Model training method, battery replacement time prediction method, and battery replacement demand prediction method and system

Abstract

The invention discloses a model training method, a battery replacement time prediction method and a battery replacement demand prediction method and system, and the method comprises the steps: obtaining a training set which comprises the historical battery replacement time interval of a battery replacement vehicle; and training the network model based on the historical battery replacement time interval of the battery replacement vehicle to obtain a battery replacement time prediction model of the battery replacement vehicle. According to the method, the network model is trained through the obtained historical battery replacement time interval of the battery replacement vehicle in the training set, so that the battery replacement time prediction model of the battery replacement vehicle is obtained, the accuracy and prediction precision of model prediction are improved, the accurate prediction of the battery replacement demand of the battery replacement station by combining the battery replacement time prediction model of the battery replacement vehicle is realized, and the prediction efficiency is improved. And beforehand preparation and flexible adjustment of the battery replacement station are facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of battery swapping in battery swapping stations, and particularly to a model training method, a battery swapping time prediction method, a battery swapping demand prediction method, and a system. Background Art

[0002] With the vigorous promotion and popularization of new energy vehicles, more and more vehicle manufacturers and drivers have joined the battery swapping network. With the development of the battery swapping industry, the construction of battery swapping stations has also opened a new chapter. However, in the context of the increasing and complex battery swapping scenarios and demands, higher requirements are also placed on the daily operation of battery swapping stations.

[0003] In the prior art during the operation of battery swapping stations, generally, the needs of vehicles and battery swapping stations are judged based on the experience of staff. It is impossible to predict the battery swapping patterns of battery swapping vehicles, nor can it predict the battery swapping demands of battery swapping stations, and thus it is difficult to conduct refined management of battery swapping stations. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the defect in the prior art that refined management of the battery swapping demands of battery swapping stations cannot be achieved, and to provide a model training method, a battery swapping time prediction method, a battery swapping demand prediction method, and a system.

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

[0006] The first aspect of the present invention provides a model training method, including:

[0007] Obtain a training set, where the training set includes the historical battery swapping time intervals of battery swapping vehicles;

[0008] Train a network model based on the historical battery swapping time intervals of the battery swapping vehicles to obtain a battery swapping time prediction model for the battery swapping vehicles.

[0009] In this embodiment, by training a network model with the historical battery swapping time intervals in the obtained training set to obtain a battery swapping time prediction model for battery swapping vehicles, the accuracy and precision of model prediction are improved.

[0010] Preferably, before the step of obtaining the training set, the model training method further includes:

[0011] Obtain the historical battery swapping data set of the battery swapping vehicles;

[0012] Conduct distribution analysis and abnormal data analysis and processing on the historical battery swapping data set to obtain a processed historical battery swapping data set;

[0013] Conduct feature construction and feature engineering on the processed historical battery swapping data set to obtain the training set.

[0014] In this embodiment, distribution analysis and abnormal data analysis and processing are performed on the historical battery replacement data set to obtain the processed historical battery replacement data set, improving the accuracy of the data. Through feature construction and feature engineering on the processed historical battery replacement data set, a training set is obtained; enabling the performance of the battery replacement time prediction model on unknown data to reach the optimal or near-optimal performance.

[0015] Preferably, the step of performing feature construction and feature engineering on the processed historical battery replacement data set to obtain the training set includes:

[0016] Dividing the training set, validation set, and test set according to a preset ratio;

[0017] Using the validation set to verify the battery replacement time prediction model of the battery replacement vehicle to obtain the verified battery replacement time prediction model of the battery replacement vehicle;

[0018] Using the test set to test the prediction results of the battery replacement time prediction model of the battery replacement vehicle.

[0019] In this embodiment, using the training set to train the battery replacement time prediction model of the battery replacement vehicle and using the validation set to verify the battery replacement time prediction model of the battery replacement vehicle can obtain a relatively good battery replacement time prediction model of the battery replacement vehicle; using the test set to test the prediction results of the battery replacement time prediction model of the battery replacement vehicle verifies the effect of the battery replacement time prediction model of the battery replacement vehicle.

[0020] Preferably, the step of performing abnormal data analysis and processing on the historical battery replacement data set includes:

[0021] Obtaining the historical battery replacement time interval, the difference in the state of charge (SOC) of the replaced battery, the difference in the replaced battery power, and the difference in the replaced battery mileage between any two adjacent battery replacement times of the battery replacement vehicle;

[0022] Identifying abnormal data based on the comparison results of the historical battery replacement time interval, the difference in the SOC of the replaced battery, the difference in the replaced battery power, and the difference in the replaced battery mileage with their respective preset thresholds.

[0023] In this embodiment, identifying abnormal data based on the comparison results of the obtained historical battery replacement time interval, the difference in the SOC of the replaced battery, the difference in the replaced battery power, and the difference in the replaced battery mileage with their respective preset thresholds can accurately identify abnormal data and improve the accuracy of the data.

[0024] The second aspect of the present invention provides a method for predicting the battery replacement time of a battery replacement vehicle, including:

[0025] Obtaining the historical battery replacement time interval of the battery replacement vehicle as the input of the battery replacement time prediction model of the battery replacement vehicle;

[0026] Predict the battery swapping time interval of the battery swapping vehicle by using the battery swapping time prediction model of the battery swapping vehicle to obtain the predicted battery swapping time interval of the battery swapping vehicle;

[0027] Obtain the upper limit value and the lower limit value of the predicted battery swapping time interval according to the predicted battery swapping time interval;

[0028] Obtain the predicted battery swapping probability value of the battery swapping vehicle according to the predicted battery swapping time interval, the upper limit value and the lower limit value;

[0029] Wherein, the battery swapping time prediction model of the battery swapping vehicle is trained by using the model training method described in the first aspect.

[0030] In this embodiment, the future battery swapping time interval is predicted based on historical battery swapping data, and different prediction methods are adopted for the historical battery swapping data of different battery swapping vehicles. Combining the predicted number of days until the next battery swap, as well as the upper and lower limit values, it is possible to accurately output the daily battery swapping probability values of the battery swapping vehicle within a preset future time starting from the current battery swapping time.

[0031] The third aspect of the present invention provides a method for predicting the battery swapping demand of a battery swapping station, including:

[0032] Predict the predicted battery swapping probability values of multiple battery swapping vehicles by using the battery swapping vehicle battery swapping time prediction method described in the second aspect;

[0033] Obtain the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations in the historical time period;

[0034] According to the predicted battery swapping probability values of the multiple battery swapping vehicles and the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations, statistically calculate the predicted battery swapping demand of each battery swapping station within a preset time interval, and the predicted battery swapping demand includes the number of battery swapping vehicles whose predicted battery swapping time interval is within the preset time interval.

[0035] In this embodiment, the predicted battery swapping probability values of multiple battery swapping vehicles are predicted by using the battery swapping vehicle battery swapping time prediction method; according to the predicted battery swapping probability values of the multiple battery swapping vehicles and the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations, statistically calculate the predicted battery swapping demand of each battery swapping station within a preset time interval; realizing the accurate prediction of the battery swapping demand of the battery swapping station by combining the battery swapping time prediction model of the battery swapping vehicle, which facilitates the prior preparation and flexible adjustment of the battery swapping station.

[0036] Preferably, the step of obtaining the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations in the historical time period includes:

[0037] Obtain the total historical battery swapping times of the multiple battery swapping vehicles and the total historical battery swapping times of the different battery swapping stations in the historical time period;

[0038] Obtain the ratio of the number of battery replacements of each battery replacement vehicle to different battery replacement stations according to the total historical number of battery replacements of the multiple battery replacement vehicles and the total historical number of battery replacements of different battery replacement stations;

[0039] Obtain the battery replacement probability values of the multiple battery replacement vehicles to different battery replacement stations during the historical time period according to the ratio of the number of battery replacements and the weight coefficient of each battery replacement station.

[0040] In this embodiment, obtaining the battery replacement probability values of multiple battery replacement vehicles to different battery replacement stations during the historical time period according to the ratio of the number of battery replacements and the weight coefficient of each battery replacement station can perform queuing reminders based on the battery replacement vehicle data, realizing the prediction from the battery replacement vehicle to the battery replacement station, that is, the battery replacement demand interval of the battery replacement vehicle and the demand interval of the battery replacement station can be obtained, so as to suggest whether the battery replacement vehicle needs to delay battery replacement according to the actual battery power of the battery replacement vehicle to avoid high electricity price time periods, thereby reducing the charging cost.

[0041] The fourth aspect of the present invention provides a model training system, including:

[0042] A training set acquisition module, configured to acquire a training set, where the training set includes the historical battery replacement time intervals of battery replacement vehicles;

[0043] A training module, configured to train a network model based on the historical battery replacement time intervals of the battery replacement vehicles to obtain a battery replacement time prediction model for the battery replacement vehicles.

