An electric vehicle cluster charging and discharging load dynamic prediction method and device

Abstract

This invention provides a method and device for dynamic prediction of charging and discharging loads of electric vehicle clusters, relating to the field of data processing technology. The method includes: acquiring an electric vehicle cluster; collecting initial State of Charge (SOC) information and continuously collecting driving data sets; constructing an integrated charging and discharging predictor; predicting multiple discharge load information, calculating multiple predicted SOC information, predicting multiple charging time information, and analyzing multiple charging load information; predicting and optimizing charging power to obtain multiple optimal predicted charging power information; controlling the charging power when multiple electric vehicles are charging at charging piles; and combining the multiple discharge load information and multiple charging load information as multiple charging and discharging load prediction results. This invention solves the technical problem in the prior art where electric vehicle data is mostly scattered, and the distribution of charging power is unreasonable when multiple electric vehicles are charging simultaneously, leading to low utilization of charging facilities.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and device for dynamic prediction of charging and discharging load of electric vehicle clusters. Background Technology

[0002] Optimizing the charging power allocation for electric vehicles can improve the utilization rate of charging stations, avoid excessive grid load during peak charging periods, and ensure the safety and efficiency of the charging process. In existing technologies, the operating and charging / discharging data of electric vehicles are mostly scattered, lacking effective integration and utilization, and failing to form a global management and optimization strategy. Furthermore, traditional charging / discharging load prediction methods are mostly simple linear models or models based on single features, lacking in-depth mining and utilization of multi-dimensional features and historical data, resulting in low prediction accuracy and an inability to effectively cope with complex driving and charging behaviors. During peak charging periods, the lack of effective optimization strategies often leads to unreasonable power allocation when multiple electric vehicles are charging simultaneously, resulting in low utilization of charging facilities, slow charging speeds, and even overheating issues, affecting battery life and safety. Summary of the Invention

[0003] This application provides a method and device for dynamic prediction of charging and discharging load of electric vehicle clusters, aiming to solve the technical problem that in the prior art, the charging and discharging data of electric vehicles are mostly scattered, lack effective integration and utilization, and cannot form a global management, resulting in unreasonable distribution of charging power when multiple electric vehicles are charging at the same time, leading to low utilization of charging facilities.

[0004] The first aspect of this application discloses a method for dynamic prediction of charging and discharging load in an electric vehicle cluster. The method includes: acquiring multiple similar electric vehicles connected via a clustering protocol to form an electric vehicle cluster; collecting initial SOC information of multiple electric vehicles within the cluster at a preset time node, and continuously collecting driving behavior datasets within a preset time range, wherein each driving behavior dataset includes multiple types of driving characteristic information; and constructing an integrated charging and discharging predictor based on driving record data and charging and discharging record data from the historical time of the multiple electric vehicles. The integrated charging and discharging predictor includes multiple discharging prediction paths and multiple charging prediction paths corresponding to multiple types of driving characteristic information, and each discharging prediction path and charging prediction path includes multiple discharge prediction paths. The system includes an electric vehicle prediction branch and multiple charging prediction branches. Based on multiple driving datasets, multiple discharge prediction paths within the integrated charge / discharge predictor are used to predict multiple discharge load information. Combined with multiple initial SOC information, multiple predicted SOC information is calculated. Using the multiple charging prediction paths, multiple charging time information is predicted. Combined with the multiple predicted SOC information, multiple charging load information is analyzed and obtained. Based on the multiple charging load information, the charging power of the multiple electric vehicles is predicted and optimized to obtain multiple optimal predicted charging power information. When the multiple electric vehicles are charging using charging piles, the charging power is controlled, and the multiple discharge load information and multiple charging load information are combined as multiple charge / discharge load prediction results.

[0005] The second aspect of this application discloses a device for dynamic prediction of charging and discharging load of an electric vehicle cluster. This device is used in the aforementioned method for dynamic prediction of charging and discharging load of an electric vehicle cluster. The device includes: a module for acquiring similar electric vehicles, used to acquire multiple similar electric vehicles connected via a cluster protocol to form an electric vehicle cluster; a driving data acquisition module, used to acquire initial SOC information of multiple electric vehicles within the electric vehicle cluster when a preset time node is reached, and continuously acquire driving data sets within a preset time range, wherein each driving behavior data set includes multiple types of driving feature information; and a predictor construction module, used to construct an integrated charging and discharging predictor based on driving record data and charging and discharging record data of multiple electric vehicles over historical time periods, the integrated charging and discharging predictor including multiple discharging prediction paths and multiple charging prediction paths corresponding to multiple types of driving feature information. The system includes a power prediction path, where each discharge prediction path and charging prediction path comprises multiple discharge prediction branches and multiple charging prediction branches; a charging load acquisition module, which uses multiple discharge prediction paths within the integrated charge / discharge predictor to predict multiple discharge load information based on multiple driving datasets, calculates multiple predicted SOC information by combining multiple initial SOC information, predicts multiple charging time information using the multiple charging prediction paths, and analyzes and obtains multiple charging load information by combining the multiple predicted SOC information; and a prediction result acquisition module, which optimizes the charging power of the multiple electric vehicles based on the multiple charging load information to obtain multiple optimal predicted charging power information, controls the charging power when the multiple electric vehicles are charging at charging piles, and combines the multiple discharge load information and multiple charging load information as multiple charge / discharge load prediction results.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By connecting similar electric vehicles through cluster protocol communication to form an electric vehicle cluster, multiple electric vehicles within the cluster can be managed and scheduled in a unified manner. This not only improves the comprehensiveness of data but also enables collaborative optimization, enhancing the utilization rate of charging facilities and the intelligence level of the charging process. Collecting initial State of Charge (SOC) information at preset time nodes and continuously collecting various driving characteristic information within a preset time range allows for comprehensive and real-time monitoring of the electric vehicle's operating status, providing an accurate data foundation for subsequent prediction and optimization, and improving prediction accuracy. Based on historical driving record data and charge / discharge record data, an integrated charge / discharge predictor is constructed, which can fully utilize the characteristics of historical data to improve the accuracy of the prediction model through multiple predictions. The synergistic effect of paths and branches further improves the accuracy of predictions. By employing multiple discharge and charge prediction paths within the integrated charge and discharge predictor, the prediction results of different paths and branches can be combined to obtain more accurate discharge load and charging time information. Combined with initial SOC information and predicted SOC information, the charging load can be accurately calculated, improving the overall prediction effect. Optimizing charging power prediction based on charging load information enables globally optimal allocation of charging power for multiple electric vehicles. This not only improves charging efficiency but also effectively controls charging overheating temperature, protecting battery life. In actual charging, dynamic control of charging power can adapt to changes in actual demand, ensuring efficient and safe charging.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0009] Figure 1 This is a schematic flowchart of a method for dynamic prediction of charging and discharging load of electric vehicle clusters, provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the structure of a dynamic prediction device for the charging and discharging load of an electric vehicle cluster, provided in an embodiment of this application.

