Battery replacement demand prediction method and system of battery replacement station

Abstract

The invention discloses a battery replacing demand prediction method and system for a battery replacing station, and the method comprises the steps: obtaining the historical battery replacing data of the battery replacing station, wherein the historical battery replacing data of the battery replacing station comprises a training set; training a multiple regression model based on the training set to obtain a trained multiple regression model; and inputting to-be-predicted battery replacement data of a to-be-predicted battery replacement station into the trained multiple regression model, and outputting a predicted battery replacement demand of the to-be-predicted battery replacement station, the predicted battery replacement demand comprising predicted battery replacement order data of the to-be-predicted battery replacement station in a future preset time. According to the method, the attribute data and the vehicle data of the battery swap station are used as input, the historical battery swap order data are used as output to train the multiple regression model, and the to-be-predicted battery swap data of the to-be-predicted battery swap station are input into the trained multiple regression model to output the predicted battery swap demand of the to-be-predicted battery swap station. The method can accurately predict the battery swap demand of the battery swap station in the future time based on the multiple regression model, and improves the prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of battery swapping stations, and particularly relates to a method and a system for predicting battery swapping demands of a battery swapping station. Background Art

[0002] In the operation of a battery swapping station, being able to accurately estimate how many battery swapping demands (battery swapping orders / battery swapping times) the battery swapping station will have in a future period of time plays a key role in the implementation of refined operation projects such as reasonably allocating the number of personnel in the battery swapping station, diverting vehicles in the battery swapping station, evaluating electricity prices for charging, and scientifically matching batteries.

[0003] Currently, the statistical method - moving average method is used to predict the battery swapping demand of the battery swapping station in the next 1 day based on the historical order volume of the battery swapping station itself. The moving average method is simple to implement and easy to understand, but it only has one dimension of order volume. In addition, the prediction accuracy of the moving average method is relatively low. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the defect of low accuracy in the method for predicting the battery swapping demand of a battery swapping station in the prior art, and to provide a method and a system for predicting the battery swapping demand of a battery swapping station.

[0005] The present invention solves the above technical problem through the following technical solutions:

[0006] The first aspect of the present invention provides a method for predicting the battery swapping demand of a battery swapping station, and the method for predicting the battery swapping demand includes:

[0007] Obtain historical battery swapping data of the battery swapping station, where the historical battery swapping data of the battery swapping station includes a training set, and the training set includes battery swapping station attribute data and vehicle data as inputs and historical battery swapping order data as an output;

[0008] Train a multiple regression model based on the training set to obtain a trained multiple regression model;

[0009] Input the to - be - predicted battery swapping data of the to - be - predicted battery swapping station into the trained multiple regression model to output the predicted battery swapping demand of the to - be - predicted battery swapping station, where the predicted battery swapping demand includes predicted battery swapping order data of the to - be - predicted battery swapping station within a future preset time.

[0010] Training a multiple regression model with battery swapping station attribute data and vehicle data as inputs and historical battery swapping order data as an output, and inputting the to - be - predicted battery swapping data of the to - be - predicted battery swapping station into the trained multiple regression model to output the predicted battery swapping demand of the to - be - predicted battery swapping station can accurately predict the battery swapping demand of the battery swapping station in the future time based on the multiple regression model, and improve the prediction accuracy.

[0011] Preferably, the method for predicting the battery swapping demand further includes:

[0012] Perform data anomaly processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing;

[0013] Perform feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0014] By performing data anomaly processing on the historical battery swapping data of the battery swapping station, the accuracy of the historical battery swapping data of the battery swapping station is improved. By performing feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing, the calculation cost is reduced and the model upper limit is increased.

[0015] Preferably, the historical battery swapping data of the battery swapping station further includes a test set; the battery swapping demand prediction method further includes:

[0016] Use the test set to test the prediction result of the trained multiple regression model to obtain the predicted battery swapping demand corresponding to the test set;

[0017] Obtain the actual battery swapping demand corresponding to the test set;

[0018] Based on the predicted battery swapping demand corresponding to the test set and the actual battery swapping demand corresponding to the test set, obtain the root mean square error and accuracy of the trained multiple regression model as the evaluation value of the trained multiple regression model;

[0019] Optimize the trained multiple regression model based on the comparison result between the evaluation value and the target evaluation value.

[0020] Using the test set to evaluate the quality of the multiple regression model and optimizing the trained multiple regression model based on the comparison result between the evaluation value and the target evaluation value improves the accuracy of the multiple regression model.

[0021] Preferably, the battery swapping demand prediction method further includes:

[0022] Tag the historical battery swapping data of the battery swapping station to obtain an order level tag and an order volatility tag.

[0023] By tagging the historical battery swapping data of the battery swapping station with an order level tag and an order volatility tag, the historical battery swapping data of the battery swapping station is divided into finer granularities. Based on the divided historical battery swapping data of the battery swapping station, multiple regression models are respectively established for training, further improving the accuracy of the multiple regression model.

[0024] Preferably, the step of performing data anomaly processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing includes:

[0025] Process the missing data, discrete data, and normalize the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing.

[0026] By processing the missing data, discrete data, and normalizing the historical battery swapping data of the battery swapping station, the accuracy of the historical battery swapping data of the battery swapping station is improved.

[0027] Preferably, the step of performing feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing includes:

[0028] Perform feature construction and filtering variable processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0029] By performing feature construction and filtering variable processing on the historical battery swapping data of the battery swapping station after data anomaly processing, the calculation cost is reduced and the model upper limit is improved.

[0030] Preferably, the battery swapping station attribute data includes at least one of the battery swapping station version, supported vehicle models, number of business sites, start business hours, end business hours, preset business hours, cumulative operation hours, city where the station end is located, county where the station end is located, charging price at the station end, configuration quantity of each model battery at the station end, and number of positions at the station end;

[0031] and / or,

[0032] The vehicle data includes at least one of the cumulative number of networked vehicles, distribution quantity of vehicle operation types, number of active vehicles, vehicle new addition month-on-month ratio, and vehicle network entry registration sites;

[0033] and / or,

[0034] The historical battery swapping order data includes at least one of the order distribution of each package type, average daily number of battery swapping vehicles, average daily battery swapping mileage, number of battery swapping orders per day, battery swapping mileage per day, and average battery swapping duration.