[0044] In this embodiment, by training a network model with the historical battery replacement time intervals in the acquired training set to obtain a battery replacement time prediction model for the battery replacement vehicles, the accuracy and precision of the model prediction are improved.

[0045] Preferably, the model training system further includes:

[0046] A historical battery replacement data set acquisition module, configured to acquire the historical battery replacement data set of the battery replacement vehicles;

[0047] A first processing module, configured to perform distribution analysis and abnormal data analysis processing on the historical battery replacement data set to obtain a processed historical battery replacement data set;

[0048] A second processing module, configured to perform feature construction and feature engineering on the processed historical battery replacement data set to obtain the training set.

[0049] In this embodiment, the historical battery swapping data set is analyzed for distribution and abnormal data analysis and processing to obtain a processed historical battery swapping data set, improving the accuracy of the data. By performing feature construction and feature engineering on the processed historical battery swapping data set, a training set is obtained, enabling the battery swapping time prediction model to achieve optimal or near-optimal performance on unknown data.

[0050] Preferably, the second processing module includes:

[0051] A partitioning unit for partitioning the training set, validation set, and test set according to a preset ratio;

[0052] A validation unit for validating the battery swapping time prediction model of the battery swapping vehicle using the validation set to obtain a validated battery swapping time prediction model of the battery swapping vehicle;

[0053] A testing unit for testing the prediction results of the battery swapping time prediction model of the battery swapping vehicle using the test set.

[0054] In this embodiment, training the battery swapping time prediction model of the battery swapping vehicle using the training set and validating the battery swapping time prediction model of the battery swapping vehicle using the validation set can obtain a relatively good battery swapping time prediction model of the battery swapping vehicle. Testing the prediction results of the battery swapping time prediction model of the battery swapping vehicle using the test set verifies the effectiveness of the battery swapping time prediction model of the battery swapping vehicle.

[0055] Preferably, the first processing module includes:

[0056] A first acquisition unit for acquiring the historical battery swapping time intervals, differences in battery swapping battery SOC, differences in battery swapping power, and differences in battery swapping mileage between any two adjacent battery swapping times of the battery swapping vehicle;

[0057] An identification unit for identifying abnormal data based on the comparison results of the historical battery swapping time intervals, differences in battery swapping battery SOC, differences in battery swapping power, and differences in battery swapping mileage with their respective preset thresholds.

[0058] In this embodiment, identifying abnormal data based on the comparison results of the acquired historical battery swapping time intervals, differences in battery swapping battery SOC, differences in battery swapping power, and differences in battery swapping mileage with their respective preset thresholds can accurately identify abnormal data and improve the accuracy of the data.

[0059] The fifth aspect of the present invention provides a battery swapping time prediction system for a battery swapping vehicle, including:

[0060] A battery swapping time interval acquisition module for acquiring the historical battery swapping time intervals of the battery swapping vehicle as the input of the battery swapping time prediction model of the battery swapping vehicle;

[0061] The first prediction module is used to predict the replacement time interval of the replacement vehicle by using the replacement time prediction model of the replacement vehicle, so as to obtain the predicted replacement time interval of the replacement vehicle;

[0062] The first acquisition module is used to obtain the upper limit value and the lower limit value of the predicted replacement time interval according to the predicted replacement time interval;

[0063] The second acquisition module is used to obtain the predicted replacement probability value of the replacement vehicle according to the predicted replacement time interval, the upper limit value and the lower limit value;

[0064] Wherein, the replacement time prediction model of the replacement vehicle is trained by using the model training system described in the fourth aspect.

[0065] In this embodiment, the future replacement time interval is predicted based on historical replacement data, and different prediction methods are adopted for the historical replacement data of different replacement vehicles. Combining the predicted number of days until the next replacement and the upper and lower limit values, the replacement probability value of each day within a preset future time from the current replacement time of the replacement vehicle can be accurately output.

[0066] The sixth aspect of the present invention provides a replacement demand prediction system for a replacement station, including:

[0067] The second prediction module is used to predict the predicted replacement probability values of multiple replacement vehicles by using the replacement vehicle replacement time prediction system described in the fifth aspect;

[0068] The third acquisition module is used to obtain the replacement probability values of the multiple replacement vehicles to different replacement stations in a historical time period;

[0069] The statistics module is used to count the predicted replacement demand of each replacement station within a preset time interval according to the predicted replacement probability values of the multiple replacement vehicles and the replacement probability values of the multiple replacement vehicles to different replacement stations, and the predicted replacement demand includes the number of replacement vehicles whose predicted replacement time interval is within the preset time interval.

[0070] In this embodiment, the predicted replacement probability values of multiple replacement vehicles are predicted by using the replacement vehicle replacement time prediction method; the predicted replacement demand of each replacement station within a preset time interval is counted according to the predicted replacement probability values of the multiple replacement vehicles and the replacement probability values of the multiple replacement vehicles to different replacement stations; realizing the accurate prediction of the replacement demand of the replacement station by combining the replacement time prediction model of the replacement vehicle, which facilitates the prior preparation and flexible adjustment of the replacement station.

[0071] Preferably, the third acquisition module includes:

[0072] A second acquisition unit, configured to acquire the total historical battery swapping times of the multiple battery swapping vehicles and the total historical battery swapping times of different battery swapping stations during the historical time period;

[0073] A third acquisition unit, configured to obtain the ratio of the battery swapping times of each battery swapping vehicle to different battery swapping stations according to the total historical battery swapping times of the multiple battery swapping vehicles and the total historical battery swapping times of different battery swapping stations;

[0074] A fourth acquisition unit, configured to obtain the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations during the historical time period according to the ratio of the battery swapping times and the weight coefficient of each battery swapping station.

[0075] In this embodiment, obtaining the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations during the historical time period according to the ratio of the battery swapping times and the weight coefficient of each battery swapping station can perform queuing reminders based on the battery swapping vehicle data, realizing the prediction from the battery swapping vehicles to the battery swapping stations, that is, the battery swapping demand intervals of the battery swapping vehicles and the demand intervals of the battery swapping stations can be obtained, so as to recommend whether the battery swapping vehicles need to delay battery swapping according to the actual battery levels of the battery swapping vehicles to avoid high electricity price time periods, thereby reducing the charging cost.

[0076] The seventh 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, it implements at least one of the model training method described in the first aspect, the battery swapping time prediction method for battery swapping vehicles described in the second aspect, and the battery swapping demand prediction method for battery swapping stations described in the third aspect.

[0077] The eighth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. The computer program, when executed by a processor, implements at least one of the model training method described in the first aspect, the battery swapping time prediction method for battery swapping vehicles described in the second aspect, and the battery swapping demand prediction method for battery swapping stations described in the third aspect.

[0078] On the basis of conforming to the common knowledge in the art, the above preferred conditions can be combined arbitrarily to obtain various preferred embodiments of the present invention.

[0079] The positive and progressive effects of the present invention are as follows:

[0080] The present invention trains a network model through the historical battery swapping time intervals of battery swapping vehicles in the obtained training set to obtain a battery swapping time prediction model for battery swapping vehicles, improving the accuracy and prediction precision of the model prediction, and further realizing the accurate prediction of the battery swapping demand of the battery swapping station in combination with the battery swapping time prediction model of the battery swapping vehicle, facilitating the prior preparation and flexible adjustment of the battery swapping station. Description of the Drawings

[0081] Figure 1 Flow chart of the model training method according to Embodiment 1 of the present invention.

[0082] Figure 2 Schematic diagram of the modules of the model training system according to Embodiment 2 of the present invention.

[0083] Figure 3 Schematic diagram of the structure of the electronic device according to Embodiments 3, 7 and 11 of the present invention.

[0084] Figure 4 Flow chart of the method for predicting the battery swapping time of a battery swapping vehicle according to Embodiment 5 of the present invention.

[0085] Figure 5 Schematic diagram of the modules of the system for predicting the battery swapping time of a battery swapping vehicle according to Embodiment 6 of the present invention.

[0086] Figure 6 Flow chart of the method for predicting the battery swapping demand of a battery swapping station according to Embodiment 9 of the present invention.

[0087] Figure 7 Schematic diagram of the modules of the system for predicting the battery swapping demand of a battery swapping station according to Embodiment 10 of the present invention. Detailed implementation manners

[0088] 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.

[0089] Embodiment 1

[0090] A model training method provided in this embodiment is as Figure 1 shown and includes:

[0091] Step 101, obtain a training set, where the training set includes the historical battery swapping time intervals of battery swapping vehicles;

[0092] Step 102, train a network model based on the historical battery swapping time intervals of battery swapping vehicles to obtain a battery swapping time prediction model for battery swapping vehicles.

[0093] In an optional embodiment, before step 101, the model training method further includes:

[0094] Step 100, obtain the historical battery swapping data set of battery swapping vehicles;

[0095] In this embodiment, the historical battery swapping data set of battery swapping vehicles includes at least one of vehicle battery swapping order data, vehicle basic information, and station-side basic information.

