New energy vehicle battery charging state management method and system based on artificial intelligence

Abstract

The present invention discloses a method and system for managing the battery charging state of a new energy vehicle based on artificial intelligence, and the method includes data collection, data preprocessing, charging demand prediction, charging strategy optimization and charging safety management. The present invention relates to the technical field of battery management for new energy vehicles, and specifically refers to a method and system for managing the battery charging state of a new energy vehicle based on artificial intelligence. The present solution adopts a battery charging state management method that combines long-term and short-term charging demand prediction with charging strategy optimization. By comprehensively predicting the demand for battery charging and the specific strategy for battery charging, and considering the user's personal charging strategy, the intelligence and feasibility of the method are improved; an integrated classification method combining modal decomposition time series convolution and long-term and short-term memory is used to predict charging demand; and a reinforcement learning method for multi-objective optimization of charging strategy is used to optimize charging strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy vehicle battery management, and specifically to a new energy vehicle battery charging state management method and system based on artificial intelligence. Background Art

[0002] The AI-based new energy vehicle battery charging state management method refers to a system that uses AI technologies, such as machine learning and deep learning algorithms, to monitor, predict and optimize the battery of electric vehicles in real time. Its core functions include: predicting the life and performance of the battery through intelligent algorithms, optimizing the charging strategy to extend the battery life, ensuring the safety and efficiency of the charging process, and providing users with personalized charging suggestions and management solutions through data analysis, thereby improving the overall performance and economy of new energy vehicles.

[0003] However, in the existing new energy vehicle battery charging state management methods, there are traditional intelligent methods that focus on the battery state assessment and overall data management, but lack the perspective of charging demand and strategy analysis from the user's perspective for charging management, which leads to insufficient data dimensions for battery charging state management and reduces the technical problem of the depth of analysis of comprehensive analysis management of new energy vehicle batteries; in the existing charging demand prediction methods, there are new energy vehicle battery charging demand that needs to consider both short-term charging specific time analysis and trend demand long-term analysis, while the existing intelligent technology is difficult to take both into account at the same time. In the commonly used time series feature mode decomposition process, it is also prone to noise residual and pseudo-mode interference, which affects the accuracy and scalability of charging demand prediction; in the existing charging strategy optimization methods, there is the existing new energy vehicle charging is a multi-dimensional charging demand process, which may involve charging speed, charging battery protection, and may also involve comprehensive consideration of charging cost, while traditional strategy optimization algorithms often focus on optimizing a fixed established strategy, ignoring the technical problem of multiple possible objectives. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a new energy vehicle battery charging state management method and system based on artificial intelligence. In view of the fact that in the existing new energy vehicle battery charging state management method, the traditional intelligent method focuses on the battery state evaluation and overall data management, but lacks the perspective of charging demand and strategy analysis from the user's perspective for charging management, which leads to insufficient data dimensions for battery charging state management and reduces the technical problem of the depth of analysis of comprehensive analysis management of new energy vehicle batteries. This solution creatively adopts a battery charging state management method that combines long-term and short-term charging demand prediction and charging strategy optimization. By comprehensively predicting the battery charging demand and the specific battery charging strategy, combined with considering the user's personal charging strategy, the intelligence and feasibility of the method are improved, and the research and practice path of new energy vehicle battery charging state management is broadened; in view of the fact that in the existing charging demand prediction method, the battery charging demand of new energy vehicles needs to consider both the specific charging time analysis in the short term and the long-term analysis of trend demand, and the existing intelligent technology is difficult to take both predictions into account at the same time, in the commonly used time series feature model In the process of state decomposition, noise residue and pseudo-modal interference are prone to occur, which affects the accuracy and scalability of charging demand prediction. This solution creatively adopts an integrated classification method combining modal decomposition time series convolution and long short-term memory to predict charging demand. By introducing white noise to improve modal decomposition, and using time series convolution and long short-term memory networks to extract short-term and long-term features respectively, the predictable dimension of charging demand prediction and the referenceability of the results are optimized, and good data support is provided for subsequent charging strategy optimization. In the existing charging strategy optimization methods, there is a problem that the existing new energy vehicle charging is a multi-dimensional charging demand process, which may involve charging speed, charging battery protection, and charging cost. The comprehensive consideration, while the traditional strategy optimization algorithm often focuses on optimizing a fixed established strategy, ignoring the technical problems of multiple objectives that may exist. This solution creatively adopts a reinforcement learning method for multi-objective optimization of charging strategy to optimize the charging strategy. While performing multi-objective optimization through demand optimization modeling, it provides a flexible demand weight adjustment strategy, improves the meticulousness of charging strategy optimization, and provides an important management basis for charging state management.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a new energy vehicle battery charging state management method based on artificial intelligence, the method comprising the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data preprocessing;

[0008] Step S3: Charging demand prediction;

[0009] Step S4: Charging strategy optimization;

[0010] Step S5: Charging safety management.