[0035] By training the multiple regression model with various different battery swapping station attribute data and vehicle data as inputs and various different historical battery swapping order data as outputs, the performance of the multiple regression model can be made more abundant and the prediction range can be wider.

[0036] The second aspect of the present invention provides a battery swapping demand prediction system for a battery swapping station, and the battery swapping demand prediction system includes:

[0037] A first acquisition module, configured to acquire historical battery swapping data of a battery swapping station, where the historical battery swapping data of the battery swapping station includes a training set, and the training set includes battery swapping station attribute data and vehicle data as inputs and historical battery swapping order data as an output;

[0038] A training module, configured to train a multiple regression model based on the training set to obtain a trained multiple regression model;

[0039] A prediction module, configured to input the to-be-predicted battery swapping data of a to-be-predicted battery swapping station into the trained multiple regression model to output a predicted battery swapping demand of the to-be-predicted battery swapping station, where the predicted battery swapping demand includes predicted battery swapping order data of the to-be-predicted battery swapping station within a preset future time.

[0040] Training a multiple regression model with battery swapping station attribute data and vehicle data as inputs and historical battery swapping order data as an output, and inputting the to-be-predicted battery swapping data of a to-be-predicted battery swapping station into the trained multiple regression model to output a predicted battery swapping demand of the to-be-predicted battery swapping station can accurately predict the battery swapping demand of a battery swapping station at a future time based on the multiple regression model, improving the prediction accuracy.

[0041] Preferably, the battery swapping demand prediction system further includes:

[0042] A data anomaly processing module, configured to perform data anomaly processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing;

[0043] A feature engineering processing module, configured to perform feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0044] By performing data anomaly processing on the historical battery swapping data of the battery swapping station, the accuracy of the historical battery swapping data of the battery swapping station is improved. By performing feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing, the calculation cost is reduced and the model upper limit is improved.

[0045] Preferably, the historical battery swapping data of the battery swapping station further includes a test set; the battery swapping demand prediction system further includes:

[0046] A test module, configured to use the test set to test the prediction result of the trained multiple regression model to obtain a predicted battery swapping demand corresponding to the test set;

[0047] A second acquisition module, configured to acquire the true battery swapping demand corresponding to the test set;

[0048] A third acquisition module, configured to obtain the root mean square error and accuracy of the trained multiple regression model based on the predicted battery swapping demand corresponding to the test set and the actual battery swapping demand corresponding to the test set, as the evaluation value of the trained multiple regression model;

[0049] An optimization module, configured to optimize the trained multiple regression model based on the comparison result between the evaluation value and the target evaluation value.

[0050] Evaluating the quality of the multiple regression model using a test set and optimizing the trained multiple regression model based on the comparison result between the evaluation value and the target evaluation value improves the accuracy of the multiple regression model.

[0051] Preferably, the battery swapping demand prediction system further includes:

[0052] A label processing module, configured to label the historical battery swapping data of the battery swapping station to obtain an order level label and an order volatility label.

[0053] By labeling the historical battery swapping data of the battery swapping station with an order level label and an order volatility label, the historical battery swapping data of the battery swapping station is divided into finer granularity. Based on the divided historical battery swapping data of the battery swapping station, multiple regression models are respectively established for training, further improving the accuracy of the multiple regression model.

[0054] Preferably, the data anomaly processing module is configured to perform missing data processing, discrete data processing, and normalization processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing.

[0055] By performing missing data processing, discrete data processing, and normalization processing on the historical battery swapping data of the battery swapping station, the accuracy of the historical battery swapping data of the battery swapping station is improved.

[0056] Preferably, the feature engineering processing module is configured to perform feature construction and variable filtering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0057] By performing feature construction and variable filtering processing on the historical battery swapping data of the battery swapping station after data anomaly processing, the calculation cost is reduced and the model upper limit is improved.

[0058] Preferably, the battery swapping station attribute data includes at least one of the battery swapping station version, supported vehicle models, number of business sites, start business hours, end business hours, business hours during a preset time, cumulative operation hours, city where the station is located, county where the station is located, charging price at the station, configuration quantity of each model of battery at the station, and number of positions at the station;

[0059] and / or

[0060] The vehicle data includes at least one of the cumulative number of networked vehicles, the distribution quantity of vehicle operation types, the number of active vehicles, the month-on-month increase in newly added vehicles, and the vehicle network registration sites.

[0061] and / or

[0062] The historical battery swapping order data includes at least one of the order distribution of each package type, the daily average number of battery swapping vehicles, the daily average battery swapping mileage, the number of battery swapping orders per day, the number of battery swapping mileage per day, and the average battery swapping duration.

[0063] By training a multiple regression model with various different battery swapping station attribute data and vehicle data as inputs and various different historical battery swapping order data as outputs, the performance of the multiple regression model can be made more abundant and the prediction range can be wider.

[0064] The third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and configured to run on the processor. When the processor executes the computer program, the battery swapping demand prediction method of the battery swapping station as described in the first aspect is implemented.

[0065] The fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the battery swapping demand prediction method of the battery swapping station as described in the first aspect is implemented.

[0066] On the basis of conforming to common knowledge in the art, the above preferred conditions can be combined arbitrarily to obtain various preferred embodiments of the present invention.

[0067] The positive and progressive effects of the present invention are as follows:

[0068] The present invention uses battery swapping station attribute data and vehicle data as inputs and historical battery swapping order data as outputs to train a multiple regression model, and inputs the to-be-predicted battery swapping data of the to-be-predicted battery swapping station into the trained multiple regression model to output the predicted battery swapping demand of the to-be-predicted battery swapping station, which can accurately predict the battery swapping demand of the battery swapping station at a future time based on the multiple regression model and improve the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flowchart of the battery swapping demand prediction method of the battery swapping station according to Embodiment 1 of the present invention.

[0070] Figure 2 It is a schematic diagram of the modules of the battery swapping demand prediction system of the battery swapping station according to Embodiment 2 of the present invention.