[0011] Figure labeling: Module 10 for acquiring similar electric vehicles, Module 20 for acquiring driving data, Module 30 for constructing predictors, Module 40 for acquiring charging load, and Module 50 for acquiring prediction results. Detailed Implementation

[0012] This application provides a method for dynamic prediction of charging and discharging load of electric vehicle clusters. It solves the technical problem in the prior art that the charging and discharging data of electric vehicles are mostly scattered, lack effective integration and utilization, and cannot form a global management system. This results in unreasonable distribution of charging power when multiple electric vehicles are charging at the same time, leading to low utilization of charging facilities.

[0013] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0014] like Figure 1 As shown in the figure, this application provides a method for dynamic prediction of charging and discharging load of electric vehicle clusters, the method comprising:

[0015] Multiple electric vehicles of the same type that are connected via a clustering protocol are acquired to form an electric vehicle cluster.

[0016] First, a standardized clustering protocol is established to unify communication among similar electric vehicles. This method can be implemented by charging service providers. Vehicle owners submit an application through their vehicle's control system, such as an onboard computer or a mobile application, filling in vehicle information including model and region. The cluster management system verifies the vehicle information, confirming that the vehicle meets cluster requirements, including being from the same region and model. After authentication, the vehicle obtains a unique cluster ID. Once a vehicle is authenticated as a cluster member, network parameters are configured on the vehicle to ensure communication within the cluster network. Vehicles discover nodes using the clustering protocol, identifying and connecting to other cluster members that accept the protocol. These cluster members are similar electric vehicles, including those from the same region and model.

[0017] When a preset time node is reached, the initial SOC information of multiple electric vehicles in the electric vehicle cluster is collected, and driving data sets are continuously collected within a preset time range. Each driving behavior data set includes multiple types of driving feature information.

[0018] The preset time node can be the time when the electric vehicle starts. This time node can be automatically detected by the onboard system. When the vehicle starts, the onboard system records the current time and sets it as the preset time node. At the preset time node, the initial SOC information of each electric vehicle in the electric vehicle cluster is collected. The SOC information represents the current state of charge of the battery, expressed as a percentage. For example, an SOC of 80% means that the battery currently has 80% charge.

[0019] Within a preset time frame, such as one hour or half a day, driving data sets of electric vehicles are continuously collected. These datasets include various driving characteristic information such as driving speed records, driving time records, and driving location records. The collected driving data is uploaded to the cluster management system via the cluster network for subsequent discharge load prediction.

[0020] Furthermore, upon reaching a preset time node, the initial SOC information of multiple electric vehicles within the electric vehicle cluster is collected, and driving data sets within a preset time range are continuously collected, including:

[0021] When a preset time node is reached, the initial SOC information of multiple electric vehicles in the electric vehicle cluster is collected; within a preset time range after the preset time node, the driving speed record, driving time record and driving location record of the multiple electric vehicles are collected as multiple types of driving feature information to obtain multiple driving datasets.

[0022] The preset time node can be the time when the electric vehicle starts. This time node can be automatically detected by the onboard system. When the vehicle starts, the onboard system records the current time and sets it as the preset time node. At the preset time node, the initial SOC information of each electric vehicle in the electric vehicle cluster is collected. The SOC information represents the current state of charge of the battery, expressed as a percentage. For example, an SOC of 80% means that the battery currently has 80% charge.

[0023] The preset time range can be one hour, half a day, or a whole day, depending on the specific needs. Within the preset time range after the preset time node, multiple types of driving characteristic information are continuously collected, including recording the vehicle's speed information during the driving process, recording the total driving time and specific time periods of the vehicle, using GPS to record the vehicle's driving trajectory and location, and summarizing the data to obtain multiple driving datasets, providing accurate data support for subsequent prediction and optimization.

[0024] Based on driving record data and charging / discharging record data from multiple electric vehicles over a historical period, an integrated charging / discharging predictor is constructed. The integrated charging / discharging predictor includes multiple discharge prediction paths and multiple charging prediction paths corresponding to multiple types of driving characteristic information. Each discharge prediction path and charging prediction path includes multiple discharge prediction branches and multiple charging prediction branches.

[0025] Acquire driving and charging / discharging data from multiple electric vehicles over historical periods. Driving data includes driving speed, driving time, and driving location records, while charging / discharging data includes the time for each charge and discharge, SOC changes, and charging time.