[0011] Further, in step S1, the data collection is used to collect the original data set required for the new energy vehicle battery charging state management, specifically, by collecting the vehicle-mounted records and driving environment data records to obtain the charging state management original data set;

[0012] The charging status management raw data set specifically includes vehicle raw data, driving raw data, environmental raw data, vehicle owner behavior data and historical charging data.

[0013] Further, in step S2, the data preprocessing is used to optimize and process the original data set, specifically, performing outlier processing, data cleaning, feature extraction, data normalization and time series processing operations on the charging state management original data set in sequence to obtain an optimized charging state management data set;

[0014] The optimized charging state management data set specifically includes charging demand prediction time series characteristic data, charging strategy optimization characteristic data and cleaning optimization original data.

[0015] Further, in step S3, the charging demand prediction is used to provide long-term and short-term charging demand predictions, specifically, based on the optimized charging state management data set, an integrated classification method combining modal decomposition time series convolution and long short-term memory is used to perform charging demand prediction to obtain charging demand prediction reference data;

[0016] The integrated classification method combining modal decomposition temporal convolution and long short-term memory specifically includes an empirical mode decomposition operator, a temporal convolution subnet, a long short-term memory subnet and an integrated classification subnet;

[0017] The step of using an integrated classification method combining modal decomposition time series convolution and long short-term memory to predict charging demand and obtain charging demand prediction reference data includes:

[0018] Step S31: improving the ensemble empirical mode decomposition, specifically by introducing an initialization white noise with an expected value and a deviation of zero in the initialization step of the standard ensemble empirical mode decomposition process, optimizing the initialization data, and then performing ensemble empirical mode decomposition on the characteristic data in the optimized charging state management data set to obtain characteristic mode component data, wherein the characteristic mode component data specifically includes a mode component and a residual number;

[0019] Step S32: constructing a temporal convolutional subnet for extracting long-term and short-term time step features, specifically constructing a standard temporal convolutional network as the temporal convolutional subnet, and extracting features of long-term and short-term time steps by introducing an extended convolutional layer to obtain temporal convolutional feature data;

[0020] Step S33: constructing a long short-term memory subnet to optimize the processing of long time series data, specifically constructing a standard long short-term memory neural network including a forget gate, an input gate and an output gate as a long short-term memory subnet, and performing time dependency decomposition according to the long time series feature data in the time series convolution feature data to obtain long time series optimized feature data;

[0021] Step S34: construct an integrated classification subnet, specifically a three-structure classification subnet using a support vector machine, a random forest model, and an extreme gradient boosting model, and obtain integrated classification output data through mean prediction. The calculation formula is:

[0022] ;

[0023] In the formula, is the integrated classification output data, It is the input data of the integrated classification subnet, which is used to represent the feature connection data of long time series optimization feature data and time series convolution feature data, argmax n is the summation function of the classification subnet prediction results n, Q is the total number of classification subnets, and the specific value here is 3. q is the classification subnet index. When q is 1, it indicates the support vector machine subnet, when q is 2, it indicates the random forest model subnet, and when q is 3, it indicates the extreme gradient boosting model subnet. I(·) is the subnet output normalization function, and f q (·) is the prediction function of the qth subnetwork;

[0024] Step S35: training the charging demand prediction model, specifically, training the charging demand prediction model by improving the set empirical mode decomposition, building the time series convolution subnet, building the long short-term memory subnet and building the integrated classification subnet, and applying the model loss function to optimize the training process to obtain the charging demand prediction model Model OP ;

[0025] Step S36: Charging demand prediction, specifically using the charging demand prediction model Model OP , performing charging demand prediction based on the optimized charging state management data set to obtain charging demand prediction reference data;

[0026] The charging demand forecast reference data specifically includes a short-term charging demand forecast reference and a long-term charging demand trend reference;

[0027] The short-term charging demand forecast reference specifically refers to the real-time predicted charging demand time and charging demand mode; the charging demand time specifically refers to the predicted time when charging is required next time, and the charging demand mode specifically includes a fast charging mode and a slow charging mode;

[0028] The long-term charging demand trend reference specifically refers to the charging demand trend in the future, which is used for battery power consumption warning and prompting.