[0071] Figure 3Schematic structural diagram of an electronic device for implementing the method for predicting the battery replacement demand of a battery replacement station according to Embodiment 3 of the present invention. Detailed implementation manners

[0072] The present invention will be further described below by way of embodiments, but the present invention is not limited to the scope of the described embodiments.

[0073] Embodiment 1

[0074] A method for predicting the battery replacement demand of a battery replacement station provided in this embodiment is as Figure 1 shown. The method for predicting the battery replacement demand includes:

[0075] Step 101, obtain the historical battery replacement data of the battery replacement station. The historical battery replacement data of the battery replacement station includes a training set, and the training set includes the battery replacement station attribute data and vehicle data as inputs and the historical battery replacement order data as outputs;

[0076] In an optional embodiment, the battery replacement station attribute data (for example, the basic data of the battery replacement station) includes at least one of the battery replacement station version, supported vehicle models, number of business sites, start business hours, end business hours, business hours during a preset time, cumulative operation duration, city where the station end is located, county where the station end is located, charging price at the station end, configured quantity of each model battery at the station end, and number of positions at the station end;

[0077] In this embodiment, the business sites include cities, counties, etc.

[0078] It should be noted that the preset time is set according to the actual situation. For example, the preset time can be set to 1 day or other values, and no specific limitation is made here.

[0079] In an optional embodiment, the vehicle data (for example, vehicle end data) includes at least one of the cumulative number of networked vehicles, distribution quantity of vehicle operation types, number of active vehicles, vehicle new addition month-on-month ratio, and vehicle network entry registration sites;

[0080] In this embodiment, the cumulative number of networked vehicles and the number of active vehicles are both counted by brand; the vehicle new addition month-on-month ratio is counted by brand and month.

[0081] In an optional embodiment, the historical battery replacement order data (for example, consumption data) includes at least one of the order distribution of each package type, average daily number of battery replacement vehicles (for example, average daily number of battery replacement vehicles in the recent week), average daily battery replacement mileage (for example, average daily battery replacement mileage in the recent week), number of battery replacement orders per day, battery replacement mileage per day, and average battery replacement duration.

[0082] In this embodiment, the number of battery replacement orders per day and the battery replacement mileage per day are both counted by model.

[0083] Step 102: Train a multiple regression model based on the training set to obtain the trained multiple regression model.

[0084] In this embodiment, in the actual business scenario, since factors such as weekdays, non-weekdays, and seasons of the battery swapping station will affect the battery swapping demand of the battery swapping station, when training the multiple regression model, information such as weekdays, non-weekdays, and seasons of the battery swapping station is also included in the training set to train the multiple regression model.

[0085] Step 103: Input the to-be-predicted battery swapping data of the to-be-predicted battery swapping station into the trained multiple regression model to output the predicted battery swapping demand of the to-be-predicted battery swapping station, where the predicted battery swapping demand includes the predicted battery swapping order data of the to-be-predicted battery swapping station within a future preset time.

[0086] In this embodiment, the battery swapping station attribute data and vehicle data are used as inputs, and the historical battery swapping order data is used as the output to train the multiple regression model. Then, the to-be-predicted battery swapping data of the to-be-predicted battery swapping station is input into the trained multiple regression model to output the predicted battery swapping demand of the to-be-predicted battery swapping station, which can accurately predict the battery swapping demand of the battery swapping station in the future based on the multiple regression model and improve the prediction accuracy.

[0087] In an alternative embodiment, the battery swapping demand prediction method further includes:

[0088] Step 101-1: Perform data anomaly processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing.

[0089] In an alternative embodiment, step 101-1 includes:

[0090] Perform missing data processing, discrete data processing, and normalization processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing.

[0091] In this embodiment, for missing data processing, data with a missing ratio higher than a preset ratio (for example, the preset ratio can be 30%) is deleted, and data with a missing ratio within the preset ratio (for example, 30%) is filled with 0 or the mean value. For example, for battery swapping orders, the number of battery swapping vehicles, and the business rate, if there is no corresponding record on the same day, it is filled with 0.

[0092] For discrete data processing, one-hot (one valid) encoding is performed on the historical battery swapping data of the power station, such as battery swapping vehicle models, seasons, weekdays, holidays, and whether it is open all day. Before one-hot encoding, it is shown in Table 1, and after one-hot encoding, it is shown in Table 2:

[0093] Table 1

[0094] Serial number Vehicle model 1 A 2 B 3 C

[0095] Table 2

[0096] Serial number Vehicle model A Vehicle model B Vehicle model C 1 1 0 0 2 0 1 0 3 0 0 1

[0097] To eliminate the influence of dimension, normalization processing ([0, 1]) is performed on the historical battery swapping data of the battery swapping station, such as variables like the number of battery swapping vehicles, the number of battery swapping orders, the number of newly networked vehicles, and the cumulative number of networked vehicles. The calculation formula for normalization processing is shown in Formula (1):

[0098]

[0099] Among them, X represents an independent variable (for example, the number of battery swapping vehicles), and X i represents the value of variable X in the i-th record.

[0100] Step 101-2: Perform feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0101] In an optional embodiment, Step 101-2 includes:[[]]

[0102] Perform feature construction and filtered variable processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0103] In this embodiment, the feature construction of the historical battery swapping data of the battery swapping station after data anomaly processing includes the following contents:

[0104] Actual business rate (days) of the battery swapping station = actual business hours / theoretical business hours;

[0105] Whether it operates throughout the day: If end business time (seconds) - start business time (seconds) = 86400, then it operates throughout the day, otherwise it does not operate throughout the day;

[0106] Cumulative business hours of the battery swapping station = number of days between the statistical date and the time of the first order of the battery swapping station + 1;

[0107] Average number of battery swapping orders in the recent N days: Average value of the order values of the battery swapping station in the recent N days;

[0108] Fluctuation value of battery swapping orders in the recent N days: Ratio of the standard deviation of the order values in the recent N days to the average of the order values in the recent N days;