[0026] Based on historical data, an ensemble learning approach is used to construct and train multiple discharge prediction paths and charging prediction paths. Each path corresponds to one type of data, and each path contains multiple prediction branches. These branches predict the discharge load and the available charging time based on driving conditions within a preset time range. For example, if driving characteristics indicate a busy period with only one hour of charging time, or if charging time is longer, the accuracy of each prediction path is evaluated using a test set, and weighted according to the accuracy. Integrating these multiple discharge and charging prediction paths forms an integrated charge / discharge predictor, providing accurate predictive support for the charging and discharging management of electric vehicle clusters.

[0027] Furthermore, based on driving and charging / discharging record data from multiple historical periods of electric vehicles, an integrated charging / discharging predictor is constructed, including:

[0028] Based on driving and charging / discharging record data from multiple electric vehicles over historical periods, sample driving speed record sets, sample driving time record sets, sample driving location record sets, sample discharge load information sets, and sample charging time information sets are collected. These sets are then combined with the sample discharge load information sets and sample charging time information sets as outputs to obtain discharge speed training datasets, charging speed training datasets, discharge time training datasets, charging time training datasets, discharge location training datasets, and charging location training datasets. Based on ensemble learning, multiple discharge prediction paths and multiple charging prediction paths are trained to obtain an integrated charge / discharge predictor.

[0029] Based on driving record data from multiple electric vehicles over a historical period, sample driving speed record set, sample driving time record set, and sample driving location record set are extracted. The sample driving speed record set records the driving speed data of electric vehicles at different points in time during the historical period, the sample driving time record set records the driving time data of electric vehicles at different points in time during the historical period, and the sample driving location record set records the driving location data of electric vehicles at different points in time during the historical period.

[0030] Based on charging and discharging records of multiple electric vehicles over a historical period, sample discharge load information set and sample charging time information set are extracted. The sample discharge load information set records the discharge load of electric vehicles at different time points during the historical period, i.e., the power consumption data. The sample charging time information set records the charging time data of electric vehicles at different time points during the historical period.

[0031] Training datasets are constructed using sample data sets. Specifically, a set of sample driving speed records is used as input, and a set of sample discharge load information is used as output to obtain a discharge speed training dataset; a set of sample driving speed records is used as input, and a set of sample charging time information is used as output to obtain a charging speed training dataset; a set of sample driving time records is used as input, and a set of sample discharge load information is used as output to obtain a discharge time training dataset; a set of sample driving time records is used as input, and a set of sample charging time information is used as output to obtain a charging time training dataset; a set of sample driving location records is used as input, and a set of sample discharge load information is used as output to obtain a discharge location training dataset; and a set of sample driving location records is used as input, and a set of sample charging time information is used as output to obtain a charging location training dataset.

[0032] An ensemble learning approach is used to train multiple discharge prediction paths and multiple charging prediction paths. Specifically, based on the training dataset, ensemble learning algorithms, such as ensemble tree models and ensemble neural networks, are used to train multiple discharge prediction paths and multiple charging prediction paths. Each discharge prediction path predicts the discharge load based on different input features, and each charging prediction path predicts the charging time based on different input features.

[0033] Furthermore, based on ensemble learning, multiple discharge prediction paths and multiple charge prediction paths are trained to obtain an integrated charge / discharge predictor, including:

[0034] A predetermined proportion of training data is randomly selected from the discharge speed training dataset to train the first speed discharge prediction branch within the speed discharge prediction path; a predetermined proportion of training data is randomly selected again from the discharge speed training dataset to train the second speed discharge prediction branch; training continues to obtain multiple speed discharge prediction branches, which are then integrated to obtain the first discharge prediction path; the charging speed training dataset, discharge time training dataset, charging time training dataset, discharge location training dataset, and charging location training dataset are then used to train multiple discharge prediction paths and multiple charging prediction paths to obtain an integrated charge and discharge predictor.

[0035] By using a random sampling method, a certain proportion of data is randomly selected from the discharge speed training dataset as training data to ensure the diversity of the training data. The training data uses the sample driving speed record set as input and the sample discharge load information set as output to train the model. For example, machine learning models such as random forests and neural networks are used to train and obtain the first speed discharge prediction branch so that it can accurately predict the discharge load of electric vehicles at different speeds.

[0036] Another portion of data is randomly extracted from the discharge speed training dataset as training data. Similarly, the sample driving speed record set is used as input and the sample discharge load information set is used as output to train the second speed discharge prediction branch.

[0037] Repeat the above steps to train more speed-based discharge prediction branches. Using ensemble learning techniques, such as voting and weighted averaging, combine multiple trained discharge prediction branches to form the first discharge prediction path. By integrating multiple prediction branches, the accuracy and stability of discharge load prediction can be improved to cope with different driving conditions and data changes.

[0038] The training process for the first discharge prediction path is repeated, using discharge time and discharge location training datasets respectively to train the second and third discharge prediction paths, which are then integrated to obtain multiple discharge prediction paths. Similarly, using charging speed, charging time, and charging location training datasets respectively, the first, second, and third charging prediction paths are trained, which are then integrated to obtain multiple charging prediction paths. Finally, the multiple discharge and charging prediction paths are integrated to obtain an integrated charge / discharge predictor.

[0039] Based on multiple driving datasets, multiple discharge prediction paths within the integrated charge and discharge predictor are used to predict multiple discharge load information. Combined with multiple initial SOC information, multiple predicted SOC information is calculated. Using the multiple charging prediction paths, multiple charging time information is predicted. Combined with the multiple predicted SOC information, multiple charging load information is analyzed.

[0040] The discharge load information of each electric vehicle is predicted using multiple discharge prediction paths in the integrated charge and discharge predictor. Specifically, the driving dataset of each electric vehicle is input into the discharge prediction path. Each discharge prediction path predicts the discharge load based on the driving data, resulting in multiple discharge load prediction results. The prediction results of multiple branches of each prediction path are averaged and weighted to obtain the final discharge load information.