[0029] Further, in step S4, the charging strategy optimization is used to generate and optimize a multi-objective charging strategy, specifically, based on the optimized charging state management data set and the charging demand prediction reference data, a reinforcement learning method for multi-objective optimization of the charging strategy is used to optimize the charging strategy to obtain charging strategy optimization reference data, specifically including the following steps:

[0030] Step S41: multi-objective demand modeling of charging strategy, specifically constructing a multi-objective optimization demand model for charging measurement and recording optimization, and designing a multi-objective reward function based on the multi-objective optimization demand model, wherein the multi-objective optimization demand model specifically includes the objectives of maximizing charging speed, extending battery life, and minimizing charging cost;

[0031] Step S42: defining initial parameters of reinforcement learning, specifically constructing a reinforcement learning parameter combination and a multi-objective reward function, wherein the reinforcement learning parameter combination specifically includes state parameters, action parameters and reward parameters;

[0032] The state parameters specifically refer to the current state parameters of the vehicle, including battery power, remaining charge, current charging power and battery temperature;

[0033] The action parameters specifically refer to the current charging strategy of the vehicle, including fast charging action, slow charging action and charging power adjustment action;

[0034] The reward parameter specifically refers to the output of the multi-objective reward function;

[0035] Step S43: constructing a basic network structure for reinforcement learning, specifically, sequentially using an actor network and a critic network as the basic network structure for reinforcement learning, wherein the actor network is used to select a charging strategy, specifically, outputting action parameters according to state parameters, and the critic network is used to evaluate the value of the current charging strategy, specifically, calculating a state value function and an advantage function according to state parameters and action parameters, and obtaining a basic network structure for reinforcement learning and an optimized output of a charging strategy by performing strategy updates and value evaluations;

[0036] Step S44: updating the dynamic weights of multiple objectives of the charging strategy, specifically constructing a multi-objective reward weight update rule, and dynamically adjusting the multi-objective weights according to the multi-objective reward weight update rule to obtain a dynamic weight update output of the charging strategy;

[0037] The multi-objective reward weight update rule specifically includes a fast charging weight sequence, a battery protection weight sequence and a cost saving weight sequence;

[0038] The calculation formula of the fast charging weight sequence is:

[0039] ;

[0040] In the formula, is the target reward weight for maximizing charging speed;

[0041] The calculation formula of the battery protection weight sequence is:

[0042] ;

[0043] In the formula, is the battery life extension target reward weight;

[0044] The calculation formula of the cost saving weight sequence is:

[0045] ;

[0046] In the formula, is the objective reward weight for minimizing charging cost;

[0047] Step S45: strategy evaluation calculation, specifically, performing strategy evaluation after each round of reinforcement learning training, and combining the Pareto optimal evaluation method to calculate the optimal balance point to obtain the optimal strategy parameter combination;

[0048] Step S46: Charging strategy optimization model training, specifically, through the charging strategy multi-objective demand modeling, the reinforcement learning basic network structure, the charging strategy multi-objective dynamic weight update and the strategy evaluation calculation, the charging strategy optimization model training is performed to obtain the charging strategy optimization model Model OC ;

[0049] Step S47: Charging strategy optimization, specifically using the charging strategy optimization model Model OC , based on the optimized charging state management data set and the charging demand prediction reference data, the charging strategy is optimized to obtain the charging strategy optimization reference data.

[0050] Furthermore, in step S5, the charging safety management is used to perform comprehensive charging safety management in combination with the results of long-term and short-term charging demand forecasts and the results of multi-objective charging strategy optimization. Specifically, charging safety management is performed in combination with the charging demand forecast reference data and the charging strategy optimization reference data, and real-time battery charging status management and charging status monitoring are performed by setting battery status thresholds to obtain battery status management reference data.

