service fee, so as to comprehensively evaluate the overall impact of changes in charging costs and charging amount on its own revenue, and thus formulate a reasonable pricing strategy to maximize its own revenue while meeting the charging demand.

[0004] However, the current prediction of charging volume after price adjustment mainly relies on the experience of service operators, and its prediction accuracy and efficiency are relatively low. Summary of the invention

[0005] The embodiments of the present application provide a charging capacity prediction method and device to solve the technical problem that the current prediction of charging capacity after price adjustment mainly relies on the experience of service operators, and the prediction accuracy and efficiency are low.

[0006] In a first aspect, an embodiment of the present application provides a charging capacity prediction method, comprising: Obtain multiple features that affect the hourly charging volume after the charging fee is adjusted in the target charging station; Inputting the historical data of the multiple features into a gradient boosting tree model, and screening out multiple target features with greater influence based on the output results of the gradient boosting tree model; Inputting the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

[0007] In one embodiment, the historical data of the multiple features are input into a gradient boosting tree model, and based on the output result of the gradient boosting tree model, multiple target features with greater influence are screened out, including: Inputting the historical data of the plurality of features into the gradient boosting tree model, and obtaining a first historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Randomly shuffle historical data of candidate features among the multiple features to obtain shuffled historical data of the multiple features; the candidate feature is any feature among the multiple features; Inputting the historical shuffled data into the gradient boosting tree model, and obtaining a second historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Based on the first historical prediction value and the second historical prediction value, calculating the prediction impact value of the candidate feature on the gradient boosting tree model; Restoring the historical data of the candidate feature, selecting another feature from the multiple features as the candidate feature, and returning to the step of randomly shuffling the historical data of the candidate feature from the multiple features to obtain the historical shuffled data of the multiple features, until obtaining the predicted influence value of each feature from the multiple features for the gradient boosting tree model; The predicted influence values ​​of each feature on the gradient boosting tree model are sorted from large to small, and the features corresponding to the top-ranked predicted influence values ​​are determined as target features.

[0008] In one embodiment, the calculating the prediction impact value of the candidate feature on the gradient boosting tree model based on the first historical prediction value and the second historical prediction value includes: Calculating an absolute value of a difference between the first historical prediction value and the second historical prediction value; Determine the absolute value of the difference as the predicted impact value of the candidate feature on the gradient boosting tree model; or Calculating a difference between the first historical prediction value and the second historical prediction value; Calculating an absolute value of a ratio between the difference and the first historical prediction value; The absolute value of the ratio is determined as the prediction influence value of the candidate feature on the gradient boosting tree model.

[0009] In one embodiment, the charge capacity prediction model is obtained based on the following steps: Inputting the historical data of the multiple target features and the average charging amount of the target hour one week after the historical price adjustment into the gradient boosting tree model, and obtaining a third historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Calculating the mean absolute percentage error between the third historical predicted value and its corresponding historical actual value to obtain the prediction error of the gradient boosting tree model; Based on the prediction error, calculating the prediction accuracy of the gradient boosting tree model; Correcting the prediction accuracy to obtain a corrected accuracy; If the revised accuracy is less than the accuracy threshold, then after adjusting the learning rate of the gradient boosting tree model, return to the step of inputting the historical data of the multiple target features and the average charging amount of the target hours one week after the historical price adjustment into the gradient boosting tree model, and obtain the third historical prediction value of the average charging amount of the target hours one week after the historical price adjustment output by the gradient boosting tree model, until the revised accuracy is greater than or equal to the accuracy threshold, and the gradient boosting tree model at this time is determined as the charging amount prediction model.

[0010] In one embodiment, after correcting the prediction accuracy to obtain the corrected accuracy, the method further includes: If the corrected accuracy is less than the accuracy threshold, then after adjusting the learning rate of the gradient boosting tree model, return to the step of inputting the historical data of the multiple features into the gradient boosting tree model, and based on the output results of the gradient boosting tree model, screen out multiple target features with greater influence until the corrected accuracy is greater than or equal to the accuracy threshold, and determine the gradient boosting tree model at this time as the charging capacity prediction model.