[0041] By combining the initial SOC information of each electric vehicle, the predicted SOC information is calculated. The predicted SOC is the predicted battery level after the trip. Specifically, the initial SOC is the battery level of each electric vehicle at the start, and the discharge load information is the predicted battery level consumed during the trip. The predicted SOC is calculated by subtracting the discharge load information from the initial SOC.

[0042] Using multiple charging prediction paths in the integrated charge and discharge predictor, the charging time information for each electric vehicle is predicted. Specifically, the driving dataset and predicted SOC information of each electric vehicle are input into the charging prediction path. Each charging prediction path predicts the charging time based on the driving data and predicted SOC, resulting in multiple charging time information. The prediction results of multiple branches of each prediction path are averaged and weighted to obtain the final charging time information.

[0043] By combining multiple predicted SOC information and charging time information, the charging load information is obtained through analysis. Specifically, the predicted SOC is the predicted battery level after driving is completed, the charging time information is the predicted charging time required to fully charge the battery, and the charging load information is calculated by dividing the predicted SOC by the charging time, which represents the charging power requirement to fully charge the battery, that is, the predicted charging load per unit of time required to fully charge the battery.

[0044] Furthermore, based on multiple driving datasets, multiple discharge prediction paths within the integrated charge / discharge predictor are used to predict multiple discharge load information. Combined with multiple initial SOC information, multiple predicted SOC information is calculated. Using the multiple charging prediction paths, multiple charging time information is predicted. Combining the multiple predicted SOC information, multiple charging load information is analyzed, including:

[0045] The multiple driving datasets are input into multiple discharge prediction paths within the integrated charge / discharge predictor for prediction, and the average of the prediction results of multiple discharge prediction branches is calculated to obtain multiple sets of discharge load information. The multiple discharge prediction paths are tested to obtain multiple discharge prediction accuracies, and the multiple sets of discharge load information are weighted to obtain multiple sets of discharge load information. Based on the multiple discharge load information and the multiple initial SOC information, multiple predicted SOC information is calculated. The multiple driving datasets are input into multiple charging prediction paths within the integrated charge / discharge predictor for prediction, and the average of the prediction results of multiple charging prediction branches is calculated to obtain multiple sets of charging time information. The multiple charging prediction paths are tested to obtain multiple charging prediction accuracies, and the multiple sets of charging time information are weighted to obtain multiple sets of charging time information. Based on the multiple predicted SOC information, multiple charging SOC information is calculated, and combined with the multiple charging time information, multiple charging load information is analyzed and calculated.

[0046] Each driving dataset is input into multiple discharge prediction paths within the integrated charge and discharge predictor for prediction. Each discharge prediction path predicts future electric vehicle discharge load information based on historical data and model training results. For any discharge prediction path, multiple discharge load prediction results of multiple discharge prediction branches of that path are obtained. The average of these prediction results is taken to obtain the discharge load information of that discharge prediction path. Multiple discharge prediction paths are traversed in sequence to obtain multiple sets of discharge load information.

[0047] For each discharge prediction path, a test dataset (which can be a separately reserved dataset) is used to test and evaluate its prediction accuracy and generalization ability. The prediction accuracy is calculated for each discharge prediction path, and this is measured by comparing the predicted results with the actual observations. Based on the accuracy of each discharge prediction path, the predicted discharge load information is weighted to obtain the discharge load information for each path. Prediction paths with higher accuracy are assigned higher weights to improve the overall reliability and accuracy of the prediction results. Through weighted calculation, the prediction results of multiple discharge prediction paths can be comprehensively utilized to obtain more accurate electric vehicle discharge load information.

[0048] Predicted SOC is the prediction of the remaining battery power after the trip. Specifically, the initial SOC is the battery power of each electric vehicle at the start, and the discharge load information is the predicted battery power consumed during the trip. The predicted SOC is calculated by subtracting the discharge load information from the initial SOC. Multiple predicted SOC values ​​are obtained through this subtraction calculation.

[0049] Each driving dataset is input into multiple charging prediction paths within the integrated charge and discharge predictor for prediction. Each charging prediction path predicts the charging time information of electric vehicles based on historical data and model training results. For each charging prediction path, multiple charging time prediction results from multiple charging prediction branches are obtained. These prediction results are averaged to obtain the charging time information set for each charging prediction path.

[0050] For each charging prediction path, a test dataset is used to evaluate its prediction accuracy and generalization ability, and the prediction accuracy of each charging prediction path is calculated. Based on the accuracy, the predicted charging time information is weighted and calculated to obtain multiple charging time information.

[0051] Predicted SOC information reflects the predicted SOC of each electric vehicle at the end of charging, while charging time information represents the expected charging period for each electric vehicle. Based on each predicted SOC and the corresponding charging time, the change in SOC value at each time point during charging is calculated. This can be estimated using a simple integration method: multiplying the charging rate (i.e., charging power) at each time point by the time period and accumulating the results to obtain the change in charging SOC. Multiple charging SOC information points are obtained through calculation, reflecting the actual SOC change of each electric vehicle during charging. Based on the charging SOC and charging time information, the charging power demand and charging efficiency of electric vehicles during charging can be analyzed. The average charging power demand of each electric vehicle at different time periods can be calculated, thereby assessing energy consumption and charging efficiency during charging. Multiple charging load information points are obtained through calculation, providing data support and decision-making basis for optimizing charging strategies and improving charging efficiency.

[0052] Based on the multiple charging load information, the charging power of the multiple electric vehicles is predicted and optimized to obtain multiple optimal predicted charging power information. When the multiple electric vehicles are charged using charging piles, the charging power is controlled, and the multiple discharge load information and multiple charging load information are combined as multiple charging and discharging load prediction results.