[0051] The artificial intelligence-based new energy vehicle battery charging state management system provided by the present invention includes a data collection module, a data preprocessing module, a charging demand prediction module, a charging strategy optimization module and a charging safety management module;

[0052] The data collection module is used for data collection, and obtains a charging state management original data set through data collection, and sends the charging state management original data set to the data preprocessing module;

[0053] The data preprocessing module is used for data preprocessing, and obtains an optimized charging state management data set through data preprocessing, and sends the optimized charging state management data set to the charging demand prediction module and the charging strategy optimization module;

[0054] The charging demand prediction module is used for charging demand prediction, obtains charging demand prediction reference data through charging demand prediction, and sends the charging demand prediction reference data to the charging strategy optimization module and the charging safety management module;

[0055] The charging strategy optimization module is used for charging strategy optimization, obtains charging strategy optimization reference data through charging strategy optimization, and sends the charging strategy optimization reference data to the charging safety management module;

[0056] The charging safety management module is used for charging safety management, and obtains battery status management reference data through charging safety management.

[0057] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0058] (1) In the existing new energy vehicle battery charging status management methods, there are traditional intelligent methods that focus on the battery status assessment and overall data management, but lack the perspective of charging management based on charging demand and strategy analysis from the user's perspective, which leads to insufficient data dimensions for battery charging status management and reduces the technical problem of the depth of analysis of comprehensive analysis and management of new energy vehicle batteries. This solution creatively adopts a battery charging status management method that combines long-term and short-term charging demand prediction and charging strategy optimization. By comprehensively predicting the battery charging demand and the specific battery charging strategy, combined with the user's personal charging strategy, the intelligence and feasibility of the method are improved, and the research and practice path of new energy vehicle battery charging status management is broadened;

[0059] (2) In the existing charging demand prediction methods, the battery charging demand of new energy vehicles needs to consider both the specific charging time analysis in the short term and the long-term analysis of the trend demand. The existing intelligent technology is difficult to take both into account at the same time. In the commonly used time series feature modal decomposition process, it is also easy to have noise residue and pseudo-modal interference, which affects the accuracy and scalability of charging demand prediction. This solution creatively adopts an integrated classification method that combines modal decomposition, time series convolution and long short-term memory to predict charging demand. By introducing white noise to improve modal decomposition, and using time series convolution and long short-term memory networks to extract short-term and long-term features respectively, the predictable dimension of charging demand prediction and the referenceability of the results are optimized, and good data support is provided for subsequent charging strategy optimization.

[0060] (3) In the existing charging strategy optimization methods, there is a problem that the existing new energy vehicle charging is a multi-dimensional charging demand process, which may involve charging speed, charging battery protection, and comprehensive consideration of charging cost. Traditional strategy optimization algorithms often focus on optimizing a fixed established strategy, ignoring the technical problems of multiple objectives that may exist. This scheme creatively adopts the reinforcement learning method of multi-objective optimization of charging strategy to optimize the charging strategy. While performing multi-objective optimization through demand optimization modeling, it provides a flexible demand weight adjustment strategy, improves the meticulousness of charging strategy optimization, and provides an important management basis for charging status management. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of the process of the artificial intelligence-based new energy vehicle battery charging state management method provided by the present invention;

[0062] Figure 2 A schematic diagram of an artificial intelligence-based new energy vehicle battery charging state management system provided by the present invention;

[0063] Figure 3 This is a schematic diagram of the process of data preprocessing in step S2;

[0064] Figure 4 This is a schematic diagram of the process of charging demand prediction in step S3;

[0065] Figure 5 Schematic diagram of the process of optimizing the charging strategy in step S4.

[0066] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0068] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0069] Example 1, see Figure 1 The present invention provides a new energy vehicle battery charging state management method based on artificial intelligence, the method comprising the following steps:

[0070] Step S1: data collection;

[0071] Step S2: data preprocessing;

[0072] Step S3: Charging demand prediction;

[0073] Step S4: Charging strategy optimization;

[0074] Step S5: Charging safety management.

[0075] By executing the above operations, in the existing new energy vehicle battery charging status management methods, there are traditional intelligent methods that focus on the battery status evaluation and overall data management, but lack the perspective of charging demand and strategy analysis from the user's perspective for charging management, which leads to insufficient data dimensions for battery charging status management and reduces the technical problem of the depth of analysis of comprehensive analysis management of new energy vehicle batteries. This solution creatively adopts a battery charging status management method that combines long-term and short-term charging demand prediction and charging strategy optimization. By comprehensively predicting the battery charging demand and the specific battery charging strategy, combined with considering the user's personal charging strategy, the intelligence and feasibility of the method are improved, and the research and practice path of new energy vehicle battery charging status management is broadened.