[0096] The vehicle battery swapping order data includes at least one of payment time, vehicle number, vehicle code, station code, city ID, order type, battery code, order validity, consumption type, mileage, order status, operation type, battery SOC at the time of battery swapping in and out, time, and power.

[0097] The vehicle basic information includes at least one of operation type, vehicle model, account status, vehicle type, and the city to which it belongs.

[0098] The station terminal basic information includes at least one of station terminal code, station terminal operation type, the city to which it belongs, business hours, and supported vehicle models.

[0099] Step 100-1: Perform distribution analysis and abnormal data analysis and processing on the historical battery swapping data set to obtain the processed historical battery swapping data set.

[0100] In this embodiment, the historical battery swapping data set is screened based on whether a certain vehicle needs battery swapping. For example, the most recent battery swapping records are selected for analysis. However, abnormal data may also appear in the time intervals of the most recent battery swapping records. For example, among the most recent 10 battery swapping records, if the time interval of a certain battery swapping becomes dozens of days or hundreds of days, etc., then the data with a relatively large time interval is considered abnormal data. Usually, the time intervals of each battery swapping are relatively small (for example, battery swapping once every 10 days), and those with a large difference are considered abnormal data. The abnormal data is found and removed.

[0101] In this embodiment, combined with the data distribution situation, statistical analysis is performed on the abnormal data. The data distribution mainly uses the 20th percentile, and the abnormal data is mainly judged in combination with the business situation. For example, battery swapping should be carried out once within 30 days, and the differences in battery SOC, power, and battery swapping mileage between two battery swappings should not be too low or too high.

[0102] Step 100-2: Perform feature construction and feature engineering on the processed historical battery swapping data set to obtain a training set.

[0103] In this embodiment, feature selection is performed from the processed historical battery swapping data set to know which feature data can be used to estimate the battery swapping demand of battery swapping vehicles on a certain future day. Specifically, different algorithms are used to construct different features. For example, time series analysis algorithms are used to construct time-related feature data. After feature construction, further data analysis needs to be performed on the constructed features to finally determine the useful feature data. It should be noted that this process is a cyclic process. After all features are cyclically analyzed, the final useful feature data can be obtained. In addition, the algorithms used in this embodiment are related to the selected features. If different features need to be obtained, algorithms corresponding to the features can be used according to the requirements. A certain feature may be obtained through one or more algorithms.

[0104] In this embodiment, feature engineering is to convert data into features that can better represent potential problems (for example, converting the processed historical battery swap data set into features), thereby improving machine learning capabilities. Furthermore, by converting the processed historical battery swap data set into features through feature engineering, these features can well describe the processed historical battery swap data set, and the performance of the model established using them on unknown data can achieve optimal or near-optimal performance.

[0105] It should be noted that the better the features, the greater the flexibility, the simpler the model built, and the better the performance of the model.

[0106] In addition, the time series analysis algorithm can be adjusted according to actual conditions.

[0107] In the specific implementation process, the following contents of the historical battery swapping data set of the battery swapping vehicle are obtained: city ID, vehicle ID, consumption type, and vehicle type;

[0108] Total battery swap times: calculated based on the battery swap vehicle dimension, multiple battery swaps in a day are counted as one;

[0109] Time since the first battery swap: calculated by the battery swap vehicle dimension, the number of days from the earliest battery swap record to the current date;

[0110] Time since the last battery swap: calculated based on the battery swap vehicle dimension, the number of days since the last battery swap record is from the current date;

[0111] Total number of battery swaps in the past year: calculated based on the battery swap vehicle dimension, the total number of battery swaps in the past year, multiple battery swaps in one day are counted as one;

[0112] Average of the middle 50% quantiles - before correction: Sort all the battery swap time intervals of the battery swap vehicles, take the middle 50% quantiles, and then calculate the average quantile_avg (average quantile);

[0113] Lower quartile - before correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, and take the lower quartile;

[0114] Upper quartile - before correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, and take the upper quartile;

[0115] Average of the middle 50% quantiles - after correction: Calculate and sort all battery swap time intervals by the battery swap vehicle dimension, eliminate abnormal battery swap records, and then take the middle 50% quantiles of all records after eliminating the corrections, and calculate the average quantile_avg.

[0116] Among them, the judgment process of abnormal records: all battery replacement time intervals are compared with the "middle 50% quantile average - before correction". Records with a battery replacement time interval greater than "middle 50% quantile average - before correction" * 10 and greater than 40 days and all battery replacement records of the battery replacement vehicle before the date are abnormal records. The replacements here are: 50% can be replaced by 70%, 80%; the average can be replaced by the weighted average, mode, etc.; * 10 can be replaced by (Q3-Q1) * 1.5 times, greater than 40 days can be replaced by 50 days, etc.);

[0117] Lower quartile - before correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, sort by battery swap time interval, and take the lower quartile;

[0118] Upper quartile - before correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, sort by battery swap time interval, and take the upper quartile;

[0119] Minimum value of all records - after correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, remove abnormal battery swapping records, and take the minimum value;

[0120] Maximum value of all records - after correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, remove abnormal battery swapping records, and take the maximum value;

[0121] Average value of all records - after correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, remove abnormal battery swapping records, and take the average value;

[0122] Average of the battery swap time intervals of the last three battery swap records - after correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, and take the moving average of the battery swap time intervals of the last three battery swap records;

[0123] This embodiment can refer to the moving average algorithm in the time series. The simple moving average algorithm used here can be replaced by weighted moving average method, trend moving average method, exponential smoothing method, and differential exponential smoothing method.

[0124] Most recent battery replacement time - after correction: Calculate all battery replacement time intervals by battery replacement vehicle dimension and sort them, and take the most recent battery replacement time;

[0125] The battery replacement time interval of the most recent battery replacement record - after correction: calculate and sort all battery replacement time intervals by the battery replacement vehicle dimension, remove abnormal battery replacement records, and take the battery replacement time interval corresponding to the most recent battery replacement record;

[0126] Total battery swap times - after correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, and summarize the battery swap times;

[0127] Whether it is stable - after correction (is_steady): Calculate the sorting of all battery swapping time intervals according to the battery swapping vehicle dimension, eliminate abnormal battery swapping records, and then perform stability detection and judgment.

[0128] Stability detection logic: ADF test in unit root test in time series analysis. Replaceable methods here include: time series plot, autocorrelation plot and partial autocorrelation plot, DF test, PP test, DF-GLS test, KPSS test, etc.

[0129] In an optional embodiment, step 100-2 includes:

[0130] Step 100-21: Divide the training set, validation set and test set according to a preset ratio;

[0131] In this embodiment, the preset ratio is set according to the actual situation. For example, divide the constructed feature data into a training set, a validation set and a test set according to a ratio, such as dividing the feature data into a training set + validation set: test set according to 8:2 or 9:1, and then divide the training set and the validation set according to 2:8.

[0132] It should be noted that the training set, validation set and test set are only randomly divided according to the ratio. When the quantity ratio is quite different, stratified sampling will be adopted.

[0133] Step 100-22: Use the validation set to verify the battery swapping time prediction model of the battery swapping vehicle to obtain the verified battery swapping time prediction model of the battery swapping vehicle;

[0134] Step 100-23: Use the test set to test the prediction result of the battery swapping time prediction model of the battery swapping vehicle.

[0135] In the specific implementation process, train the network model based on the training set. The network model includes but is not limited to the regression model and the decision tree model. Specifically, train the regression model and the decision tree model respectively based on the training set, obtain two trained models respectively, and then combine these two models to finally obtain the battery swapping time prediction model of the battery swapping vehicle;

[0136] It should be noted that the trained model can be replaced, and multiple models can be trained at the same time. Finally, the battery swapping time prediction model of the battery swapping vehicle is obtained after combination. The above-mentioned trained battery swapping time prediction model of the battery swapping vehicle can predict which time period the battery swapping vehicle will go to swap batteries in the future, that is, predict the battery swapping time of the battery swapping vehicle.

[0137] In addition, the training set and the validation set are mainly used in the process of training the network model. Specifically, the training set is used to train the network model to obtain the prediction model for the battery swapping time of battery swapping vehicles. Then, the prediction model for the battery swapping time of battery swapping vehicles is verified based on the validation set to obtain a better prediction model for the battery swapping time of battery swapping vehicles (such as electric vehicles). The test set is used to verify the effect of the prediction model for the battery swapping time of battery swapping vehicles. Finally, the battery swapping demand interval of the battery swapping vehicle is predicted through the prediction model for the battery swapping time of the battery swapping vehicle (that is, it is predicted that a certain battery swapping vehicle is expected to come for battery swapping in a certain period of time in the future. For example, it is expected to come for battery swapping in the next 1-5 days). The test set is the time period in the future when it has been determined that the battery swapping vehicle will come back for battery swapping. Therefore, the test set is used to verify whether the result of the battery swapping time of the battery swapping vehicle predicted by the prediction model for the battery swapping time of the battery swapping vehicle is consistent with the time in the test set;

[0138] Then, from the dimension of the battery swapping vehicle, the estimated battery swapping demand of a certain battery swapping station in a certain period of time in the future is predicted, and which station the battery swapping vehicle will go to for battery swapping is determined according to the recent battery swapping behavior attribute data of a certain battery swapping vehicle.