[0053] A charging power optimization model is established with the goals of increasing charging capacity and reducing charging overheating temperature. Specifically, the possible range of electric vehicle charging power is defined, which must meet a safety threshold, i.e., it cannot exceed the maximum allowable power. The optimization objectives are to increase charging capacity and reduce charging overheating temperature. A cluster charging function is constructed to predict and optimize the actual charging power of multiple electric vehicles within the charging power space, obtaining multiple optimal predicted charging power information.

[0054] During actual charging, the optimized predicted charging power information is used to control the charging power of each electric vehicle, and the charging status and temperature of each electric vehicle are monitored in real time. The charging power is adjusted to ensure it remains within a safe range. Based on real-time data, the charging power is dynamically adjusted to ensure charging efficiency and safety. The combined information from multiple discharge loads and multiple charging loads serves as the multiple predicted charging and discharging loads.

[0055] By following the steps above, the charging power of electric vehicles has been optimized, charging efficiency has been improved, and the risk of overheating during charging has been reduced, providing a safe and effective charging management solution for electric vehicle clusters.

[0056] Furthermore, based on the multiple charging load information, the charging power of the multiple electric vehicles is predicted and optimized to obtain multiple optimal predicted charging power information, including;

[0057] Obtain the charging power space of the multiple electric vehicles; with the aim of increasing the charging capacity of the multiple electric vehicles and reducing the charging overheating temperature, construct a cluster charging function to predict and optimize the charging power of the multiple electric vehicles, as shown in the following equation:

[0058]

[0059] Where CCF is the cluster fitness, w1 and w2 are the weights, and M is the number of electric vehicles in the electric vehicle cluster. To optimize the predicted charging power information of the i-th electric vehicle, For the charging load information of the i-th electric vehicle, T y The standard temperature for electric vehicle batteries, To optimize the charging overheating temperature of the i-th electric vehicle when charging according to the predicted charging power information; within the electric vehicle charging power space, the charging power of the multiple electric vehicles is predicted and optimized according to the cluster charging function to obtain multiple optimal predicted charging power information.

[0060] The electric vehicle charging power space is the range of possible charging power for each electric vehicle during the charging process. This range can be affected by factors such as the capacity of charging stations, electric vehicle battery technology, and safety requirements.

[0061] Construct the cluster charging function as follows:

[0062]

[0063] and These represent predicted charging power information and charging load information, respectively. The first part of the function reflects the relationship between predicted charging power and actual charging demand; T y This is the battery's standard temperature, which affects the battery's thermal management and efficiency during charging. This function predicts the overheating temperature during charging based on the predicted charging power information, guiding temperature control during the charging process. Weights w1 and w2 are set according to actual needs and optimization objectives to adjust the balance between charging power optimization and thermal management. The sum of the two parts is calculated separately to obtain the final cluster fitness. This cluster charging function allows for the evaluation of charging strategies for electric vehicle clusters, maximizing charging capacity while effectively controlling temperature during charging to extend battery life and improve charging efficiency.

[0064] Within the electric vehicle charging power space, the charging power of each electric vehicle is predicted and optimized according to the cluster charging function. The optimization objectives include increasing the charging capacity of electric vehicles while reducing the overheating temperature during charging to ensure charging safety and battery life. The specific optimization process will be detailed in subsequent steps and will not be elaborated here.

[0065] Furthermore, within the electric vehicle charging power space, the charging power of the multiple electric vehicles is predicted and optimized according to the cluster charging function, including:

[0066] Within the electric vehicle charging power space, multiple first predicted charging power information for multiple electric vehicles is randomly generated. Based on the multiple first predicted charging power information, multiple first charging overheating temperatures are predicted. Combining the multiple charging load information, a first cluster fitness is calculated based on the cluster charging function. Specifically, a charging temperature predictor is trained by collecting a sample charging power set and a sample charging overheating temperature set to predict multiple first charging overheating temperatures. Based on the deviation between the multiple first charging overheating temperatures and the first average charging overheating temperature, multiple adjustment ranges are set. Within the electric vehicle charging power space, the multiple first predicted charging power information is adjusted to obtain multiple second predicted charging power information, and a second cluster fitness is calculated. The prediction optimization of the charging power of multiple electric vehicles continues within the charging power space until convergence. The multiple predicted charging power information with the highest cluster fitness is output, resulting in multiple optimal predicted charging power information.

[0067] Within a defined charging power space, a predicted charging power value is randomly generated for each electric vehicle, resulting in multiple first predicted charging power information values. For each electric vehicle, based on its predicted charging power information, a corresponding charging overheating temperature is predicted. This is accomplished using a pre-trained charging temperature predictor, which is trained with a sample charging power set as input data and a sample charging overheating temperature set as output data, capable of mapping charging power to the expected charging overheating temperature. Multiple first charging overheating temperatures are obtained through the predictor's prediction. These multiple first predicted charging power information values, multiple first charging overheating temperatures, and multiple charging load information are input into the cluster charging function to calculate the first cluster fitness.

[0068] Calculate the average of multiple first charging overheat temperatures as the first average charging overheat temperature. For each electric vehicle, calculate the deviation between its predicted charging overheat temperature and the first average charging overheat temperature to obtain multiple deviation ranges. Set multiple adjustment ranges based on the multiple deviation ranges, for example, by directly normalizing the deviation ranges and using them as adjustment ranges.

[0069] These adjustment increments are applied to the first predicted charging power information for each electric vehicle to generate multiple second predicted charging power information sets. The second cluster fitness is then calculated using a cluster charging function, similar to the calculation method described above. This allows for adjustments to the predicted charging power information based on actual conditions, and evaluation of the effectiveness of the adjusted charging strategy, thereby optimizing power distribution and temperature control during the charging process and improving charging efficiency.