[0011] In one embodiment, the correcting the prediction accuracy to obtain a corrected accuracy includes: Setting a cutoff value of the prediction error based on the hourly charging capacity characteristics of the target charging station; Correcting the prediction error based on the cutoff value to obtain a corrected error; The prediction accuracy is corrected based on the correction error to obtain a corrected accuracy.

[0012] In a second aspect, an embodiment of the present application provides a charging capacity prediction device, comprising: The hourly charging capacity impact feature acquisition module is used to: acquire multiple features that affect the hourly charging capacity after the charging fee price adjustment in the target charging station; An hourly charging target feature screening module is used to: input the historical data of the multiple features into a gradient boosting tree model, and screen out multiple target features with greater influence based on the output results of the gradient boosting tree model; An hourly charging capacity prediction module is used to: input the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory storing a computer program, wherein the processor implements the steps of the charge capacity prediction method described in the first aspect when executing the program.

[0014] In a fourth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the charge capacity prediction method described in the first aspect.

[0015] In a fifth aspect, an embodiment of the present application provides a non-transitory computer-readable storage medium, including a computer program, which, when executed by a processor, implements the steps of the charge capacity prediction method described in the first aspect.

[0016] The charging capacity prediction method and device provided in the present application obtain multiple features that affect the hourly charging capacity after the charging fee price adjustment in the target charging station, input the historical data of the multiple features into the gradient boosting tree model, and based on the output result of the gradient boosting tree model, screen out multiple target features with a greater degree of influence, input the current data of the multiple target features into the charging capacity prediction model, and obtain the average charging capacity of the target hour one week after the current price adjustment output by the charging capacity prediction model. The present application first uses the gradient boosting tree model to screen the target features with a greater degree of influence on the hourly charging capacity after the price adjustment, and then uses the charging capacity prediction model based on these target features to predict the hourly charging capacity after the price adjustment. Since the charging capacity prediction model is obtained by training the gradient boosting tree model using the historical data of multiple target features and the average charging capacity of the target hour one week after the historical price adjustment, the present application uses the gradient boosting tree model to perform preliminary screening of the target features, and then predicts the hourly charging capacity based on the charging capacity prediction model with a high degree of adaptability to the target features. On the one hand, it can fully explore the historical data rules and the contribution of the target features to the prediction results, thereby improving the prediction accuracy. On the other hand, it can use mechanized means to replace manual experience prediction and improve prediction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 This is one of the flow charts of the charging capacity prediction method provided in the embodiment of the present application; Figure 2 This is the second flow chart of the charging capacity prediction method provided in the embodiment of the present application; Figure 3 This is the third flow chart of the charging capacity prediction method provided in the embodiment of the present application; Figure 4 This is the fourth flow chart of the charging capacity prediction method provided in the embodiment of the present application; Figure 5 is a schematic diagram of the structure of a charging capacity prediction device provided in an embodiment of the present application; Figure 6 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] It should be noted that in the description of the embodiments of the present application, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements. The orientation or position relationship indicated by the terms "upper", "lower" and the like is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. Unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be a connection between the two elements. For those skilled in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0021] The terms "first", "second", etc. in this application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0022] Figure 1 This is one of the flow charts of the charging capacity prediction method provided in the embodiment of the present application. Figure 1 , the embodiment of the present application provides a charging capacity prediction method, which may include: 101. Obtain multiple features that affect the hourly charging volume after the charging fee is adjusted in the target charging station; 102. Input historical data of multiple features into the gradient boosting tree model, and screen out multiple target features with greater influence based on the output results of the gradient boosting tree model; 103. Input the current data of multiple target features into the charging capacity prediction model to obtain the average charging capacity of the target hours in one week after the current price adjustment output by the charging capacity prediction model.

[0023] The charging capacity prediction model is trained on the basis of the gradient boosting tree model using historical data of multiple target features and the average charging capacity of the target hours in one week after the historical price adjustment.

[0024] In step 101, the charging fee includes a basic electricity price and a service fee, wherein the basic electricity price is set by the power supply unit, and each day is divided into peak hours, peak hours, parity hours and valley hours of electricity consumption according to hours, and different basic electricity prices are set for each time period. On the one hand, different regions have different divisions of different time periods, and the divisions of different time periods will also change with factors such as seasonal changes. On the other hand, there are differences in the basic electricity prices for different time periods in the same region and for the same time periods in different regions. Therefore, the data of dynamic features in multiple features in this application, such as charging volume, charging costs and other data, are adapted to use aggregated data within hours. Since the dynamic feature data on the charging order is usually across hours, these data can be split by minutes and aggregated into different hours to obtain aggregated data within hours. The use of these hourly aggregated data can maximize the realization of detailed quantitative predictions while ensuring the comparability of data during changes.