[0139] In this embodiment, for example, the constructed feature data can be split into a training set and a test set according to a ratio of 7:3 (this ratio is adjustable and other ratios can also be used). Among them, the splitting method is stratified sampling. The specific stratified sampling logic can be as follows: First, the number of days of the battery swapping time interval is divided into 8 categories according to (0,1], (1,2], (2,3], (3,6], (6,12], (12,20], (20,25], (25,30]. The division standard is judged through the data statistical distribution. The purpose is to make the data volume between various categories relatively balanced. There is no strict standard for the specific division. When performing stratified sampling on the 8 categories of data, other sampling methods can also be used, such as simple sampling and systematic sampling;

[0140] After sampling, null value data processing is performed. Here, the null value imputation method of filling with 0 is adopted. Then, a standardization method is used to perform feature scaling on numerical variables; one-hot encoding is used to numericalize categorical variables; the training set is input into the network model, and a regression model and a decision tree model are trained respectively. Here, the target variable is the battery swapping time interval. The evaluation method used for the battery swapping time prediction model of battery swapping vehicles is RMSE (root mean square error, also known as standard error); the training set is divided into a smaller training set and a validation set. Here, 10-fold cross-validation is adopted (it randomly divides the training set into 10 different subsets, each subset is called a fold, and then the network model is trained and evaluated 10 times. Each time, 1 fold is selected for evaluation, and the other 9 folds are used for training); the test set is used for validation analysis. Specifically, the test set is input into the trained battery swapping time prediction model of battery swapping vehicles, and the weighted average is used to output the final predicted value predict_day_interval (the predicted interval days for the next battery swapping). Here, the result weight values of the battery swapping time prediction model of battery swapping vehicles are all 0.5. predict_day_interval = model predicted value; predict_day_interval_low = floor(model predicted value * 0.5); predict_day_interval_up = ceil(model predicted value * 1.5), and then compare with 30 and take the smaller value.

[0141] In an optional embodiment, the abnormal data analysis and processing of the historical battery swapping data set in step 100-1 includes:

[0142] Step 100-11: Obtain the historical battery swapping time interval, the difference in battery swapping battery SOC, the difference in battery swapping power, and the difference in battery swapping mileage between any two adjacent battery swapping times of the battery swapping vehicle;

[0143] Step 100-12: Identify abnormal data according to the comparison results of the historical battery swapping time interval, the difference in battery swapping battery SOC, the difference in battery swapping power, and the difference in battery swapping mileage with their respective preset thresholds.

[0144] In this embodiment, the battery swapping time interval = the time when the last battery swapping battery was replaced - the time of this battery swapping order): there are abnormal battery swapping time interval data. For example, the battery swapping time interval is 0 and the battery swapping time interval is greater than the preset battery swapping time interval. From the distribution, 15% of the battery swapping records have a battery swapping time interval greater than 26 hours, 5% of the battery swapping records have a battery swapping time interval less than 4 hours, and 80% of the battery swapping intervals are concentrated between 4 and 26 hours. From 85% to 90%, the span is 13 hours, and from 90% to 95%, the span is 32 hours.

[0145] SOC difference of battery swapping = SOC value of the battery installed during the last battery swapping - SOC value of the battery removed during the current battery swapping: There are negative values for the SOC difference of battery swapping. For example, the maximum negative value of the SOC difference of battery swapping is 87, and the maximum positive difference of the SOC difference of battery swapping is 100. There are relatively large abnormal records. 90% of the SOC differences of battery swapping are between 26 and 87. Except for some abnormalities, for example, the SOC differences of the starting battery swapping are all 0. The recorded battery swapping time intervals corresponding to too small SOC differences of battery swapping are relatively large. Generally, for too large SOC differences of battery swapping, the driving time is relatively long, but there are many records where the SOC difference of the battery at the end of the swap is less than 2.

[0146] Battery swapping power difference = power of the battery installed during the last battery swapping - power of the battery removed during the current battery swapping: 90% of the recorded battery swapping power differences are between 11 and 39 kWh. There are situations where the battery swapping power difference is abnormally negative and abnormally large. For example, the minimum negative value of the battery swapping power difference is -38 kWh, and the maximum positive difference of the battery swapping power difference is 10000052.

[0147] Battery swapping mileage difference = vehicle mileage during the current battery swapping - vehicle mileage during the last battery swapping: 90% of the battery swapping mileage differences are concentrated between 42 and 367 km. For example, the minimum battery swapping mileage of the battery swapping mileage difference is 0, and the maximum battery swapping mileage of the battery swapping mileage difference is 1224302.

[0148] In this embodiment, the preset threshold is set according to the actual situation and is not specifically limited here.

[0149] In this embodiment, the network model is trained through the historical battery swapping time intervals of battery swapping vehicles in the obtained training set to obtain a battery swapping time prediction model for battery swapping vehicles, improving the accuracy and prediction precision of model prediction.

[0150] Embodiment 2

[0151] A model training system provided in this embodiment, as Figure 2 shown, includes: a training set acquisition module 21, a training module 22;

[0152] The training set acquisition module 21 is used to acquire a training set, and the training set includes the historical battery swapping time intervals of battery swapping vehicles;

[0153] The training module 22 is used to train a network model based on the historical battery swapping time intervals of battery swapping vehicles to obtain a battery swapping time prediction model for battery swapping vehicles.

[0154] In an optional embodiment, as Figure 2 shown, this model training system further includes: a historical battery swapping data set acquisition module 23, a first processing module 24, a second processing module 25;

[0155] The historical battery swapping dataset acquisition module 23 is used to acquire the historical battery swapping dataset of battery swapping vehicles;

[0156] In this embodiment, the historical battery swapping dataset of battery swapping vehicles includes at least one of vehicle battery swapping order data, vehicle basic information, and station terminal basic information.

[0157] The vehicle battery swapping order data includes at least one of payment time, vehicle number, vehicle code, station code, city ID, order type, battery code, whether the order is valid, consumption type, mileage, order status, operation type, battery SOC when swapping on and off, time, and power.

[0158] The vehicle basic information includes at least one of operation type, vehicle model, account status, vehicle type, and the city to which it belongs.

[0159] The station terminal basic information includes at least one of station terminal code, station terminal operation type, the city to which it belongs, business hours, and supported vehicle models.

[0160] The first processing module 24 is used to perform distribution analysis and abnormal data analysis and processing on the historical battery swapping dataset to obtain the processed historical battery swapping dataset;

[0161] In this embodiment, the historical battery swapping dataset is screened based on whether a certain vehicle needs battery swapping. For example, the most recent battery swapping records are selected for analysis. However, abnormal data may also appear in the time intervals of the most recent battery swapping records. For example, among the most recent 10 battery swapping records, if the time interval of a certain battery swapping becomes dozens of days or hundreds of days, etc., then the data with a relatively large time interval is considered abnormal data. Usually, the time intervals of each battery swapping are relatively small (for example, swapping once every 10 days), and those with a large difference are considered abnormal data. The abnormal data is found and removed.

[0162] In this embodiment, combined with the data distribution situation, statistical analysis is performed on the abnormal data. The data distribution mainly uses the 20th percentile, and the abnormal data is mainly judged in combination with the business situation. For example, a battery swap should be carried out within 30 days, and the differences in battery SOC, power, and battery swapping mileage between two battery swaps should not be too low or too high.

[0163] The second processing module 25 is used to perform feature construction and feature engineering on the processed historical battery swapping dataset to obtain a training set.

[0164] In this embodiment, feature selection is performed on the processed historical battery swapping dataset to identify which feature data can be used to estimate the battery swapping demand of battery swapping vehicles on a certain future day. Specifically, different algorithms are used to construct different features. For example, time series analysis algorithms are used to construct time-related feature data. After the features are constructed, further data analysis needs to be performed on the constructed features to finally determine the useful feature data. It should be noted that this process is a cyclic process. After all the features are analyzed cyclically, the final useful feature data can be obtained. In addition, the algorithms used in this embodiment are related to the selected features. If different features are required, algorithms corresponding to the features can be used according to the requirements. A certain feature may be obtained through one or more algorithms.