[0109] Filter the variable of the historical battery swapping data of the battery swapping station after data anomaly processing. Specifically, filter the variables with low relevance: In this embodiment, the variance filtering method is used to screen. If the variance of a feature or variable itself is very small, it means that there is basically no difference in the sample for this feature or variable. It is possible that most of the values in the feature are the same, or even the values of the entire feature are the same. Then this feature has little effect on sample discrimination. In this embodiment, variables with low relevance to the predicted target variable, such as the battery swapping station version and the vehicle network registration site, are filtered. So far, the data used to construct the multiple regression model is finally sorted as shown in Table 3:

[0110] Table 3

[0111]

[0112]

[0113]

[0114] In this embodiment, data anomaly processing is a process of detecting, correcting, or deleting inaccurate or inapplicable records from the historical battery swapping data of the battery swapping station for the multiple regression model. Feature engineering processing is a process of converting the original historical battery swapping data of the battery swapping station into features that can better represent the potential problems of the predicted multiple regression model, which can be achieved by selecting the most relevant features, extracting features, and creating features. In addition, through feature engineering processing, the calculation cost can be reduced and the model upper limit can be improved.

[0115] In an optional embodiment, the historical battery swapping data of the battery swapping station further includes a test set; the battery swapping demand prediction method further includes:

[0116] Step 201: Use the test set to test the prediction result of the trained multiple regression model to obtain the predicted battery swapping demand corresponding to the test set;

[0117] Step 202: Obtain the actual battery swapping demand corresponding to the test set;

[0118] Step 203: Obtain the root mean square error and accuracy of the trained multiple regression model based on the predicted battery swapping demand corresponding to the test set and the actual battery swapping demand corresponding to the test set as the evaluation value of the trained multiple regression model;

[0119] Step 204: Optimize the trained multiple regression model based on the comparison result between the evaluation value and the target evaluation value.

[0120] In the specific implementation process, a multivariate regression model is constructed through the OLS (ordinary least squares) algorithm to predict the battery swap demand of the battery swap station. OLS is a supervised machine learning model. The multivariate regression model is trained based on a large amount of historical battery swap data of the battery swap station to predict the predicted battery swap order data of the battery swap station within a preset time in the future. The supervised learning model needs to divide the historical battery swap data of the battery swap station into a training set and a test set. The training set is used to train the multivariate regression model, and the test set is used to evaluate the quality of the multivariate regression model.

[0121] The target variable (for example, battery replacement demand) is related to the time sequence. This embodiment divides the training set and the test set according to the timeline. For example, the historical battery replacement data of the battery replacement station from 2020-08-31 to T is used as the training set to predict the battery replacement demand (and battery replacement order data) of the site T+1, where T slides a window every day from 2021-08-31 to 2021-11-09, and the training set only takes data from the past year.

[0122] For example, the forecast of battery swapping demand by model at battery swapping stations on 2021-09-01;

[0123] Training set: historical battery swap data of battery swap stations from 2020-08-31 to 2021-08-31;

[0124] Test set: The test set needs to input the values ​​of all battery swapping demands on 2021-09-01, but 2021-09-01 is a future date, and the actual operating rate of the battery swap station, the number of new vehicles connected to the network, the cumulative number of vehicles connected to the network, etc. cannot be obtained. Therefore, this solution can use the data of 2021-08-31 for filling.

[0125] In addition, OLS predicts the dependent variable (i.e., the target variable) through a series of independent variables. The basic idea of ​​OLS is that the process of training the multivariate regression model is to find a set of optimal learning parameters (a1, a2, ... an) to minimize the sum of squared errors between the predicted value and the actual value of the multivariate regression model. The mathematical expression of the OLS model is shown in formula (2):

[0126] y=T0+a1×x1+a2×x2+a3×x3+...+a n × n (2)

[0127] Where y represents the target variable (e.g., battery replacement demand), x i represents the independent variable (e.g., battery swap station attribute data and vehicle data), a i Represents the parameters to be learned.

[0128] In this embodiment, the evaluation of the multiple regression model mainly measures the error between the predicted value and the true value (i.e., the error between the predicted battery swapping demand and the true battery swapping demand). The commonly used evaluation metrics for the multiple regression model include the root mean square error rmse, accuracy rate R 2 and so on. The root mean square error rmse is the square root of the ratio of the sum of the squares of the differences between the predicted value and the true value to the number of samples n. The smaller the root mean square error rmse, the smaller the error between the predicted value and the true value of a set of data. The calculation formula is shown in Formula (3):

[0129]

[0130] where n represents the number of the test set, y i represents the true battery swapping demand corresponding to the test set (i.e., the true battery swapping order data), represents the predicted battery swapping demand corresponding to the test set.

[0131] It should be noted that the accuracy rate R 2 is to measure the goodness of fit of the predicted value to the true value. The calculation formula is shown in Formula (4):

[0132]

[0133] where n represents the number of the test set, y i represents the true battery swapping demand corresponding to the test set (i.e., the true battery swapping order data), represents the predicted battery swapping demand corresponding to the test set, represents the mean value of the true battery swapping demand (i.e., the true battery swapping order data). The range of R 2 is (-∞, 1]. In the prediction practice, the multiple regression model with the highest R-squared is often adopted.

[0134] In an optional embodiment, the battery swapping demand prediction method further includes:

[0135] Labeling the historical battery swapping data of the battery swapping station to obtain the order level label and the order volatility label.

[0136] In this embodiment, there are many operating battery swapping stations, and there are multiple versions of the battery swapping stations. The maximum service capabilities (times / day) of different versions of the battery swapping stations are different, and the actual service capabilities of the battery swapping stations are also affected by the number of batteries configured in the station. Therefore, the interval range of the daily battery swapping order data (such as the number of battery swapping orders) of the battery swapping station is between [0, 900), and the interval span is large. Moreover, the daily stability of the number of battery swapping orders of different battery swapping stations is also different, and the prediction accuracy of the previous model is quite average.