[0025] Specifically, assuming that you want to adjust the charging fee from 20:00 to 21:00 at the target charging station, the multiple features that affect the charging amount from 20:00 to 21:00 after the price adjustment may include the charging amount from 20:00 to 21:00 every day in the week before the price adjustment, that is, the 7 days before the price adjustment, the charging fee from 20:00 to 21:00 before and after the price adjustment, the time characteristics from 20:00 to 21:00 before and after the price adjustment (for example, which month it belongs to, which week of a month, which day of the week, which hour of a day, etc.), the charging amount in the adjacent hours from 20:00 to 21:00 before the price adjustment (for example, the charging amount from 19:00 to 20:00 and the charging amount from 21:00 to 22:00), the latitude and longitude coordinates of the target charging station, the number of charging piles in the target charging station, the number of charging guns in the target charging station, and the charging fee and charging amount of other charging stations from 20:00 to 21:00 within a specific range around the target charging station before the price adjustment.

[0026] In step 102, that is, after screening by the gradient boosting tree model, multiple target features that have a greater impact on the charging amount from 20 o'clock to 21 o'clock after the price adjustment can be screened out from the multiple features in step 101.

[0027] In step 103, the current data of these target features are input into the charging capacity prediction model, and the average charging capacity prediction value from 20:00 to 21:00 one week after the current price adjustment can be obtained.

[0028] In this way, the average charging capacity forecast value of other target hours in the week after the current price adjustment is obtained, and then the average charging capacity forecast value of each target hour is added up to obtain the daily average charging capacity forecast value in the week after the current price adjustment.

[0029] The charging capacity prediction method provided in this embodiment obtains multiple features that affect the hourly charging capacity after the charging fee price adjustment in the target charging station, inputs the historical data of the multiple features into the gradient boosting tree model, and based on the output result of the gradient boosting tree model, screens out multiple target features with a greater degree of influence, and inputs the current data of the multiple target features into the charging capacity prediction model to obtain the average charging capacity of the target hour one week after the current price adjustment output by the charging capacity prediction model. In this embodiment, the gradient boosting tree model is first used to screen the target features with a greater degree of influence on the hourly charging capacity after the price adjustment, and then the charging capacity prediction model is used to predict the hourly charging capacity after the price adjustment based on these target features. Since the charging capacity prediction model is obtained by training the gradient boosting tree model using the historical data of multiple target features and the average charging capacity of the target hour one week after the historical price adjustment, this embodiment uses the gradient boosting tree model to perform preliminary screening of the target features, and then predicts the hourly charging capacity based on the charging capacity prediction model with a high degree of adaptability to the target features. On the one hand, it can fully explore the historical data rules and the contribution of the target features to the prediction results, thereby improving the prediction accuracy, and on the other hand, it can use mechanized means to replace manual experience prediction to improve prediction efficiency.

[0030] Figure 2 This is the second flow chart of the charging capacity prediction method provided in the embodiment of the present application. Figure 2 In one embodiment, historical data of multiple features are input into a gradient boosting tree model, and multiple target features with greater influence are screened out based on the output results of the gradient boosting tree model, which may include: 201. Input historical data of multiple features into a gradient boosting tree model to obtain a first historical prediction value of an average charging amount in a target hour in a week after the historical price adjustment output by the gradient boosting tree model; 202. Randomly shuffle historical data of candidate features among the multiple features to obtain shuffled historical data of the multiple features; A candidate feature is any one of multiple features; 203. Input the historical shuffled data into the gradient boosting tree model to obtain a second historical prediction value of the average charging amount in the target hour of one week after the historical price adjustment output by the gradient boosting tree model; 204. Calculate the prediction impact value of the candidate feature on the gradient boosting tree model based on the first historical prediction value and the second historical prediction value; 205. Restore the historical data of the candidate feature, select another feature from the multiple features as the candidate feature, and return to step 202; 206. After obtaining the predicted influence value of each feature of the multiple features for the gradient boosting tree model, the predicted influence values ​​of each feature for the gradient boosting tree model are sorted from large to small, and the features corresponding to the multiple predicted influence values ​​with the highest sorting are determined as target features.