[0165] In this embodiment, feature engineering is to convert data into features that can better represent potential problems (for example, converting the processed historical battery swapping dataset into features), thereby improving the machine learning ability. Further, through feature engineering, the processed historical battery swapping dataset is converted into features. These features can well describe the processed historical battery swapping dataset, and the performance of the model established using them on unknown data can reach the optimal or near-optimal performance.

[0166] It should be noted that the better the features, the stronger the flexibility, the simpler the constructed model, and the more excellent the performance of the model.

[0167] In addition, the time series analysis algorithm can be adjusted according to the actual situation.

[0168] In the specific implementation process, the following content of the historical battery swapping dataset of battery swapping vehicles is obtained: city ID, vehicle ID, consumption type, vehicle type;

[0169] Total number of battery swaps: Calculated by battery swapping vehicle dimension, multiple battery swaps in one day are counted as one;

[0170] Time since the first battery swap: Calculated by battery swapping vehicle dimension, the number of days from the earliest battery swap record to the current date;

[0171] Time since the most recent battery swap: Calculated by battery swapping vehicle dimension, the number of days from the most recent battery swap record to the current date;

[0172] Total number of battery swaps in the most recent year: Calculated by battery swapping vehicle dimension, the total number of battery swaps in the most recent 1 year, multiple battery swaps in one day are counted as one;

[0173] Median 50% quantile average - before correction: Sort all the battery swapping time intervals of the battery swapping vehicle, take the median 50% quantile, and then calculate the average quantile_avg (average quantile);

[0174] Lower quartile - before correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, and take the lower quartile;

[0175] Upper quartile - before correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, and take the upper quartile;

[0176] Average of the middle 50% quantiles - after correction: Calculate and sort all battery swap time intervals by the battery swap vehicle dimension, eliminate abnormal battery swap records, and then take the middle 50% quantiles of all records after eliminating the corrections, and calculate the average quantile_avg.

[0177] Among them, the judgment process of abnormal records: all battery replacement time intervals are compared with the "middle 50% quantile average - before correction". Records with a battery replacement time interval greater than "middle 50% quantile average - before correction" * 10 and greater than 40 days and all battery replacement records of the battery replacement vehicle before the date are abnormal records. The replacements here are: 50% can be replaced by 70%, 80%; the average can be replaced by the weighted average, mode, etc.; * 10 can be replaced by (Q3-Q1) * 1.5 times, greater than 40 days can be replaced by 50 days, etc.);

[0178] Lower quartile - before correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, sort by battery swap time interval, and take the lower quartile;

[0179] Upper quartile - before correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, sort by battery swap time interval, and take the upper quartile;

[0180] Minimum value of all records - after correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, remove abnormal battery swapping records, and take the minimum value;

[0181] Maximum value of all records - after correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, remove abnormal battery swapping records, and take the maximum value;

[0182] Average value of all records - after correction: Calculate and sort all battery swapping time intervals by battery swapping vehicle dimension, remove abnormal battery swapping records, and take the average value;

[0183] Average of the battery swap time intervals of the last three battery swap records - after correction: Calculate and sort all battery swap time intervals by battery swap vehicle dimension, remove abnormal battery swap records, and take the moving average of the battery swap time intervals of the last three battery swap records;

[0184] This embodiment can refer to the moving average algorithm in the time series. The simple moving average algorithm used here can be replaced by weighted moving average method, trend moving average method, exponential smoothing method, and differential exponential smoothing method.

[0185] The most recent battery swapping time - after correction: Sort all battery swapping time intervals by battery swapping vehicle dimension, and take the most recent battery swapping time;

[0186] The battery swapping time interval of the most recent battery swapping record - after correction: Sort all battery swapping time intervals by battery swapping vehicle dimension, eliminate abnormal battery swapping records, and take the battery swapping time interval corresponding to the most recent 1 battery swapping record;

[0187] The total number of battery swaps - after correction: Sort all battery swapping time intervals by battery swapping vehicle dimension, eliminate abnormal battery swapping records, and summarize the number of battery swaps;

[0188] Whether it is stable - after correction (is_steady): Sort all battery swapping time intervals by battery swapping vehicle dimension, eliminate abnormal battery swapping records, and then perform stability detection and judgment.

[0189] Stability detection logic: ADF test in unit root test in time series analysis. Replaceable methods here include: time series diagram, autocorrelation diagram and partial autocorrelation diagram, DF test, PP test, DF-GLS test, KPSS test, etc.

[0190] In an optional embodiment, as Figure 2 shown, the second processing module 25 includes: a partitioning unit 251, a verification unit 252, and a testing unit 253;

[0191] The partitioning unit 251 is used to partition the training set, verification set, and test set according to a preset ratio;

[0192] In this embodiment, the preset ratio is set according to the actual situation. For example, the constructed feature data is divided into a training set, a verification set, and a test set according to a ratio. For example, the feature data is divided into training set + verification set: test set according to 8:2 or 9:1, and then the training set and verification set are partitioned according to 2:8.

[0193] It should be noted that the training set, verification set, and test set are only randomly partitioned according to a ratio. When the quantity ratio is quite different, stratified sampling will be used.

[0194] The verification unit 252 is used to verify the battery swapping time prediction model of the battery swapping vehicle by using the verification set to obtain the verified battery swapping time prediction model of the battery swapping vehicle;

[0195] The testing unit 253 is used to test the prediction result of the battery swapping time prediction model of the battery swapping vehicle by using the test set.

[0196] In the specific implementation process, the network model is trained based on the training set. The network model includes, but is not limited to, regression models and decision tree models. Specifically, the regression model and the decision tree model are respectively trained based on the training set to obtain two trained models, and then these two models are combined to finally obtain the charging time prediction model for the battery swapping vehicle;

[0197] It should be noted that the trained models can be replaced, and multiple models can be trained simultaneously. Finally, the charging time prediction model for the battery swapping vehicle is obtained after combination. The above-mentioned trained charging time prediction model for the battery swapping vehicle can predict which time period the battery swapping vehicle will go for battery swapping in the future, that is, predict the charging time of the battery swapping vehicle.

[0198] In addition, during the training process of the network model, the training set and the validation set are mainly used. Specifically, the training set is used to train the network model to obtain the charging time prediction model for the battery swapping vehicle, and then the validation set is used to verify the charging time prediction model for the battery swapping vehicle to obtain a better charging time prediction model for the battery swapping vehicle (such as an electric vehicle). The test set is used to verify the effect of the charging time prediction model for the battery swapping vehicle. Finally, the charging demand interval of the battery swapping vehicle is predicted through the charging time prediction model for the battery swapping vehicle (that is, it is predicted that a certain battery swapping vehicle is expected to come for battery swapping in a certain period of time in the future. For example, it is expected to come for battery swapping in the next 1 - 5 days). The test set is the time period when it is already determined that the battery swapping vehicle will come back for battery swapping in the future. Therefore, the test set is used to verify whether the result of the battery swapping time of the battery swapping vehicle predicted by the charging time prediction model for the battery swapping vehicle is consistent with the time in the test set;

[0199] Then, from the dimension of the battery swapping vehicle, the estimated charging demand of a certain battery swapping station in a certain period of time in the future is predicted, and which station the battery swapping vehicle will go to for battery swapping is determined according to the recent charging behavior attribute data of the battery swapping vehicle.

[0200] In this embodiment, for example, the constructed feature data can be split into a training set and a test set according to 7:3 (this ratio is adjustable and other ratios can also be used). Among them, the splitting method is stratified sampling. The specific stratified sampling logic can be: first, the number of days of the charging time interval is divided into 8 categories according to (0,1], (1,2], (2,3], (3,6], (6,12], (12,20], (20,25], (25,30]. The division standard is judged through the data statistical distribution. The purpose is to make the data volumes of each category relatively balanced. There is no strict standard for the specific division. When performing stratified sampling on the 8 categories of data, other sampling methods can also be used, such as simple sampling and systematic sampling;

[0201] After sampling, null value data processing is performed. Here, the null value imputation method of filling with 0 is adopted. Then, the standardization method is used to perform feature scaling on numerical variables; the one-hot encoding is used to numericalize categorical variables; the training set is input into the network model, and the regression model and decision tree model are trained respectively. Here, the target variable is the battery swapping time interval. The evaluation method used for the battery swapping time prediction model of battery swapping vehicles is RMSE (root mean square error, also known as standard error); the training set is divided into a smaller training set and a validation set. Here, 10-fold cross-validation is adopted (it randomly divides the training set into 10 different subsets, each subset is called a fold, and then the network model is trained and evaluated 10 times. Each time, one fold is selected for evaluation, and the other 9 folds are used for training); the test set is used for verification analysis. Specifically, the test set is input into the trained battery swapping time prediction model of battery swapping vehicles, and the weighted average is used to output the final predicted value predict_day_interval (the predicted interval days for the next battery swapping). Here, the result weighting values of the battery swapping time prediction model of battery swapping vehicles are all 0.5. predict_day_interval = model predicted value; predict_day_interval_low = floor(model predicted value * 0.5); predict_day_interval_up = ceiling(model predicted value * 1.5), and then compare with 30, and take the smaller value.