[0076] Example 2, see Figure 1 and Figure 2 In step S1, the data collection is used to collect the original data set required for the new energy vehicle battery charging state management, specifically, the original data set for charging state management is obtained through vehicle-mounted recording and driving environment data recording and collection;

[0077] The charging status management raw data set specifically includes vehicle raw data, driving raw data, environmental raw data, vehicle owner behavior data and historical charging data;

[0078] The vehicle-mounted raw data specifically refers to the battery status data collected by the vehicle-mounted sensor, including voltage, current, temperature and battery internal resistance parameter data;

[0079] The original driving data refers specifically to vehicle driving status record data, including driving speed, acceleration mode, braking frequency, driving time and driving distance data;

[0080] The environmental raw data specifically refers to external data of environmental factors, including temperature parameters and humidity parameters;

[0081] The original behavior data of the vehicle owner specifically refers to the personalized behavior data of the vehicle owner, including charging habit data, travel time data and travel frequency data;

[0082] The historical charging data specifically includes historical charging duration data, historical charging power data and historical charging times data.

[0083] Example 3, see Figure 1 , Figure 2 and Figure 3This embodiment is based on the above embodiment. In step S2, the data preprocessing is used to optimize and process the original data set, specifically, outlier processing, data cleaning, feature extraction, data normalization and time series processing operations are sequentially performed on the charging state management original data set to obtain an optimized charging state management data set, which specifically includes the following steps:

[0084] Step S21: outlier processing, specifically refers to removing the outlier data in the charging state management original data set, and using the mean filling method to fill the outlier to obtain outlier optimized data;

[0085] Step S22: data cleaning, specifically removing data noise and filling missing values ​​from the outlier optimized data to obtain cleaned optimized data;

[0086] Step S23: feature extraction, specifically using an artificial feature extraction method to extract features from the cleaned and optimized data to obtain an original feature data set;

[0087] The original feature data set specifically includes battery health features, power consumption pattern features, charging frequency features, and owner behavior features;

[0088] Step S24: data normalization, specifically, performing Z-mean normalization processing on the original feature data set to obtain a standardized feature data set;

[0089] Step S25: time series data processing, specifically, performing time series data conversion on the cleaning optimization data according to the timestamp to obtain time series conversion raw data, and performing time series modeling on the standardized feature data set and the time series conversion raw data to obtain time series feature data;

[0090] Step S26: data preprocessing, specifically, performing comprehensive data preprocessing through the abnormal value processing, the data cleaning, the feature extraction, the data normalization and the time series data processing to obtain an optimized charging state management data set;

[0091] The optimized charging state management data set specifically includes charging demand prediction time series characteristic data, charging strategy optimization characteristic data and cleaning optimization original data.

[0092] Example 4, see Figure 1 , Figure 2 and Figure 4 , This embodiment is based on the above embodiment. In step S3, the charging demand prediction is used to provide long-term and short-term charging demand predictions. Specifically, based on the optimized charging state management data set, an integrated classification method combining modal decomposition time series convolution and long short-term memory is used to perform charging demand prediction to obtain charging demand prediction reference data;

[0093] The integrated classification method combining modal decomposition temporal convolution and long short-term memory specifically includes an empirical mode decomposition operator, a temporal convolution subnet, a long short-term memory subnet and an integrated classification subnet;

[0094] The step of using an integrated classification method combining modal decomposition time series convolution and long short-term memory to predict charging demand and obtain charging demand prediction reference data includes:

[0095] Step S31: improving the ensemble empirical mode decomposition, specifically by introducing an initialization white noise with an expected value and a deviation of zero in the initialization step of the standard ensemble empirical mode decomposition process, optimizing the initialization data, and then performing ensemble empirical mode decomposition on the characteristic data in the optimized charging state management data set to obtain characteristic mode component data, wherein the characteristic mode component data specifically includes a mode component and a residual number;