[0137] In the case where the number and stability of battery swapping orders vary, in order to improve the model accuracy, in this embodiment, it is necessary to perform a more fine-grained division of the historical battery swapping data of the battery swapping station in combination with the number of battery swapping orders and the order stability, and establish a multiple regression model for training on the divided historical battery swapping data of the battery swapping station. To achieve this goal, first, it is necessary to label the historical battery swapping data of the battery swapping station with order level labels and order volatility (measuring stability) labels. The method of labeling the historical battery swapping data of the battery swapping station is described as follows:

[0138] Take the number of battery swapping orders of the battery swapping station itself in the recent N days to calculate the average value and volatility value of each battery swapping station. Sort the order average value and volatility value from small to large respectively, and perform segmented processing according to the quantile and actual business experience, and label values according to the interval. Table 4 shows the order level labels and the order average values corresponding to the order level labels, and Table 5 shows the order volatility labels and the order average values corresponding to the order volatility labels. For example, sort the average value of the battery swapping orders of the battery swapping station itself in the recent 30 days from small to large, and combine the quantile and actual business experience to extract several demarcation points of 100, 300, 500, and 800. One interval is formed between two demarcation points, and each interval is labeled (the same applies to the volatility label). Finally, there are a total of 6 order level labels and 5 order volatility labels.

[0139] Table 4

[0140] Order level label Order mean 1 (0,100] 2 (100,300] 3 (300,500] 4 (500,800] 5 (800, inf)

[0141] Table 5

[0142]

[0143] In this embodiment, the historical battery swapping data of the battery swapping station is divided based on all pairwise combinations of level_label (set of order level labels) and var_label (set of order volatility labels). Multiple regression models are respectively constructed based on the divided historical battery swapping data of the battery swapping station. For example, it can be divided into 30 pieces of historical battery swapping data of the battery swapping station (i.e., data sets).

[0144] Furthermore, evaluate the 30 pieces of historical battery swapping data of the battery swapping station generated by pairwise combinations in the set of order level labels and the set of order volatility labels respectively. Each piece of historical battery swapping data of the battery swapping station has a root mean square error rmse and an accuracy rate R 2 evaluation value. Finally, comprehensively calculate the overall root mean square error rmse and accuracy rate R 2 value.

[0145] It should be noted that the overall evaluation index is calculated based on the predicted values and true values of the entire test set. If the overall evaluation index is less than the target set value, continue to optimize the parameters of the multiple regression model and continuously train the multiple regression model until the overall evaluation index reaches the target set value.

[0146] For example, when predicting data for more than two months from September 1, 2021 to November 9, 2021, there are more than 10,000 records of Beijing's battery swapping stations, almost covering all stations in Beijing. The evaluation indicators of the multiple regression model are good, where the root mean square error rmse = 16.7 and the accuracy rate R 2 = 86%; The above root mean square error and accuracy rate are used as the evaluation values of the trained multiple regression model; The trained multiple regression model is optimized based on the comparison result between the evaluation value and the target evaluation value.

[0147] Based on the historical battery swapping order data of the battery swapping station, the number of in-network vehicles, seasonal attributes, business parameters of the battery swapping station, etc., this embodiment predicts the future battery swapping demand (such as battery swapping order data or the number of battery swaps) of different vehicle models at the battery swapping station within a preset time. Specifically, the data samples are classified according to the order quantity and stability of the battery swapping station, the multiple regression model is trained, the optimal multiple regression model is selected according to the evaluation indicators, and finally the optimal multiple regression model is used to predict the battery swapping demand of different models at the battery swapping station within the future preset time, improving the prediction accuracy.

[0148] Embodiment 2

[0149] A battery swapping demand prediction system for a battery swapping station provided in this embodiment, as Figure 2 shown, the battery swapping demand prediction system includes: a first acquisition module 21, a training module 22, and a prediction module 23;

[0150] The first acquisition module 21 is used to acquire the historical battery swapping data of the battery swapping station. The historical battery swapping data of the battery swapping station includes a training set, and the training set includes the battery swapping station attribute data and vehicle data as inputs and the historical battery swapping order data as outputs;

[0151] In an optional embodiment, the battery swapping station attribute data (for example, the basic data of the battery swapping station) includes at least one of the battery swapping station version, supported vehicle models, number of business sites, start business hours, end business hours, preset time business duration, cumulative operation duration, city where the station is located, county where the station is located, station charging price, configuration quantity of each model of battery at the station, and number of station positions;

[0152] In this embodiment, the business sites include cities, counties, etc.

[0153] It should be noted that the preset time is set according to the actual situation. For example, the preset time can be set to 1 day or other values, and no specific limitation is made here.

[0154] In an alternative embodiment, the vehicle data (e.g., vehicle - end data) includes at least one of the cumulative number of network - connected vehicles, the distribution quantity of vehicle operation types, the number of active vehicles, the month - on - month increase in newly added vehicles, and the vehicle network - entry registration sites;

[0155] In this embodiment, the cumulative number of network - connected vehicles and the number of active vehicles are both counted by brand; the month - on - month increase in newly added vehicles is counted by brand and month.

[0156] In an alternative embodiment, the historical battery - swapping order data (e.g., consumption data) includes at least one of the order distribution of each package type, the average number of battery - swapping vehicles per day (e.g., the average number of battery - swapping vehicles in the most recent week), the average number of battery - swapping mileage per day (e.g., the average number of battery - swapping mileage in the most recent week), the number of battery - swapping orders per day, the number of battery - swapping mileage per day, and the average battery - swapping duration.

[0157] In this embodiment, the number of battery - swapping orders per day and the number of battery - swapping mileage per day are both counted by model.

[0158] The training module 22 is used to train a multiple - regression model based on a training set to obtain a trained multiple - regression model;

[0159] In this embodiment, in the actual business scenario, since factors such as weekdays, non - weekdays, and the seasons to which the battery - swapping station belongs will affect the battery - swapping demand of the battery - swapping station, when training the multiple - regression model, information such as weekdays, non - weekdays, and the seasons to which the battery - swapping station belongs is also included in the training set to train the multiple - regression model.

[0160] The prediction module is used to input the to - be - predicted battery - swapping data of the to - be - predicted battery - swapping station into the trained multiple - regression model to output the predicted battery - swapping demand of the to - be - predicted battery - swapping station. The predicted battery - swapping demand includes the predicted battery - swapping order data of the to - be - predicted battery - swapping station within a preset future time.