[0031] In step 201, it is assumed that multiple features include A, B, and C, each feature includes multiple historical data, and there is a corresponding relationship between the historical data of these features. For example, the historical data of A includes a1, a2, and a3, the historical data of B includes b1, b2, and b3, and the historical data of C includes c1, c2, and c3, where a1, b1, and c1 correspond to form sample 1, a2, b2, and c2 correspond to form sample 2, and a3, b3, and c3 correspond to form sample 3. The data of the three samples are input into the gradient boosting tree model, then what is input is not only the data, but also the corresponding relationship between the data in each sample.

[0032] In step 202, A is used as a candidate feature, and the historical data of A is randomly shuffled, that is, a1, a2 and a3 are randomly shuffled, then the data in sample 1 becomes a2, b1, c1, the data in sample 2 becomes a3, b2, c2, and the data in sample 3 becomes a1, b3, c3, and the corresponding relationship between the data in each sample has changed.

[0033] In step 203, the historical disrupted data is input into the gradient boosting tree model, that is, the data of the three samples after the data change are input into the gradient boosting tree model, and at this time, the data and the corresponding relationship between the data in each sample are also input.

[0034] In step 204, the first historical prediction value and the second historical prediction value are the historical prediction values ​​of the charge capacity output by the gradient boosting tree model before and after the corresponding relationship between the historical data of A and the historical data of other features is changed. If A has a greater impact on the charge capacity prediction of the gradient boosting tree model, the historical prediction value of the charge capacity output by the gradient boosting tree model will fluctuate greatly before and after the corresponding relationship between the historical data of A and the historical data of other features is changed, that is, the gap between the first historical prediction value and the second historical prediction value will be large. Therefore, the prediction influence value of A on the gradient boosting tree model can be calculated based on the first historical prediction value and the second historical prediction value. Specifically, the following two methods can be used: 1. Calculate the absolute value of the difference between the first historical prediction value and the second historical prediction value, and determine the absolute value of the difference as the prediction impact value of A for the gradient boosting tree model; 2. Calculate the difference between the first historical prediction value and the second historical prediction value, calculate the absolute value of the ratio between the difference and the first historical prediction value, and determine the absolute value of the ratio as the prediction impact value of A for the gradient boosting tree model.

[0035] Method 1 directly uses the absolute value of the difference between the first historical prediction value and the second historical prediction value as the prediction impact value of A on the gradient boosting tree model. It is efficient and convenient, and is suitable for scenarios where the gap between the first historical prediction value and the second historical prediction value is small. Method 2 is equivalent to scaling the absolute value of the difference obtained by Method 1 using the first historical prediction value, which can reduce the order of magnitude of the absolute value of the difference and is suitable for scenarios where the gap between the first historical prediction value and the second historical prediction value is large.

[0036] In step 205, the correspondence between the historical data of A and the historical data of other features is restored, and after selecting another feature, such as B as a candidate feature, return to step 202, randomly shuffle the historical data of B, and calculate the prediction impact value of B on the gradient boosting tree model.

[0037] In step 206, after obtaining the predicted influence values ​​of A, B, and C for the gradient boosting tree model respectively, these predicted influence values ​​are sorted from large to small. Assuming that the predicted influence value corresponding to A is greater than the predicted influence value corresponding to C, and the predicted influence value corresponding to C is greater than the predicted influence value corresponding to B, the features corresponding to the two predicted influence values ​​with the highest ranking are selected as the target features, and A and C are selected as the target features.

[0038] This embodiment randomly shuffles the historical data of each of the multiple features one by one, and calculates the difference in the historical prediction values ​​of the charging capacity output by the gradient boosting tree model before and after the shuffling based on multiple algorithms. It can accurately quantify the prediction impact of each feature on the gradient boosting tree model, thereby accurately screening out the target features that have a greater impact on the hourly charging capacity after the price adjustment.