[0202] In an optional embodiment, as Figure 2 shown, the first processing module 24 includes: a first acquisition unit 241 and an identification unit 242;

[0203] The first acquisition unit 241 is used to acquire the historical battery swapping time interval, the difference in battery swapping battery SOC, the difference in battery swapping power, and the difference in battery swapping mileage between any two adjacent battery swapping times of the battery swapping vehicle;

[0204] The identification unit 242 is used to identify abnormal data according to the comparison results of the historical battery swapping time interval, the difference in battery swapping battery SOC, the difference in battery swapping power, and the difference in battery swapping mileage with their respective preset thresholds.

[0205] In this embodiment, the battery swapping time interval = the time when the last battery swapping battery was replaced - the time of this battery swapping order): there are abnormal battery swapping time interval data. For example, the battery swapping time interval is 0 and the battery swapping time interval is greater than the preset battery swapping time interval. From the distribution, 15% of the battery swapping records have a battery swapping time interval greater than 26 hours, 5% of the battery swapping records have a battery swapping time interval less than 4 hours, and 80% of the battery swapping intervals are concentrated between 4 and 26 hours. From 85% to 90%, the span is 13 hours, and from 90% to 95%, the span is 32 hours.

[0206] SOC difference of battery swapping = SOC value of the battery installed during the previous battery swapping - SOC value of the battery removed during the current battery swapping: There are negative values for the SOC difference of battery swapping. For example, the maximum negative value of the SOC difference of battery swapping is 87, and the maximum positive difference of the SOC difference of battery swapping is 100. There are relatively large abnormal records. 90% of the SOC differences of battery swapping are between 26 and 87. Except for some abnormalities, for example, the SOC differences of the starting battery swapping are all 0, the recorded battery swapping time intervals corresponding to too small SOC differences of battery swapping are relatively large, and generally, the driving time is relatively long for too large SOC differences of battery swapping, but there are many records where the SOC difference of the battery at the end of the swap is less than 2.

[0207] Battery swapping power difference = power of the battery installed during the previous battery swapping - power of the battery removed during the current battery swapping: 90% of the recorded battery swapping power differences are between 11 and 39 kWh. There are situations where the battery swapping power difference is abnormally negative and abnormally large. For example, the minimum negative value of the battery swapping power difference is -38 kWh, and the maximum positive difference of the battery swapping power difference is 10000052.

[0208] Battery swapping mileage difference = vehicle mileage during the current battery swapping - vehicle mileage during the previous battery swapping: 90% of the battery swapping mileage differences are concentrated between 42 and 367 km. For example, the minimum battery swapping mileage of the battery swapping mileage difference is 0, and the maximum battery swapping mileage of the battery swapping mileage difference is 1224302.

[0209] In this embodiment, the preset threshold is set according to the actual situation and is not specifically limited here.

[0210] In this embodiment, the network model is trained through the historical battery swapping time intervals of battery swapping vehicles in the obtained training set to obtain a battery swapping time prediction model for battery swapping vehicles, improving the accuracy and prediction precision of model prediction.

[0211] Embodiment 3

[0212] Figure 3 It 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 configured to run on the processor. When the processor executes the program, it implements the model training method of Embodiment 1. Figure 3 The displayed electronic device 30 is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present invention.

[0213] As Figure 3 shown, the electronic device 30 can be presented in the form of a general computing device. For example, it can 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).

[0214] The bus 33 includes a data bus, an address bus, and a control bus.

[0215] 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.

[0216] The memory 32 may also include a program / utilities 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.

[0217] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32, such as the model training method of Embodiment 1 of the present invention.

[0218] 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 model generating device 30 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 combination with the model generating device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (redundant array of independent disks) systems, tape drives, and data backup storage systems, etc.

[0219] 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.

[0220] Embodiment 4

[0221] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the model training method provided in Embodiment 1 is implemented.

[0222] Among them, the readable storage medium may more specifically include, but is 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.

[0223] In a possible implementation, 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 model training method described in Embodiment 1.

[0224] 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.

[0225] Embodiment 5

[0226] A method for predicting the battery swapping time of a battery swapping vehicle provided in this embodiment is as Figure 4 shown, and includes:

[0227] Step 201, obtain the historical battery swapping time interval of the battery swapping vehicle as the input of the battery swapping time prediction model of the battery swapping vehicle;

[0228] Step 202, use the battery swapping time prediction model of the battery swapping vehicle to predict the battery swapping time interval of the battery swapping vehicle to obtain the predicted battery swapping time interval of the battery swapping vehicle;

[0229] Step 203, obtain the upper limit value and the lower limit value of the predicted battery swapping time interval according to the predicted battery swapping time interval;

[0230] Step 204, obtain the predicted battery swapping probability value of the battery swapping vehicle according to the predicted battery swapping time interval, the upper limit value and the lower limit value;

[0231] Among them, the battery swapping time prediction model of the battery swapping vehicle is trained by using the model training method in Embodiment 1.

[0232] This embodiment mainly predicts the future battery swapping time interval based on historical battery swapping data, and adopts different prediction methods for the historical battery swapping data of different battery swapping vehicles. Combining the predicted number of days for the next battery swapping and the upper limit value and the lower limit value, the daily battery swapping probability value of the battery swapping vehicle for the next 30 days starting from the current battery swapping time is output.

[0233] Embodiment 6

[0234] A battery swapping vehicle battery swapping time prediction system provided in this embodiment is as Figure 5As shown in the figure, it includes: a battery swapping time interval acquisition module 11, a first prediction module 12, a first acquisition module 13, and a second acquisition module 14;

[0235] The battery swapping time interval acquisition module 11 is configured to acquire the historical battery swapping time intervals of battery swapping vehicles as the input of the battery swapping time prediction model of the battery swapping vehicles;

[0236] The first prediction module 12 is configured to predict the battery swapping time intervals of battery swapping vehicles by using the battery swapping time prediction model of the battery swapping vehicles, so as to obtain the predicted battery swapping time intervals of the battery swapping vehicles;

[0237] The first acquisition module 13 is configured to acquire the upper limit value and the lower limit value of the predicted battery swapping time interval according to the predicted battery swapping time interval;

[0238] The second acquisition module 14 is configured to acquire the predicted battery swapping probability value of the battery swapping vehicle according to the predicted battery swapping time interval, the upper limit value, and the lower limit value;

[0239] Among them, the battery swapping time prediction model of the battery swapping vehicle is trained by using the model training system of Embodiment 2.

[0240] This embodiment mainly predicts the future battery swapping time intervals based on historical battery swapping data, and different prediction methods are adopted for the historical battery swapping data of different battery swapping vehicles. Combining the predicted number of days until the next battery swapping and the upper limit value and the lower limit value, the daily battery swapping probability values of the battery swapping vehicle for the next 30 days starting from the current battery swapping time are output.

[0241] Embodiment 7

[0242] A schematic structural diagram of an electronic device provided in Embodiment 7 of the present invention. The schematic structural diagram of the electronic device in this embodiment is the same as that of Figure 3 . The electronic device includes a memory, a processor, and a computer program stored on the memory and configured to run on the processor. When the processor executes the program, it implements the battery swapping time prediction method for battery swapping vehicles in Embodiment 5. Figure 3 The shown electronic device 30 is only an example, and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0243] As Figure 3 shown, the electronic device 30 may be presented in the form of a general 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).

[0244] The bus 33 includes a data bus, an address bus, and a control bus.

[0245] 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.

[0246] The memory 32 may also include a program / utilities 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 an implementation of a network environment.

[0247] 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 time of a battery swapping vehicle in Embodiment 5 of the present invention.

[0248] 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 performed through an input / output (I / O) interface 35. Moreover, the model generation device 30 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 model generation device 30 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 model generation device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (redundant array of independent disks) systems, tape drives, and data backup storage systems, etc.

[0249] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this 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-described units / modules may be embodied in one unit / modules. Conversely, the features and functions of one unit / modules described above may be further divided and embodied by multiple units / modules.

[0250] Embodiment 8

[0251] 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 swapping time of a battery swapping vehicle provided in Embodiment 5.

[0252] Among them, the readable storage medium can more specifically include, but is 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.

[0253] In a possible implementation, 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 swapping time of the battery swapping vehicle described in Embodiment 5.

[0254] 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 completely executed on the user device, partially executed on the user device, executed as an independent software package, partially executed on the user device and partially executed on a remote device, or completely executed on a remote device.

[0255] 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, and 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.