[0161] In this embodiment, the battery - swapping station attribute data and vehicle data are used as inputs, and the historical battery - swapping order data is used as the output to train the multiple - regression model. Then, the to - be - predicted battery - swapping data of the to - be - predicted battery - swapping station is input into the trained multiple - regression model to output the predicted battery - swapping demand of the to - be - predicted battery - swapping station, which can accurately predict the battery - swapping demand of the battery - swapping station in the future based on the multiple - regression model and improve the prediction accuracy.

[0162] In an alternative embodiment, as Figure 2 shown, the battery - swapping demand prediction system further includes: a data anomaly processing module 24 and a feature engineering processing module 25;

[0163] The data anomaly processing module 24 is used to perform data anomaly processing on the historical battery - swapping data of the battery - swapping station to obtain the historical battery - swapping data of the battery - swapping station after data anomaly processing;

[0164] In an optional embodiment, the data anomaly processing module 24 is used to perform missing data processing, discrete data processing, and normalization processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing.

[0165] In this embodiment, the missing data processing is to delete the data whose missing ratio is higher than the preset ratio (for example, the preset ratio can be 30%), and fill the missing ratio within the preset ratio (for example, 30%) with 0 or the mean value, such as battery swap orders, battery swap vehicle number, business rate, and those without corresponding records on the day are filled with 0 values;

[0166] Discrete data processing is to perform one-hot (one valid bit) encoding on the historical power exchange data of the power station, such as the power exchange model, season, working day, holiday, whether it is open all day, etc. Before one-hot encoding, it is shown in Table 1 in Example 1, and after one-hot encoding, it is shown in Table 2 in Example 1;

[0167] In order to eliminate the impact of dimension, the historical battery swap data of the battery swap station is normalized ([0,1]), such as the number of battery swap vehicles, the number of battery swap orders, the number of new vehicles connected to the network, the cumulative number of vehicles connected to the network, and other variables. The calculation formula for the normalization is shown in formula (1) in Example 1;

[0168] The feature engineering processing module 25 is used to perform feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing, so as to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0169] In an optional embodiment, the feature engineering processing module 25 is used to perform feature construction and filter variable processing on the historical battery swapping data of the battery swapping station after data anomaly processing, so as to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

[0170] In this embodiment, feature construction of historical battery swapping data of battery swapping stations after data anomaly processing includes the following contents:

[0171] The actual operating rate of the battery swap station (days) = actual operating hours / theoretical operating hours;

[0172] Whether it is open all day: If the end business time (seconds) - the start business time (seconds) = 86400, it is open all day, otherwise it is not open all day;

[0173] Cumulative operating hours of the battery swap station = the number of days between the statistical date and the first order of the battery swap station + 1;

[0174] Average value of battery swap orders in the past N days: the average value of orders from battery swap stations in the past N days;

[0175] Fluctuation value of battery replacement orders in the past N days: the ratio of the standard deviation of the order value in the past N days to the average of the order value in the past N days;

[0176] The historical battery swap data of the battery swap station after data anomaly processing is filtered for variables. Specifically, the variables with low correlation are filtered: This embodiment uses the variance filtering method for screening. If the variance of a feature or variable itself is very small, it means that the samples have basically no difference in this feature or variable. It is possible that most of the values ​​in the feature are the same, or even the values ​​of the entire feature are the same, then this feature has no effect on sample differentiation. This embodiment filters out variables with low correlation with the predicted target variable, such as the version of the battery swap station and the vehicle network registration site. At this point, the data used to construct the multivariate regression model is finally organized as shown in Table 3 in Example 1;

[0177] In this embodiment, data anomaly processing is the process of detecting, correcting or deleting inaccurate or inappropriate records for the multivariate regression model from the historical battery swap data of the battery swap station. Feature engineering processing is the process of converting the original historical battery swap data of the battery swap station into features that are more representative of the potential problems of the multivariate regression model. This can be achieved by selecting the most relevant features, extracting features, and creating features. In addition, feature engineering processing can reduce computing costs and improve the upper limit of the model.

[0178] In an optional embodiment, the historical battery swapping data of the battery swapping station also includes a test set; Figure 2 As shown, the battery replacement demand prediction system also includes: a testing module 26, a second acquisition module 27, a third acquisition module 28, and an optimization module 29;

[0179] The test module 26 is used to test the prediction results of the trained multivariate regression model using the test set to obtain the predicted battery replacement demand corresponding to the test set;

[0180] The second acquisition module 27 is used to obtain the actual battery replacement demand corresponding to the test set;

[0181] The third acquisition module 28 is used to obtain the root mean square error and accuracy of the trained multivariate regression model based on the predicted battery swapping demand corresponding to the test set and the actual battery swapping demand corresponding to the test set, so as to serve as the evaluation value of the trained multivariate regression model;

[0182] The optimization module 29 is used to optimize the trained multivariate regression model based on the comparison result between the evaluation value and the target evaluation value.

[0183] In the specific implementation process, a multivariate regression model is constructed through the OLS (ordinary least squares) algorithm to predict the battery swap demand of the battery swap station. OLS is a supervised machine learning model. The multivariate regression model is trained based on a large amount of historical battery swap data of the battery swap station to predict the predicted battery swap order data of the battery swap station within a preset time in the future. The supervised learning model needs to divide the historical battery swap data of the battery swap station into a training set and a test set. The training set is used to train the multivariate regression model, and the test set is used to evaluate the quality of the multivariate regression model.

[0184] The target variable (for example, battery replacement demand) is related to the time sequence. This embodiment divides the training set and the test set according to the timeline. For example, the historical battery replacement data of the battery replacement station from 2020-08-31 to T is used as the training set to predict the battery replacement demand (and battery replacement order data) of the site T+1, where T slides a window every day from 2021-08-31 to 2021-11-09, and the training set only takes data from the past year.

[0185] For example, the forecast of battery swapping demand by model at battery swapping stations on 2021-09-01;

[0186] Training set: historical battery swap data of battery swap stations from 2020-08-31 to 2021-08-31;

[0187] Test set: The test set needs to input the values ​​of all battery swapping demands on 2021-09-01, but 2021-09-01 is a future date, and the actual operating rate of the battery swap station, the number of new vehicles connected to the network, the cumulative number of vehicles connected to the network, etc. cannot be obtained. Therefore, this solution can use the data of 2021-08-31 for filling.