[0039] Figure 3 This is the third flow chart of the charging capacity prediction method provided in the embodiment of the present application. Figure 3 In one embodiment, the charge capacity prediction model can be obtained based on the following steps: 301. Input historical data of multiple target features and the average charging amount in the target hour one week after the historical price adjustment into the gradient boosting tree model, and obtain a third historical prediction value of the average charging amount in the target hour one week after the historical price adjustment output by the gradient boosting tree model; 302. Calculate the mean absolute percentage error between the third historical prediction value and the corresponding historical actual value to obtain the prediction error of the gradient boosting tree model; 303. Based on the prediction error, calculate the prediction accuracy of the gradient boosting tree model; 304. Correcting the prediction accuracy to obtain a corrected accuracy; 305. If the corrected accuracy is less than the accuracy threshold, the learning rate of the gradient boosting tree model is adjusted, and then the process returns to step 301; 306. If the corrected accuracy is greater than or equal to the accuracy threshold, the gradient boosting tree model at this time is determined as a charging capacity prediction model.

[0040] In step 302, the mean absolute percentage error between the third historical prediction value and its corresponding historical actual value is It can be obtained based on the following formula: ; (3-1) in, is the current cumulative number of training times, yes In the training The third historical prediction value of the training, It is the historical actual value corresponding to the third historical prediction value.

[0041] In step 303, the prediction accuracy .

[0042] In step 304, in actual business, the actual value of the hourly charging capacity is sometimes relatively small, or even close to 0. According to formula (3-1), when When it is close to 0, since it exists as the denominator, even if and The gap between them is not large, which will also lead to Very large, even greater than 1, the prediction accuracy is If it is less than 0, the indicator loses its validity. Therefore, the prediction accuracy needs to be corrected to maintain the validity of the indicator.

[0043] In step 305 to step 306, the gradient boosting tree model is trained by adjusting the learning rate of the gradient boosting tree model with the goal of correcting the accuracy rate to be greater than or equal to the accuracy rate threshold, and finally a charging capacity prediction model is obtained.

[0044] This embodiment aims to achieve a model prediction accuracy that meets the requirements. The gradient boosting tree model is trained by adjusting the learning rate of the gradient boosting tree model, and the prediction accuracy is continuously corrected during the training process to maintain the effectiveness of its indicators. This ensures that the final charge capacity prediction model has a high prediction accuracy and a high effectiveness.

[0045] In one embodiment, after the prediction accuracy is corrected to obtain the corrected accuracy, the following steps may be further performed: If the corrected accuracy is less than the accuracy threshold, the learning rate of the gradient boosting tree model is adjusted, and the historical data of multiple features are input into the gradient boosting tree model. Based on the output results of the gradient boosting tree model, multiple target features with greater influence are screened out until the corrected accuracy is greater than or equal to the accuracy threshold. The gradient boosting tree model at this time is determined as the charging capacity prediction model.

[0046] That is, if the corrected accuracy is less than the accuracy threshold, the learning rate of the gradient boosting tree model is adjusted, and then the process returns to step 102 until the corrected accuracy is greater than or equal to the accuracy threshold, and the gradient boosting tree model at this time is determined as the charge capacity prediction model.

[0047] This embodiment not only adjusts the parameters of the gradient boosting tree model, but also re-screens the target features based on the adjusted gradient boosting tree model. The two adjustments work together to complete the training of the model more flexibly and efficiently, and obtain a charging capacity prediction model with both high prediction accuracy and effectiveness.

[0048] Figure 4 This is a fourth flow chart of the charging capacity prediction method provided in the embodiment of the present application. Figure 4 In one embodiment, correcting the prediction accuracy to obtain a corrected accuracy may include: 401. Setting a cutoff value of the prediction error based on the hourly charging capacity characteristics of the target charging station; 402. Correcting the prediction error based on the cutoff value to obtain a corrected error; 403. Correct the prediction accuracy based on the correction error to obtain a corrected accuracy.

[0049] In step 401, it is found in actual calculation that when the hourly charging capacity of the charging station increases from 0 degrees to 10 degrees, and Each 1 degree difference corresponds to are very large, resulting in a high prediction accuracy Failure, and the hourly charging capacity of most charging stations is greater than 10 degrees, so it is necessary to correspond to the charging capacity range of 0 degrees to 10 degrees For truncation, the truncation value can be 10.