[0096] Step S32: construct a temporal convolutional subnet to extract long-term and short-term time step features. Specifically, a standard temporal convolutional network is constructed as the temporal convolutional subnet, and the extended convolutional layer is introduced to extract the features of long-term and short-term time steps to obtain temporal convolutional feature data. The calculation formula is:

[0097] ;

[0098] Where o(·) is the output of the convolutional subnetwork at time step t, t is the time step index, which is used to represent the temporal convolution feature data, k is the total number of temporal convolution layers, i is the temporal convolution layer index, w(i) is the convolution kernel weight of the i-th temporal convolution layer, x(·) is the input sequence convolution function, and d is the expansion coefficient, which is used to control the receptive range of the convolution to achieve feature extraction for short-term and long-term time steps;

[0099] Step S33: constructing a long short-term memory subnet to optimize the processing of long time series data, specifically constructing a standard long short-term memory neural network including a forget gate, an input gate and an output gate as a long short-term memory subnet, and performing time dependency decomposition according to the long time series feature data in the time series convolution feature data to obtain long time series optimized feature data;

[0100] Step S34: construct an integrated classification subnet, specifically a three-structure classification subnet using a support vector machine, a random forest model, and an extreme gradient boosting model, and obtain integrated classification output data through mean prediction. The calculation formula is:

[0101] ;

[0102] In the formula, is the integrated classification output data, It is the input data of the integrated classification subnet, which is used to represent the feature connection data of long time series optimization feature data and time series convolution feature data, argmax n is the summation function of the classification subnet prediction results n, Q is the total number of classification subnets, and the specific value here is 3. q is the classification subnet index. When q is 1, it indicates the support vector machine subnet, when q is 2, it indicates the random forest model subnet, and when q is 3, it indicates the extreme gradient boosting model subnet. I(·) is the subnet output normalization function, and f q (·) is the prediction function of the qth subnetwork;

[0103] Step S35: training the charging demand prediction model, specifically, training the charging demand prediction model by improving the set empirical mode decomposition, building the time series convolution subnet, building the long short-term memory subnet and building the integrated classification subnet, and applying the model loss function to optimize the training process to obtain the charging demand prediction model Model OP ;

[0104] Step S36: Charging demand prediction, specifically using the charging demand prediction model Model OP , performing charging demand prediction based on the optimized charging state management data set to obtain charging demand prediction reference data;

[0105] The charging demand forecast reference data specifically includes a short-term charging demand forecast reference and a long-term charging demand trend reference;

[0106] The short-term charging demand forecast reference specifically refers to the real-time predicted charging demand time and charging demand mode; the charging demand time specifically refers to the predicted time when charging is required next time, and the charging demand mode specifically includes a fast charging mode and a slow charging mode;

[0107] The long-term charging demand trend reference specifically refers to the charging demand trend in the future, which is used for battery power consumption warning and prompting.

[0108] By performing the above operations, in view of the fact that in the existing charging demand prediction methods, the battery charging demand of new energy vehicles needs to consider both the specific charging time analysis in the short term and the long-term analysis of the trend demand, while the existing intelligent technology is difficult to take into account the prediction of both at the same time. In the commonly used time series feature modal decomposition process, noise residue and pseudo-modal interference are prone to occur, which affects the accuracy and scalability of charging demand prediction. The scheme creatively adopts an integrated classification method combining modal decomposition, time series convolution and long short-term memory to predict charging demand, improves modal decomposition by introducing white noise, and uses time series convolution and long short-term memory networks to extract short-term and long-term features respectively, which optimizes the predictable dimension of charging demand prediction and the reference of the results, and also provides good data support for subsequent charging strategy optimization.

[0109] Example 5, see Figure 1 , Figure 2 and Figure 5 This embodiment is based on the above embodiment. In step S4, the charging strategy optimization is used to generate and optimize a multi-objective charging strategy. Specifically, the charging strategy is optimized based on the optimized charging state management data set and the charging demand prediction reference data by using a reinforcement learning method for multi-objective optimization of the charging strategy to obtain charging strategy optimization reference data. Specifically, the following steps are included:

[0110] Step S41: multi-objective demand modeling of charging strategy, specifically constructing a multi-objective optimization demand model for charging measurement and recording optimization, and designing a multi-objective reward function based on the multi-objective optimization demand model, wherein the multi-objective optimization demand model specifically includes the objectives of maximizing charging speed, extending battery life, and minimizing charging cost;