[0188] In addition, OLS predicts the dependent variable (i.e., the target variable) through a series of independent variables. The basic idea of ​​OLS is that the process of training the multiple regression model is to find a set of optimal learning parameters (a1, a2, ... an) so that the sum of square errors between the predicted value and the actual value of the multiple regression model is minimized. The mathematical expression of the OLS model is shown in formula (2) in Example 1;

[0189] In this embodiment, the evaluation of the multivariate regression model mainly measures the error between the predicted value and the true value (i.e., the error between the predicted battery replacement demand and the actual battery replacement demand). Common evaluation indicators of the multivariate regression model include root mean square error rmse, accuracy R 2 The root mean square error rmse is the square root of the ratio of the sum of the squares of the differences between the predicted value and the true value to the number of samples n. The smaller the root mean square error rmse is, the smaller the error between the predicted value and the true value of a set of data is. The calculation formula is shown in formula (3) in Example 1.

[0190] It should be noted that the accuracy R 2It measures how well the predicted value fits the true value, and the calculation formula is as shown in formula (4) in Embodiment 1;

[0191] In an optional embodiment, as Figure 2 shown, the battery swapping demand prediction system further includes: a label processing module 291;

[0192] The label processing module 291 is used to label the historical battery swapping data of the battery swapping station to obtain an order level label and an order fluctuation label.

[0193] In this embodiment, there are a large number of operating battery swapping stations, and there are multiple versions of the battery swapping stations. The maximum service capabilities (times / day) of different versions of the battery swapping stations are different, and the actual service capabilities of the battery swapping stations are also affected by the number of configured batteries in the station. Therefore, the range of daily battery swapping order data (such as the number of battery swapping orders) of the battery swapping stations is between [0, 900), and the interval span is large. Moreover, the daily stability of the battery swapping order numbers of different battery swapping stations is also different, and the accuracy of the previous model prediction is generally average.

[0194] In the case of inconsistent battery swapping order numbers and stabilities, in order to improve the model accuracy, this embodiment needs to perform a finer-grained division of the historical battery swapping data of the battery swapping stations in combination with the battery swapping order numbers and order stabilities, and establish multiple regression models for training on the divided historical battery swapping data of the battery swapping stations. To achieve this goal, first, it is necessary to label the historical battery swapping data of the battery swapping stations with order level labels and order volatility (measuring stability) labels. The method of labeling the historical battery swapping data of the battery swapping stations is described as follows:

[0195] Take the battery swapping order numbers of the battery swapping station itself in the recent N days to calculate the average value and fluctuation value of each battery swapping station. Sort the order average value and fluctuation value from small to large respectively, and perform segmented processing according to the quantiles and actual business experience, and label values according to the intervals. For example, Table 4 in Embodiment 1 is the order level label and the order average value corresponding to the order level label, and Table 5 in Embodiment 1 is the order volatility label and the order average value corresponding to the order volatility label. For example, sort the average values of the battery swapping orders of the battery swapping station itself in the recent 30 days from small to large, and combine the quantiles and actual business experience to extract several demarcation points of 100, 300, 500, and 800. One interval is formed between two demarcation points, and labels are given to each interval (the same for the volatility label). Finally, there are a total of 6 order level labels and 5 order volatility labels.

[0196] In this embodiment, the historical battery swapping data of the battery swapping stations is divided based on all pairwise combinations of level_label (order level label set) and var_label (order volatility label set), and multiple regression models are respectively constructed based on the divided historical battery swapping data of the battery swapping stations. For example, it can be divided into 30 historical battery swapping data of the battery swapping stations (i.e., data sets).

[0197] Furthermore, 30 pieces of historical battery swapping data of battery swapping stations generated by pairwise combinations in the order level tag set and the order volatility tag set are respectively evaluated. Each piece of historical battery swapping data of a battery swapping station has 1 root mean square error rmse and accuracy rate R 2 evaluation value. Finally, the overall root mean square error rmse and accuracy rate R 2 value are calculated comprehensively.

[0198] It should be noted that the overall evaluation index is calculated based on the predicted values and true values of the entire test set. If the overall evaluation index is less than the target set value, the parameters of the multiple regression model are continuously optimized, and the multiple regression model is continuously trained until the overall evaluation index reaches the target set value.

[0199] For example, when predicting data for more than two months from September 1, 2021 to November 9, 2021, there are more than 10,000 records of battery swapping stations in Beijing, covering almost all stations in Beijing. The evaluation index of the multiple regression model is good, where the root mean square error rmse = 16.7 and the accuracy rate R 2 = 86%; the above root mean square error and accuracy rate are used as the evaluation values of the trained multiple regression model; the trained multiple regression model is optimized based on the comparison result between the evaluation value and the target evaluation value.

[0200] This embodiment predicts the future preset-time battery swapping demand (such as battery swapping order data or the number of battery swapping times) of a battery swapping station by vehicle type based on the historical battery swapping order data of the battery swapping station, the number of on-network vehicles, seasonal attributes, business parameters of the battery swapping station, etc. Specifically, the data samples are classified according to the order quantity and stability of the battery swapping station, the multiple regression model is trained, the optimal multiple regression model is selected according to the evaluation index, and finally the optimal multiple regression model is used to predict the battery swapping demand by vehicle type within the future preset time of the battery swapping station, improving the prediction accuracy.

[0201] Embodiment 3

[0202] Figure 3 FIG. 3 is a schematic structural diagram of an electronic device provided in Embodiment 3 of the present invention. The electronic device includes a memory, a processor, and a computer program stored on the memory and for running on the processor. When the processor executes the program, it implements the battery swapping demand prediction method of the battery swapping station in Embodiment 1. Figure 3 The shown electronic device 30 is only an example and should not bring any limitation to the functions and usage scope of the embodiments of the present invention.

[0203] Such as Figure 3As shown, the electronic device 30 may be embodied in the form of a general-purpose computing device. For example, it may be a server device. The components of the electronic device 30 may include, but are not limited to: at least one of the above-mentioned processors 31, at least one of the above-mentioned memories 32, and a bus 33 connecting different system components (including the memory 32 and the processor 31).