[0050] In step 402, The calculation formula is modified to: ; (4-1) in, is relative to The correction error.

[0051] In step 403, the accuracy is corrected .

[0052] In addition, if the correction accuracy calculated by this embodiment is still less than 0, the correction accuracy is determined to be 0.

[0053] This embodiment uses the error characteristics of the hourly charging capacity of the charging station between 0 degrees and 10 degrees to set the cutoff value of the prediction error, and scales the originally large prediction error based on the cutoff value, so that the obtained corrected error is reduced to an acceptable range without losing its effectiveness, and then the correction accuracy calculated based on the corrected error is greater than or equal to 0 to maintain its effectiveness. In addition, this embodiment also provides a further fallback solution, that is, when the correction accuracy is still less than 0 and fails, it is determined to be 0, thereby further ensuring its effectiveness.

[0054] The charging amount prediction device provided in an embodiment of the present application is described below. The charging amount prediction device described below and the charging amount prediction method described above can be referenced to each other.

[0055] Figure 5 Schematic diagram of the structure of the charging capacity prediction device provided in the embodiment of the present application. Figure 5 , the embodiment of the present application provides a charging capacity prediction device, which may include: The hourly charging capacity influencing feature acquisition module 501 is used to: acquire multiple features that affect the hourly charging capacity after the charging fee price adjustment in the target charging station; The hourly charging amount target feature screening module 502 is used to: input the historical data of the multiple features into the gradient boosting tree model, and screen out multiple target features with greater influence based on the output results of the gradient boosting tree model; The hourly charging capacity prediction module 503 is used to: input the current data of the multiple target features into the charging capacity prediction model to obtain the average charging capacity of the target hours in one week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

[0056] The charging capacity prediction device provided in this embodiment obtains multiple features that affect the hourly charging capacity after the charging fee price adjustment in the target charging station, inputs the historical data of the multiple features into the gradient boosting tree model, and based on the output result of the gradient boosting tree model, screens out multiple target features with a greater degree of influence, inputs the current data of the multiple target features into the charging capacity prediction model, and obtains the average charging capacity of the target hour one week after the current price adjustment output by the charging capacity prediction model. In this embodiment, the gradient boosting tree model is first used to screen the target features with a greater degree of influence on the hourly charging capacity after the price adjustment, and then the charging capacity prediction model is used to predict the hourly charging capacity after the price adjustment based on these target features. Since the charging capacity prediction model is obtained by training the gradient boosting tree model using the historical data of multiple target features and the average charging capacity of the target hour one week after the historical price adjustment, this embodiment uses the gradient boosting tree model to perform preliminary screening of the target features, and then predicts the hourly charging capacity based on the charging capacity prediction model with a high degree of adaptability to the target features. On the one hand, it can fully explore the historical data rules and the contribution of the target features to the prediction results, thereby improving the prediction accuracy, and on the other hand, it can use mechanized means to replace manual experience prediction to improve prediction efficiency.

[0057] In one embodiment, the hourly charge amount target feature screening module 502 is specifically used to: Inputting the historical data of the plurality of features into the gradient boosting tree model, and obtaining a first historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Randomly shuffle historical data of candidate features among the multiple features to obtain shuffled historical data of the multiple features; the candidate feature is any feature among the multiple features; Inputting the historical shuffled data into the gradient boosting tree model, and obtaining a second historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Based on the first historical prediction value and the second historical prediction value, calculating the prediction impact value of the candidate feature on the gradient boosting tree model; Restoring the historical data of the candidate feature, selecting another feature from the multiple features as the candidate feature, and returning to the step of randomly shuffling the historical data of the candidate feature from the multiple features to obtain the historical shuffled data of the multiple features, until obtaining the predicted influence value of each feature from the multiple features for the gradient boosting tree model; The predicted influence values ​​of each feature on the gradient boosting tree model are sorted from large to small, and the features corresponding to the top-ranked predicted influence values ​​are determined as target features.

[0058] In one embodiment, the hourly charge amount target feature screening module 502 is specifically used to: Calculating an absolute value of a difference between the first historical prediction value and the second historical prediction value; Determine the absolute value of the difference as the predicted impact value of the candidate feature on the gradient boosting tree model; or Calculating a difference between the first historical prediction value and the second historical prediction value; Calculating an absolute value of a ratio between the difference and the first historical prediction value; The absolute value of the ratio is determined as the prediction influence value of the candidate feature on the gradient boosting tree model.