[0070] Repeat the above steps to continuously adjust the predicted charging power information and obtain the corresponding cluster fitness. Compare the size of all obtained cluster fitness and finally obtain the multiple predicted charging power information with the largest cluster fitness as multiple optimal predicted charging power information.

[0071] In summary, the method for dynamic prediction of charging and discharging load of electric vehicle clusters provided in this application has the following technical effects:

[0072] By connecting similar electric vehicles through cluster protocol communication to form an electric vehicle cluster, multiple electric vehicles within the cluster can be managed and scheduled in a unified manner. This not only improves the comprehensiveness of data but also enables collaborative optimization, enhancing the utilization rate of charging facilities and the intelligence level of the charging process. Collecting initial State of Charge (SOC) information at preset time nodes and continuously collecting various driving characteristic information within a preset time range allows for comprehensive and real-time monitoring of the electric vehicle's operating status, providing an accurate data foundation for subsequent prediction and optimization, and improving prediction accuracy. Based on historical driving record data and charge / discharge record data, an integrated charge / discharge predictor is constructed, which can fully utilize the characteristics of historical data to improve the accuracy of the prediction model through multiple predictions. The synergistic effect of paths and branches further improves the accuracy of predictions. By employing multiple discharge and charge prediction paths within the integrated charge and discharge predictor, the prediction results of different paths and branches can be combined to obtain more accurate discharge load and charging time information. Combined with initial SOC information and predicted SOC information, the charging load can be accurately calculated, improving the overall prediction effect. Optimizing charging power prediction based on charging load information enables globally optimal allocation of charging power for multiple electric vehicles. This not only improves charging efficiency but also effectively controls charging overheating temperature, protecting battery life. In actual charging, dynamic control of charging power can adapt to changes in actual demand, ensuring efficient and safe charging.

[0073] Based on the same inventive concept as the electric vehicle cluster charging and discharging load dynamic prediction method in the foregoing embodiments, such as Figure 2 As shown in the figure, this application embodiment provides a device for dynamic prediction of charging and discharging load of electric vehicle clusters, the device comprising:

[0074] A module 10 for acquiring similar electric vehicles is used to acquire multiple similar electric vehicles that are connected via a clustering protocol to form an electric vehicle cluster.

[0075] The driving data acquisition module 20 is used to collect the initial SOC information of multiple electric vehicles in the electric vehicle cluster when a preset time node is reached, and to continuously collect driving data sets within a preset time range, wherein each driving behavior data set includes multiple types of driving feature information.

[0076] The predictor building module 30 is used to build an integrated charge-discharge predictor based on driving record data and charge-discharge record data of multiple electric vehicles over a historical period. The integrated charge-discharge predictor includes multiple discharge prediction paths and multiple charging prediction paths corresponding to multiple types of driving feature information. Each discharge prediction path and charging prediction path includes multiple discharge prediction branches and multiple charging prediction branches.

[0077] The charging load acquisition module 40 is used to predict multiple discharge load information based on multiple driving datasets and multiple discharge prediction paths in the integrated charge and discharge predictor, calculate multiple predicted SOC information by combining multiple initial SOC information, predict multiple charging time information by using the multiple charging prediction paths, and analyze multiple charging load information by combining the multiple predicted SOC information.

[0078] The prediction result acquisition module 50 is used to predict and optimize the charging power of the multiple electric vehicles based on the multiple charging load information, obtain multiple optimal predicted charging power information, control the charging power when the multiple electric vehicles are charged by charging piles, and combine the multiple discharge load information and the multiple charging load information as multiple charging and discharging load prediction results.

[0079] Furthermore, the device also includes a driving data set acquisition module to perform the following operational steps:

[0080] When a preset time node is reached, the initial SOC information of multiple electric vehicles in the electric vehicle cluster is collected; within a preset time range after the preset time node, the driving speed record, driving time record and driving location record of the multiple electric vehicles are collected as multiple types of driving feature information to obtain multiple driving datasets.

[0081] Furthermore, the device also includes a charge / discharge predictor acquisition module to perform the following operational steps:

[0082] Based on driving and charging / discharging record data from multiple electric vehicles over historical periods, sample driving speed record sets, sample driving time record sets, sample driving location record sets, sample discharge load information sets, and sample charging time information sets are collected. These sets are then combined with the sample discharge load information sets and sample charging time information sets as outputs to obtain discharge speed training datasets, charging speed training datasets, discharge time training datasets, charging time training datasets, discharge location training datasets, and charging location training datasets. Based on ensemble learning, multiple discharge prediction paths and multiple charging prediction paths are trained to obtain an integrated charge / discharge predictor.

[0083] Furthermore, the device also includes an integrated charge / discharge predictor acquisition module to perform the following operational steps:

[0084] A predetermined proportion of training data is randomly selected from the discharge speed training dataset to train the first speed discharge prediction branch within the speed discharge prediction path; a predetermined proportion of training data is randomly selected again from the discharge speed training dataset to train the second speed discharge prediction branch; training continues to obtain multiple speed discharge prediction branches, which are then integrated to obtain the first discharge prediction path; the charging speed training dataset, discharge time training dataset, charging time training dataset, discharge location training dataset, and charging location training dataset are then used to train multiple discharge prediction paths and multiple charging prediction paths to obtain an integrated charge and discharge predictor.