[0256] Embodiment 9

[0257] A method for predicting the battery swapping demand of a battery swapping station provided in this embodiment, as Figure 6 shown, includes:

[0258] Step 301, predicting the predicted battery swapping probability values of multiple battery swapping vehicles by using the method for predicting the battery swapping time of the battery swapping vehicle described in Embodiment 5;

[0259] Step 302, obtaining the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations in a historical time period;

[0260] Step 303, statistically calculating the predicted battery swapping demand of each battery swapping station within a preset time interval according to the predicted battery swapping probability values of multiple battery swapping vehicles and the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations. The predicted battery swapping demand includes the number of battery swapping vehicles whose predicted battery swapping time intervals are within the preset time interval.

[0261] In an optional embodiment, Step 302 includes:

[0262] Step 3021, obtaining the total historical battery swapping times of multiple battery swapping vehicles and the total historical battery swapping times of different battery swapping stations in a historical time period;

[0263] Step 3022: Obtain the ratio of the number of battery swapping times of each battery swapping vehicle to different battery swapping stations according to the total historical battery swapping times of multiple battery swapping vehicles and the total historical battery swapping times of different battery swapping stations;

[0264] Step 3023: Obtain the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations in the historical time period according to the ratio of the number of battery swapping times and the weight coefficient of each battery swapping station.

[0265] In this embodiment, for example, taking the battery swapping data of 2 months for statistics, respectively count the total historical battery swapping times of multiple battery swapping vehicles and the total historical battery swapping times of different battery swapping stations; obtain the ratio of the number of battery swapping times of each battery swapping vehicle to different battery swapping stations according to the total historical battery swapping times of multiple battery swapping vehicles and the total historical battery swapping times of different battery swapping stations; the probability values of the next battery swapping of the battery swapping vehicle to each battery swapping station are allocated according to the ratio of the number of battery swapping times, and in combination with the time sequence, multiply the weight coefficients 0.5, 0.3, and 0.2 for the recent 3 times of battery swapping stations A, B, and C respectively to obtain the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations.

[0266] In this embodiment, the battery swapping space planning model of the battery swapping vehicle with time weight can be used to predict which battery swapping station the battery swapping vehicle will go to for battery swapping and the probability value of going to which station for battery swapping. Specifically, input the historical battery swapping records of the battery swapping vehicle (i.e., which battery swapping stations it has gone to in the past) into the battery swapping space planning model of the battery swapping vehicle with time weight, and the target battery swapping station that the battery swapping vehicle will go to can be obtained.

[0267] It should be noted that the battery swapping space planning model of the battery swapping vehicle with time weight is trained based on the historical records of the battery swapping stations that the battery swapping vehicle has been to before (i.e., which battery swapping stations the battery swapping vehicle has gone to for battery swapping in the historical time period). In fact, according to which battery swapping stations the battery swapping vehicle has gone to for battery swapping in the historical time period, it is also possible to obtain which battery swapping station the battery swapping vehicle is more likely to go to for battery swapping. Specifically, train the battery swapping space planning model of the battery swapping vehicle with time weight according to the number of times the battery swapping vehicle has gone to each battery swapping station in the historical battery swapping records. This battery swapping space planning model of the battery swapping vehicle with time weight is relatively simple. This battery swapping space planning model of the battery swapping vehicle with time weight simply counts a ratio based on the number of times the battery swapping vehicle has gone to each battery swapping station, and then adds a time weight to this battery swapping space planning model of the battery swapping vehicle with time weight in combination with the future time period.

[0268] In the specific implementation process, the battery swapping record data of the target battery swapping vehicle is respectively input into the battery swapping time prediction model of the battery swapping vehicle and the battery swapping space planning model of the battery swapping vehicle with time weights, so as to respectively obtain the future battery swapping time period of the target battery swapping vehicle and which battery swapping station the target battery swapping vehicle will go to during this time period (or obtain the probability value that the target battery swapping vehicle will go to the target battery swapping station during this time period). Then, through the weight coefficient of the target battery swapping station obtained by the battery swapping space planning model of the battery swapping vehicle with time weights, it can be predicted which time period in the future the target battery swapping vehicle will go to which battery swapping station for battery swapping. Finally, according to which time period in the future the target battery swapping vehicle will go to which battery swapping station for battery swapping, the battery swapping demand of the target battery swapping station at a certain time period can be obtained (that is, it can be obtained how many battery swapping vehicles will come to the target battery swapping station for battery swapping at which time period). It is possible to perform queuing reminders based on the battery swapping vehicle data. From the prediction of the battery swapping vehicle to the battery swapping station, the battery swapping demand interval of the battery swapping vehicle and the demand interval of the battery swapping station can be obtained. Thus, according to the actual battery level of the battery swapping vehicle, it can be suggested whether the battery swapping vehicle needs to delay battery swapping to avoid high electricity price time periods, thereby reducing the charging cost.

[0269] Embodiment 10

[0270] A battery swapping demand prediction system for a battery swapping station provided in this embodiment, as Figure 7 shown, includes: a second prediction module 41, a third acquisition module 42, and a statistics module 43;

[0271] The second prediction module 41 is used to predict and obtain the predicted battery swapping probability values of multiple battery swapping vehicles by using the battery swapping time prediction system as described in Embodiment 6;

[0272] The third acquisition module 42 is used to acquire the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations in a historical time period;

[0273] The statistics module 43 is used to statistically calculate the predicted battery swapping demand of each battery swapping station at a preset time interval according to the predicted battery swapping probability values of multiple battery swapping vehicles and the battery swapping probability values of multiple battery swapping vehicles to different battery swapping stations. The predicted battery swapping demand includes the number of battery swapping vehicles whose predicted battery swapping time interval is within the preset time interval.

[0274] In an optional embodiment, as Figure 7 shown, the third acquisition module 42 includes: a second acquisition unit 421, a third acquisition unit 422, and a fourth acquisition unit 423;

[0275] The second acquisition unit 421 is used to acquire the total historical battery swapping times of multiple battery swapping vehicles in a historical time period and the total historical battery swapping times of different battery swapping stations;

[0276] A third obtaining unit 422, configured to obtain, according to the historical total number of battery replacements of multiple battery-swapping vehicles and the historical total number of battery replacements of different battery-swapping stations, the ratio of the number of battery replacements of each battery-swapping vehicle to different battery-swapping stations;

[0277] A fourth obtaining unit 423, configured to obtain, according to the ratio of the number of battery replacements and the weight coefficient of each battery-swapping station, the battery replacement probability values of multiple battery-swapping vehicles to different battery-swapping stations in a historical time period.

[0278] In this embodiment, for example, taking 2 months of battery replacement data for statistics, respectively, the historical total number of battery replacements of multiple battery-swapping vehicles and the historical total number of battery replacements of different battery-swapping stations are counted; according to the historical total number of battery replacements of multiple battery-swapping vehicles and the historical total number of battery replacements of different battery-swapping stations, the ratio of the number of battery replacements of each battery-swapping vehicle to different battery-swapping stations is obtained; the probability values of the next battery replacement of the battery-swapping vehicle going to each battery-swapping station are allocated according to the ratio of the number of battery replacements, and in combination with the time sequence, the weight coefficients 0.5, 0.3, and 0.2 are respectively multiplied by the battery-swapping stations A, B, and C in the last 3 times to obtain the battery replacement probability values of multiple battery-swapping vehicles to different battery-swapping stations.

[0279] In this embodiment, the battery replacement space planning model of the battery-swapping vehicle with time weight can be used to predict which battery-swapping station the battery-swapping vehicle will go to for battery replacement and the probability value of going to which station for battery replacement. Specifically, inputting the historical battery replacement records of the battery-swapping vehicle (that is, which battery-swapping stations the vehicle has gone to for battery replacement in history) into the battery replacement space planning model of the battery-swapping vehicle with time weight, the target battery-swapping station that the battery-swapping vehicle will go to can be obtained.

[0280] It should be noted that the battery replacement space planning model of the battery-swapping vehicle with time weight is trained based on the historical records of the battery-swapping stations that the battery-swapping vehicle has been to before (that is, which battery-swapping stations the battery-swapping vehicle has gone to for battery replacement in the historical time period). In fact, according to which battery-swapping stations the battery-swapping vehicle has gone to for battery replacement in the historical time period, it is also possible to obtain which battery-swapping station the battery-swapping vehicle is more likely to go to for battery replacement. Specifically, the battery replacement space planning model of the battery-swapping vehicle with time weight is trained according to the number of times the battery-swapping vehicle has gone to each battery-swapping station in the historical battery replacement records. The battery replacement space planning model of the battery-swapping vehicle with time weight is relatively simple. This battery replacement space planning model of the battery-swapping vehicle with time weight simply counts a ratio according to the number of times the battery-swapping vehicle has gone to each battery-swapping station, and then adds a time weight to the battery replacement space planning model of the battery-swapping vehicle with time weight in combination with the future time period.