[0059] In one embodiment, a charging capacity prediction model building module (not shown in the figure) is also included, which is used to: Inputting the historical data of the multiple target features and the average charging amount of the target hour one week after the historical price adjustment into the gradient boosting tree model, and obtaining a third historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Calculating the mean absolute percentage error between the third historical predicted value and its corresponding historical actual value to obtain the prediction error of the gradient boosting tree model; Based on the prediction error, calculating the prediction accuracy of the gradient boosting tree model; Correcting the prediction accuracy to obtain a corrected accuracy; If the revised accuracy is less than the accuracy threshold, then after adjusting the learning rate of the gradient boosting tree model, return to the step of inputting the historical data of the multiple target features and the average charging amount of the target hours one week after the historical price adjustment into the gradient boosting tree model, and obtain the third historical prediction value of the average charging amount of the target hours one week after the historical price adjustment output by the gradient boosting tree model, until the revised accuracy is greater than or equal to the accuracy threshold, and the gradient boosting tree model at this time is determined as the charging amount prediction model.

[0060] In one embodiment, the charging capacity prediction model building module is further used to: If the corrected accuracy is less than the accuracy threshold, then after adjusting the learning rate of the gradient boosting tree model, return to the step of inputting the historical data of the multiple features into the gradient boosting tree model, and based on the output results of the gradient boosting tree model, screen out multiple target features with greater influence until the corrected accuracy is greater than or equal to the accuracy threshold, and determine the gradient boosting tree model at this time as the charging capacity prediction model.

[0061] In one embodiment, the charging capacity prediction model building module is further used to: Setting a cutoff value of the prediction error based on the hourly charging capacity characteristics of the target charging station; Correcting the prediction error based on the cutoff value to obtain a corrected error; The prediction accuracy is corrected based on the correction error to obtain a corrected accuracy.

[0062] Figure 6 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call a computer program in the memory 630 to execute the steps of the charge capacity prediction method, for example including: Obtain multiple features that affect the hourly charging volume after the charging fee is adjusted in the target charging station; Inputting the historical data of the multiple features into a gradient boosting tree model, and screening out multiple target features with greater influence based on the output results of the gradient boosting tree model; Inputting the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

[0063] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0064] On the other hand, an embodiment of the present application further provides a computer program product, the computer program product including a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can perform the steps of the charge capacity prediction method provided in the above embodiments, for example, including: Obtain multiple features that affect the hourly charging volume after the charging fee is adjusted in the target charging station; Inputting the historical data of the multiple features into a gradient boosting tree model, and screening out multiple target features with greater influence based on the output results of the gradient boosting tree model; Inputting the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

[0065] On the other hand, an embodiment of the present application further provides a non-transitory computer-readable storage medium on which a computer program is stored, and the computer program is used to enable a processor to execute the steps of the charge capacity prediction method provided in the above embodiments, for example, including: Obtain multiple features that affect the hourly charging volume after the charging fee is adjusted in the target charging station; Inputting the historical data of the multiple features into a gradient boosting tree model, and screening out multiple target features with greater influence based on the output results of the gradient boosting tree model; Inputting the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

[0066] The non-transitory computer-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)), etc.

[0067] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0068] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting charging capacity, characterized in that: include: Obtain multiple features that affect the hourly charging volume after the charging fee is adjusted in the target charging station; Inputting the historical data of the multiple features into a gradient boosting tree model, and screening out multiple target features with greater influence based on the output results of the gradient boosting tree model; Inputting the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

2. The charge capacity prediction method according to claim 1, characterized in that: The historical data of the multiple features are input into the gradient boosting tree model, and based on the output result of the gradient boosting tree model, multiple target features with greater influence are screened out, including: Inputting the historical data of the plurality of features into the gradient boosting tree model, and obtaining a first historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Randomly shuffle historical data of candidate features among the multiple features to obtain shuffled historical data of the multiple features; the candidate feature is any feature among the multiple features; Inputting the historical shuffled data into the gradient boosting tree model, and obtaining a second historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Based on the first historical prediction value and the second historical prediction value, calculating the prediction impact value of the candidate feature on the gradient boosting tree model; Restoring the historical data of the candidate feature, selecting another feature from the multiple features as the candidate feature, and returning to the step of randomly shuffling the historical data of the candidate feature from the multiple features to obtain the historical shuffled data of the multiple features, until obtaining the predicted influence value of each feature from the multiple features for the gradient boosting tree model; The predicted influence values ​​of each feature on the gradient boosting tree model are sorted from large to small, and the features corresponding to the top-ranked predicted influence values ​​are determined as target features.