[0085] Furthermore, the device also includes a charging load information acquisition module to perform the following operation steps:

[0086] The multiple driving datasets are input into multiple discharge prediction paths within the integrated charge / discharge predictor for prediction, and the average of the prediction results of multiple discharge prediction branches is calculated to obtain multiple sets of discharge load information. The multiple discharge prediction paths are tested to obtain multiple discharge prediction accuracies, and the multiple sets of discharge load information are weighted to obtain multiple sets of discharge load information. Based on the multiple discharge load information and the multiple initial SOC information, multiple predicted SOC information is calculated. The multiple driving datasets are input into multiple charging prediction paths within the integrated charge / discharge predictor for prediction, and the average of the prediction results of multiple charging prediction branches is calculated to obtain multiple sets of charging time information. The multiple charging prediction paths are tested to obtain multiple charging prediction accuracies, and the multiple sets of charging time information are weighted to obtain multiple sets of charging time information. Based on the multiple predicted SOC information, multiple charging SOC information is calculated, and combined with the multiple charging time information, multiple charging load information is analyzed and calculated.

[0087] Furthermore, the device also includes an optimal predicted charging power information acquisition module to perform the following operation steps:

[0088] Obtain the charging power space of the multiple electric vehicles; with the aim of increasing the charging capacity of the multiple electric vehicles and reducing the charging overheating temperature, construct a cluster charging function to predict and optimize the charging power of the multiple electric vehicles, as shown in the following equation:

[0089]

[0090] Where CCF is the cluster fitness, w1 and w2 are the weights, and M is the number of electric vehicles in the electric vehicle cluster. To optimize the predicted charging power information of the i-th electric vehicle, For the charging load information of the i-th electric vehicle, T y The standard temperature for electric vehicle batteries, To optimize the charging overheating temperature of the i-th electric vehicle when charging according to the predicted charging power information; within the electric vehicle charging power space, the charging power of the multiple electric vehicles is predicted and optimized according to the cluster charging function to obtain multiple optimal predicted charging power information.

[0091] Furthermore, the device also includes an optimal predicted charging power information acquisition module to perform the following operation steps:

[0092] Within the electric vehicle charging power space, multiple first predicted charging power information for multiple electric vehicles is randomly generated. Based on the multiple first predicted charging power information, multiple first charging overheating temperatures are predicted. Combining the multiple charging load information, a first cluster fitness is calculated based on the cluster charging function. Specifically, a charging temperature predictor is trained by collecting a sample charging power set and a sample charging overheating temperature set to predict multiple first charging overheating temperatures. Based on the deviation between the multiple first charging overheating temperatures and the first average charging overheating temperature, multiple adjustment ranges are set. Within the electric vehicle charging power space, the multiple first predicted charging power information is adjusted to obtain multiple second predicted charging power information, and a second cluster fitness is calculated. The prediction optimization of the charging power of multiple electric vehicles continues within the charging power space until convergence. The multiple predicted charging power information with the highest cluster fitness is output, resulting in multiple optimal predicted charging power information.

[0093] Through the foregoing detailed description of a dynamic prediction method for the charging and discharging load of an electric vehicle cluster, those skilled in the art can clearly understand the dynamic prediction device for the charging and discharging load of an electric vehicle cluster in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to in the method section.

[0094] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for dynamic prediction of charging and discharging load of electric vehicle clusters, characterized in that, The method includes: Multiple electric vehicles of the same type that are connected via a clustering protocol are acquired to form an electric vehicle cluster. When a preset time node is reached, the initial SOC information of multiple electric vehicles in the electric vehicle cluster is collected, and driving data sets are continuously collected within a preset time range. Each driving behavior data set includes multiple types of driving feature information. Based on driving record data and charging / discharging record data of multiple electric vehicles over a historical period, an integrated charging / discharging predictor is constructed. The integrated charging / discharging predictor includes multiple discharge prediction paths and multiple charging prediction paths corresponding to multiple types of driving characteristic information. Each discharge prediction path and charging prediction path includes multiple discharge prediction branches and multiple charging prediction branches. Based on multiple driving datasets, multiple discharge prediction paths within the integrated charge and discharge predictor are used to predict multiple discharge load information. Combined with multiple initial SOC information, multiple predicted SOC information is calculated. Using the multiple charging prediction paths, multiple charging time information is predicted. Combined with the multiple predicted SOC information, multiple charging load information is analyzed. Based on the multiple charging load information, the charging power of the multiple electric vehicles is predicted and optimized to obtain multiple optimal predicted charging power information. When the multiple electric vehicles are charged using charging piles, the charging power is controlled, and the multiple discharge load information and multiple charging load information are combined as multiple charging and discharging load prediction results. 2.The method of claim 1, wherein, Upon reaching a preset time node, the initial SOC information of multiple electric vehicles within the electric vehicle cluster is collected, and driving data sets within a preset time range are continuously collected, including: When a preset time node is reached, the initial SOC information of multiple electric vehicles in the electric vehicle cluster is collected; Within a preset time range after the preset time node, the driving speed records, driving time records, and driving location records of the multiple electric vehicles are collected as multiple types of driving feature information to obtain multiple driving datasets. 3.The method of claim 1, wherein, Based on driving and charging / discharging record data from multiple historical periods of electric vehicles, an integrated charging / discharging predictor is constructed, including: Based on driving and charging / discharging records from multiple electric vehicles over a historical period, we collected a set of sample driving speed records, a set of sample driving time records, a set of sample driving location records, a set of sample discharge load information, and a set of sample charging time information. The sample driving speed record set, sample driving time record set, and sample driving location record set are respectively used as inputs and combined with the sample discharge load information set and sample charging time information set as outputs to obtain the discharge speed training dataset, charging speed training dataset, discharge time training dataset, charging time training dataset, discharge location training dataset, and charging location training dataset. Based on ensemble learning, multiple discharge prediction paths and multiple charge prediction paths are trained to obtain an integrated charge and discharge predictor. 4.The method of claim 3, wherein, Based on ensemble learning, multiple discharge prediction paths and multiple charge prediction paths are trained to obtain an integrated charge / discharge predictor, including: A preset proportion of training data is randomly extracted from the discharge velocity training dataset to train the first velocity discharge prediction branch within the velocity discharge prediction path. A preset proportion of training data is randomly extracted from the discharge velocity training dataset to train the second velocity discharge prediction branch. Continue training to obtain multiple velocity discharge prediction branches, and integrate them to obtain the first discharge prediction path; The charging speed training dataset, discharging time training dataset, charging time training dataset, discharging location training dataset, and charging location training dataset are used to train and obtain multiple discharging prediction paths and multiple charging prediction paths, thus obtaining an integrated charge and discharge predictor.