[0281] In the specific implementation process, the battery swapping record data of the target battery swapping vehicle is respectively input into the battery swapping time prediction model of the battery swapping vehicle and the battery swapping space planning model of the battery swapping vehicle with time weights, so as to respectively obtain the future battery swapping time period of the target battery swapping vehicle and which battery swapping station the target battery swapping vehicle will go to during this time period (or obtain the probability value that the target battery swapping vehicle will go to the target battery swapping station during this time period). Then, through the weight coefficient of the target battery swapping station obtained by the battery swapping space planning model of the battery swapping vehicle with time weights, it can be further predicted which time period in the future the target battery swapping vehicle will go to which battery swapping station for battery swapping. Finally, according to which time period in the future the target battery swapping vehicle will go to which battery swapping station for battery swapping, the battery swapping demand of the target battery swapping station in a certain time period is obtained (that is, it is obtained how many battery swapping vehicles will come to the target battery swapping station for battery swapping during which time period). Queue reminders can be made based on the battery swapping vehicle data. From the prediction of the battery swapping vehicle to the battery swapping station, the battery swapping demand interval of the battery swapping vehicle and the demand interval of the battery swapping station can be obtained. Thus, it is recommended whether the battery swapping vehicle needs to delay battery swapping according to the actual battery level of the battery swapping vehicle to avoid high electricity price time periods, thereby reducing the charging cost.

[0282] Embodiment 11

[0283] The structural schematic diagram of an electronic device provided in Embodiment 11 of the present invention. The structural schematic diagram of the electronic device in this embodiment is the same as that of Figure 3 The structure is the same. The electronic device includes a memory, a processor, and a computer program stored on the memory and used to run on the processor. When the processor executes the program, it implements the battery swapping demand prediction method of Embodiment 9. Figure 3 The electronic device 30 shown is only an example and should not bring any limitations to the functions and usage scope of the embodiments of the present invention.

[0284] As Figure 3 shown, the electronic device 30 can be presented in the form of a general computing device. For example, it can 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).

[0285] The bus 33 includes a data bus, an address bus, and a control bus.

[0286] The memory 32 may include volatile memory, such as a random access memory (RAM) 321 and / or a cache memory 322, and may further include a read-only memory (ROM) 323.

[0287] The memory 32 may also include a program / utilities 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.

[0288] 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 power replacement demand of a power replacement station in Embodiment 9 of the present invention.

[0289] The electronic device 30 can also communicate with one or more external devices 34 (such as a keyboard, a pointing device, etc.). Such communication can be carried out through the input / output (I / O) interface 35. Moreover, the device 30 for generating a model can 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 the 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 can be used in combination with the device 30 for generating a model, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (redundant array of independent disks) systems, tape drives, and data backup storage systems, etc.

[0290] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this 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 can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0291] Embodiment 12

[0292] 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 power replacement demand of a power replacement station provided in Embodiment 9.

[0293] Among them, the more specific forms that the readable storage medium can adopt can include, but are not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0294] 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 power replacement demand of the power replacement station described in Embodiment 9.

[0295] 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.

[0296] 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 model training method, characterized in that, Including: Obtain a training set, where the training set includes the historical battery swapping time intervals of battery swapping vehicles; Train a network model based on the historical battery swapping time intervals of the battery swapping vehicles to obtain a battery swapping time prediction model for the battery swapping vehicles.

2. The model training method according to claim 1, wherein Before the step of obtaining the training set, the model training method further includes: Obtain the historical battery swapping data set of the battery swapping vehicles; Conduct distribution analysis and abnormal data analysis and processing on the historical battery swapping data set to obtain a processed historical battery swapping data set; Conduct feature construction and feature engineering on the processed historical battery swapping data set to obtain the training set.

3. The model training method according to claim 2, wherein The step of conducting feature construction and feature engineering on the processed historical battery swapping data set to obtain the training set includes: Divide the training set, validation set, and test set according to a preset ratio; Use the validation set to validate the battery swapping time prediction model of the battery swapping vehicles to obtain a validated battery swapping time prediction model for the battery swapping vehicles; Use the test set to test the prediction results of the battery swapping time prediction model of the battery swapping vehicles.

4. The model training method according to claim 2, wherein The step of conducting abnormal data analysis and processing on the historical battery swapping data set includes: Obtain the historical battery swapping time intervals, differences in battery SOC during battery swapping, differences in battery swapping power, and differences in battery swapping mileage between any two adjacent battery swapping times of the battery swapping vehicles; Identify abnormal data based on the comparison results of the historical battery swapping time intervals, differences in battery SOC during battery swapping, differences in battery swapping power, and differences in battery swapping mileage with their respective preset thresholds.

5. A method for predicting the battery swapping time of a battery swapping vehicle, characterized in that, Including: Obtain the historical battery swapping time intervals of the battery swapping vehicles as the input of the battery swapping time prediction model of the battery swapping vehicles; Use the battery swapping time prediction model of the battery swapping vehicles to predict the battery swapping time intervals of the battery swapping vehicles to obtain predicted battery swapping time intervals for the battery swapping vehicles; Obtain the upper limit value and lower limit value of the predicted battery swapping time intervals according to the predicted battery swapping time intervals; Obtain the predicted battery swapping probability values of the battery swapping vehicles according to the predicted battery swapping time intervals, the upper limit value, and the lower limit value; Wherein, the battery swapping time prediction model of the battery swapping vehicles is trained by using the model training method described in any one of claims 1-4.

6. A method for predicting the battery replacement demand of a battery swapping station, characterized in that, Including: Use the battery swapping time prediction method for battery swapping vehicles described in claim 5 to predict the predicted battery swapping probability values of multiple battery swapping vehicles; Obtain the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations during a historical time period; According to the predicted battery swapping probability values of the multiple battery swapping vehicles and the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations, statistically calculate the predicted battery swapping demands of each battery swapping station within a preset time interval, where the predicted battery swapping demands include the number of battery swapping vehicles whose predicted battery swapping time intervals are within the preset time interval.

7. The method for predicting the battery replacement demand of a battery replacement station according to claim 6, wherein, The step of obtaining the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations during a historical time period includes: Obtain the total historical battery swapping times of the multiple battery swapping vehicles and the total historical battery swapping times of the different battery swapping stations during the historical time period; Obtain the ratio of the battery swapping times of each battery swapping vehicle to different battery swapping stations according to the total historical battery swapping times of the multiple battery swapping vehicles and the total historical battery swapping times of the different battery swapping stations; Obtain the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations in the historical time period according to the ratio of the number of battery swapping times and the weight coefficient of each battery swapping station.

8. A model training system, characterized in that, It includes: A training set acquisition module, configured to acquire a training set, where the training set includes the historical battery swapping time intervals of the battery swapping vehicles; A training module, configured to train a network model based on the historical battery swapping time intervals of the battery swapping vehicles to obtain a battery swapping time prediction model for the battery swapping vehicles.

9. A swapping time prediction system for a swapping power vehicle, characterized in that, It includes: A battery swapping time interval acquisition module, configured to acquire the historical battery swapping time intervals of the battery swapping vehicles as the input of the battery swapping time prediction model for the battery swapping vehicles; A first prediction module, configured to predict the battery swapping time intervals of the battery swapping vehicles by using the battery swapping time prediction model of the battery swapping vehicles to obtain the predicted battery swapping time intervals of the battery swapping vehicles; A first acquisition module, configured to obtain the upper limit value and the lower limit value of the predicted battery swapping time intervals according to the predicted battery swapping time intervals; A second acquisition module, configured to obtain the predicted battery swapping probability values of the battery swapping vehicles according to the predicted battery swapping time intervals, the upper limit value, and the lower limit value; Wherein, the battery swapping time prediction model of the battery swapping vehicles is trained by using the model training system described in Claim 8.

10. A power replacement demand prediction system for a power replacement station, characterized in that, It includes: A second prediction module, configured to predict the predicted battery swapping probability values of multiple battery swapping vehicles by using the battery swapping vehicle battery swapping time prediction system described in Claim 9; A third acquisition module, configured to acquire the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations in the historical time period; A statistics module, configured to statistically calculate the predicted battery swapping demands of each battery swapping station within a preset time interval according to the predicted battery swapping probability values of the multiple battery swapping vehicles and the battery swapping probability values of the multiple battery swapping vehicles to different battery swapping stations, where the predicted battery swapping demands include the number of battery swapping vehicles whose predicted battery swapping time intervals are within the preset time interval.

11. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and configured to run on the processor, wherein When the processor executes the computer program, it implements at least one of the model training methods described in any one of Claims 1-4, the battery swapping vehicle battery swapping time prediction method described in Claim 5, and the battery swapping station battery swapping demand prediction methods described in any one of Claims 6-7.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements at least one of the model training methods described in any one of Claims 1-4, the battery swapping vehicle battery swapping time prediction method described in Claim 5, and the battery swapping station battery swapping demand prediction methods described in any one of Claims 6-7.

Images