[0204] The bus 33 includes a data bus, an address bus, and a control bus.

[0205] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.

[0206] The memory 32 may further include a program / utility 325 having a set (at least one) of program modules 324. Such program modules 324 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment.

[0207] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32, such as the method for predicting the battery swapping demand of the battery swapping station in Embodiment 1 of the present invention.

[0208] The electronic device 30 may also communicate with one or more external devices 34 (such as a keyboard, a pointing device, etc.). Such communication may be carried out through an input / output (I / O) interface 35. Moreover, the device 30 for generating a model may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 36. As Figure 3 shown, the network adapter 36 communicates with other modules of the device 30 for generating a model through the bus 33. It should be understood that although not shown in the figure, other hardware and / or software modules may be used in conjunction with the device 30 for generating a model, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0209] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, such a division is merely exemplary and not mandatory. In fact, according to the embodiments of the present invention, the features and functions of two or more of the above-mentioned units / modules may be embodied in one unit / module. Conversely, the features and functions of one unit / module described above may be further divided and embodied by multiple units / modules.

[0210] Embodiment 4

[0211] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for predicting the battery replacement demand of the battery replacement station provided in Embodiment 1.

[0212] Among them, the more specific forms that the readable storage medium can adopt may include, but are not limited to: portable disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories, optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0213] In a possible implementation manner, the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to cause the terminal device to execute the method for predicting the battery replacement demand of the battery replacement station described in Embodiment 1.

[0214] Among them, the program code for executing the present invention can be written in any combination of one or more programming languages. The program code can be executed entirely on the user device, partially on the user device, executed as an independent software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0215] Although the specific implementation manners of the present invention have been described above, those skilled in the art should understand that this is only an example. The protection scope of the present invention is defined by the appended claims. Without departing from the principles and essence of the present invention, those skilled in the art can make various changes or modifications to these implementation manners, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A method for predicting the battery replacement demand of a battery replacement station, characterized in that, The battery swapping demand prediction method includes: Obtain the historical battery swapping data of the battery swapping station. The historical battery swapping data of the battery swapping station includes a training set, and the training set includes the battery swapping station attribute data and vehicle data as inputs, and the historical battery swapping order data as an output; Train a multiple regression model based on the training set to obtain a trained multiple regression model; Input the to-be-predicted battery swapping data of the to-be-predicted battery swapping station into the trained multiple regression model to output the predicted battery swapping demand of the to-be-predicted battery swapping station. The predicted battery swapping demand includes the predicted battery swapping order data of the to-be-predicted battery swapping station within a preset future time.

2. The method for predicting the battery replacement demand of a battery replacement station according to claim 1, wherein The battery swapping demand prediction method further includes: Perform data anomaly processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing; Perform feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

3. The method for predicting the battery replacement demand of a battery replacement station according to claim 1, wherein, The historical battery swapping data of the battery swapping station further includes a test set; the battery swapping demand prediction method further includes: Use the test set to test the prediction result of the trained multiple regression model to obtain the predicted battery swapping demand corresponding to the test set; Obtain the actual battery swapping demand corresponding to the test set; Obtain the root mean square error and accuracy of the trained multiple regression model based on the predicted battery swapping demand corresponding to the test set and the actual battery swapping demand corresponding to the test set as the evaluation value of the trained multiple regression model; Optimize the trained multiple regression model based on the comparison result between the evaluation value and the target evaluation value.

4. The method for predicting the battery replacement demand of a battery replacement station according to claim 1, wherein, The battery swapping demand prediction method further includes: Label the historical battery swapping data of the battery swapping station to obtain an order level label and an order volatility label.

5. The method for predicting the battery replacement demand of a battery replacement station according to claim 2, wherein, The step of performing data anomaly processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing includes: Perform missing data processing, discrete data processing, and normalization processing on the historical battery swapping data of the battery swapping station to obtain the historical battery swapping data of the battery swapping station after data anomaly processing.

6. The method for predicting the battery replacement demand of a battery replacement station according to claim 2, wherein, The step of performing feature engineering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing includes: Perform feature construction and variable filtering processing on the historical battery swapping data of the battery swapping station after data anomaly processing to obtain the historical battery swapping data of the battery swapping station after feature engineering processing.

7. The method for predicting the battery replacement demand of a battery replacement station according to claim 1, wherein, The battery swapping station attribute data includes at least one of the battery swapping station version, supported vehicle models, number of business sites, start business hours, end business hours, business hours during preset time, cumulative operation duration, city where the station is located, county where the station is located, station charging price, configuration quantity of each model of battery at the station, and number of station positions; And / or, The vehicle data includes at least one of the cumulative number of networked vehicles, distribution quantity of vehicle operation types, number of active vehicles, vehicle new growth month-on-month, and vehicle network entry registration sites; And / or, The historical battery swapping order data includes at least one of the order distributions of each package type, the daily average number of battery swapping vehicles, the daily average number of battery swapping mileage, the number of battery swapping orders per day, the number of battery swapping mileage per day, and the average battery swapping duration.

8. A power swapping demand prediction system for a power swapping station, characterized in that, The battery swapping demand prediction system includes: A first acquisition module, configured to acquire historical battery swapping data of a battery swapping station, where the historical battery swapping data of the battery swapping station includes a training set, and the training set includes battery swapping station attribute data and vehicle data as inputs and historical battery swapping order data as outputs; A training module, configured to train a multiple regression model based on the training set to obtain a trained multiple regression model; A prediction module, configured to input the to-be-predicted battery swapping data of the to-be-predicted battery swapping station into the trained multiple regression model to output the predicted battery swapping demand of the to-be-predicted battery swapping station, where the predicted battery swapping demand includes predicted battery swapping order data of the to-be-predicted battery swapping station within a preset future time.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and adapted to run on the processor, characterized in that, When the processor executes the computer program, it implements the battery swapping demand prediction method for a battery swapping station as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the battery swapping demand prediction method for a battery swapping station as described in any one of claims 1-7.

Images