5. The method for dynamic prediction of charging and discharging load of electric vehicle clusters according to claim 4, characterized in that, Based on multiple driving datasets, multiple discharge prediction paths within the integrated charge / discharge predictor are used to predict multiple discharge load information. Combined with multiple initial SOC information, multiple predicted SOC information is calculated. Using the multiple charging prediction paths, multiple charging time information is predicted. Combining the multiple predicted SOC information, multiple charging load information is analyzed, including: The multiple driving datasets are respectively input into the multiple discharge prediction paths in the integrated charge and discharge predictor for prediction, and the average value of the prediction results of multiple discharge prediction branches is calculated to obtain multiple discharge load information sets. The multiple discharge prediction paths are tested to obtain multiple discharge prediction accuracies, and the multiple discharge load information sets are weighted and calculated to obtain multiple discharge load information. Based on the multiple discharge load information and the multiple initial SOC information, multiple predicted SOC information are calculated and obtained; The multiple driving datasets are input into the multiple charging prediction paths in the integrated charging and discharging predictor for prediction, and the average of the prediction results of the multiple charging prediction branches is calculated to obtain multiple sets of charging time information. The multiple charging prediction paths are tested to obtain multiple charging prediction accuracies, and the multiple charging time information sets are weighted and calculated to obtain multiple charging time information. Based on the multiple predicted SOC information, multiple charging SOC information is calculated, and combined with the multiple charging time information, multiple charging load information is analyzed and calculated.

6. The method for dynamic prediction of charging and discharging load of electric vehicle clusters according to claim 1, characterized in that, Based on the multiple charging load information, the charging power of the multiple electric vehicles is predicted and optimized to obtain multiple optimal predicted charging power information, including: Obtain the charging power space of the multiple electric vehicles; To improve the charging capacity of multiple electric vehicles and reduce overheating during charging, a cluster charging function is constructed to predict and optimize the charging power of multiple electric vehicles, as shown in the following equation: Where CCF is the cluster fitness, w1 and w2 are the weights, and M is the number of electric vehicles in the electric vehicle cluster. To optimize the predicted charging power information of the i-th electric vehicle, For the charging load information of the i-th electric vehicle, T y The standard temperature for electric vehicle batteries, To optimize the charging overheating temperature of the i-th electric vehicle charging according to the predicted charging power information; Within the electric vehicle charging power space, the charging power of the multiple electric vehicles is predicted and optimized according to the cluster charging function to obtain multiple optimal predicted charging power information.

7. The method of claim 6, wherein the method further comprises: Within the electric vehicle charging power space, the charging power of the multiple electric vehicles is predicted and optimized according to the cluster charging function, including: Within the electric vehicle charging power space, multiple first predicted charging power information for multiple electric vehicles are randomly generated; Based on the multiple first predicted charging power information, multiple first charging overheating temperatures are predicted and obtained. Combining the multiple charging load information, a first cluster fitness is calculated based on the cluster charging function. In this process, by collecting a set of sample charging power and a set of sample charging overheating temperatures, a charging temperature predictor is trained to predict multiple first charging overheating temperatures. Based on the deviation range of multiple first charging overheat temperatures and first average charging overheat temperatures, multiple adjustment ranges are set. Within the electric vehicle charging power space, the multiple first predicted charging power information is adjusted to obtain multiple second predicted charging power information, and the second cluster fitness is calculated. Continue to predict and optimize the charging power of multiple electric vehicles within the charging power space until convergence, and output the multiple predicted charging power information with the highest cluster fitness to obtain multiple optimal predicted charging power information.

8. An electric vehicle cluster charge-discharge load dynamic prediction device characterized by comprising: For implementing the method for dynamic prediction of charging and discharging load of an electric vehicle cluster as described in any one of claims 1-7, the device comprises: A module for acquiring similar electric vehicles, which is used to acquire multiple similar electric vehicles that are connected through a clustering protocol to form an electric vehicle cluster. The driving data acquisition module is used to collect the initial SOC information of multiple electric vehicles in the electric vehicle cluster when a preset time node is reached, and to continuously collect driving data sets within a preset time range, wherein each driving behavior data set includes multiple types of driving feature information. A predictor building module is used to build an integrated charge-discharge predictor based on driving record data and charge-discharge record data of multiple electric vehicles over a historical period. The integrated charge-discharge predictor includes multiple discharge prediction paths and multiple charging prediction paths corresponding to multiple types of driving feature information. Each discharge prediction path and charging prediction path includes multiple discharge prediction branches and multiple charging prediction branches. The charging load acquisition module is used to predict multiple discharge load information based on multiple driving data sets and multiple discharge prediction paths in the integrated charge and discharge predictor, calculate multiple predicted SOC information by combining multiple initial SOC information, predict multiple charging time information by using the multiple charging prediction paths, and analyze multiple charging load information by combining the multiple predicted SOC information. The prediction result acquisition module is used to predict and optimize the charging power of the multiple electric vehicles based on the multiple charging load information, obtain multiple optimal predicted charging power information, control the charging power when the multiple electric vehicles are charged by charging piles, and combine the multiple discharge load information and the multiple charging load information as multiple charge and discharge load prediction results.

Images