3. The charge capacity prediction method according to claim 2, characterized in that: The calculating, based on the first historical prediction value and the second historical prediction value, the prediction influence value of the candidate feature on the gradient boosting tree model includes: Calculating an absolute value of a difference between the first historical prediction value and the second historical prediction value; Determine the absolute value of the difference as the predicted impact value of the candidate feature on the gradient boosting tree model; or Calculating a difference between the first historical prediction value and the second historical prediction value; Calculating an absolute value of a ratio between the difference and the first historical prediction value; The absolute value of the ratio is determined as the prediction influence value of the candidate feature on the gradient boosting tree model.

4. The charge capacity prediction method according to claim 1, characterized in that: The charging capacity prediction model is obtained based on the following steps: Inputting the historical data of the multiple target features and the average charging amount of the target hour one week after the historical price adjustment into the gradient boosting tree model, and obtaining a third historical prediction value of the average charging amount of the target hour one week after the historical price adjustment output by the gradient boosting tree model; Calculating the mean absolute percentage error between the third historical predicted value and its corresponding historical actual value to obtain the prediction error of the gradient boosting tree model; Based on the prediction error, calculating the prediction accuracy of the gradient boosting tree model; Correcting the prediction accuracy to obtain a corrected accuracy; If the revised accuracy is less than the accuracy threshold, then after adjusting the learning rate of the gradient boosting tree model, return to the step of inputting the historical data of the multiple target features and the average charging amount of the target hours one week after the historical price adjustment into the gradient boosting tree model, and obtain the third historical prediction value of the average charging amount of the target hours one week after the historical price adjustment output by the gradient boosting tree model, until the revised accuracy is greater than or equal to the accuracy threshold, and the gradient boosting tree model at this time is determined as the charging amount prediction model.

5. The charge capacity prediction method according to claim 4, characterized in that: After the prediction accuracy is corrected to obtain the corrected accuracy, the method further includes: If the corrected accuracy is less than the accuracy threshold, then after adjusting the learning rate of the gradient boosting tree model, return to the step of inputting the historical data of the multiple features into the gradient boosting tree model, and based on the output results of the gradient boosting tree model, screen out multiple target features with greater influence until the corrected accuracy is greater than or equal to the accuracy threshold, and determine the gradient boosting tree model at this time as the charging capacity prediction model.

6. The charge capacity prediction method according to claim 4, characterized in that: The step of correcting the prediction accuracy to obtain a corrected accuracy includes: Setting a cutoff value of the prediction error based on the hourly charging capacity characteristics of the target charging station; Correcting the prediction error based on the cutoff value to obtain a corrected error; The prediction accuracy is corrected based on the correction error to obtain a corrected accuracy.

7. A charge capacity prediction device, characterized in that: include: The hourly charging capacity impact feature acquisition module is used to: acquire multiple features that affect the hourly charging capacity after the charging fee price adjustment in the target charging station; An hourly charging target feature screening module is used to: input the historical data of the multiple features into a gradient boosting tree model, and screen out multiple target features with greater influence based on the output results of the gradient boosting tree model; An hourly charging capacity prediction module is used to: input the current data of the multiple target features into a charging capacity prediction model to obtain an average charging capacity of a target hour in a week after the current price adjustment output by the charging capacity prediction model; The charging capacity prediction model is obtained by training on the basis of the gradient boosting tree model using the historical data of the multiple target features and the average charging capacity of the target hour in one week after the historical price adjustment.

8. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the charge amount prediction method according to any one of claims 1 to 6 are implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the charge capacity prediction method according to any one of claims 1 to 6 are implemented.

10. A non-transitory 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 charge capacity prediction method according to any one of claims 1 to 6 are implemented.

Images