[0111] Step S42: defining initial parameters of reinforcement learning, specifically constructing a reinforcement learning parameter combination and a multi-objective reward function, wherein the reinforcement learning parameter combination specifically includes state parameters, action parameters and reward parameters;

[0112] The state parameters specifically refer to the current state parameters of the vehicle, including battery power, remaining charge, current charging power and battery temperature;

[0113] The action parameters specifically refer to the current charging strategy of the vehicle, including fast charging action, slow charging action and charging power adjustment action;

[0114] The reward parameter specifically refers to the output of the multi-objective reward function;

[0115] The calculation formula of the multi-objective reward function is:

[0116] ;

[0117] In the formula, r t is the reward parameter, is the target reward weight for maximizing charging speed, R speed (·) is the objective reward function for maximizing charging speed, a t is the action parameter, s t is the state parameter, t is the time step index, is the battery life extension objective reward weight, R life (·) is the battery life extension objective reward function, is the charging cost minimization objective reward weight, R cost (·) is the objective reward function for minimizing charging cost;

[0118] Step S43: constructing a basic network structure for reinforcement learning, specifically, sequentially using an actor network and a critic network as the basic network structure for reinforcement learning, wherein the actor network is used to select a charging strategy, specifically, outputting action parameters according to state parameters, and the critic network is used to evaluate the value of the current charging strategy, specifically, calculating a state value function and an advantage function according to state parameters and action parameters, and obtaining a basic network structure for reinforcement learning and an optimized output of a charging strategy by performing strategy updates and value evaluations;

[0119] Step S44: updating the dynamic weights of multiple objectives of the charging strategy, specifically constructing a multi-objective reward weight update rule, and dynamically adjusting the multi-objective weights according to the multi-objective reward weight update rule to obtain a dynamic weight update output of the charging strategy;

[0120] The multi-objective reward weight update rule specifically includes a fast charging weight sequence, a battery protection weight sequence and a cost saving weight sequence;

[0121] The calculation formula of the fast charging weight sequence is:

[0122] ;

[0123] In the formula, is the target reward weight for maximizing charging speed;

[0124] The calculation formula of the battery protection weight sequence is:

[0125] ;

[0126] In the formula, is the battery life extension target reward weight;

[0127] The calculation formula of the cost saving weight sequence is:

[0128] ;

[0129] In the formula, is the objective reward weight for minimizing charging cost;

[0130] Step S45: strategy evaluation calculation, specifically, performing strategy evaluation after each round of reinforcement learning training, and combining the Pareto optimal evaluation method to calculate the optimal balance point to obtain the optimal strategy parameter combination;

[0131] Step S46: Charging strategy optimization model training, specifically, through the charging strategy multi-objective demand modeling, the reinforcement learning basic network structure, the charging strategy multi-objective dynamic weight update and the strategy evaluation calculation, the charging strategy optimization model training is performed to obtain the charging strategy optimization model Model OC ;

[0132] Step S47: Charging strategy optimization, specifically using the charging strategy optimization model Model OC , based on the optimized charging state management data set and the charging demand prediction reference data, the charging strategy is optimized to obtain the charging strategy optimization reference data.

[0133] By executing the above operations, in the existing charging strategy optimization method, there is a problem that the existing new energy vehicle charging is a multi-dimensional charging demand process, which may involve charging speed, charging battery protection, and may also involve comprehensive consideration of charging cost. Traditional strategy optimization algorithms often focus on optimizing a fixed established strategy, ignoring the technical problems of multiple objectives that may exist. This solution creatively adopts the reinforcement learning method of multi-objective optimization of charging strategy to optimize the charging strategy. While performing multi-objective optimization through demand optimization modeling, it provides a flexible demand weight adjustment strategy, which improves the meticulousness of charging strategy optimization and provides an important management basis for charging status management.

[0134] Example 6, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S5, the charging safety management is used to combine the results of long-term and short-term charging demand predictions and the results of multi-objective charging strategy optimization to perform comprehensive charging safety management. Specifically, the charging demand prediction reference data and the charging strategy optimization reference data are combined to perform charging safety management, and by setting the battery status threshold, real-time battery charging status management and charging status monitoring are performed to obtain battery status management reference data.

[0135] Embodiment 7, see Figure 1 and Figure 2, this embodiment is based on the above embodiment, and the artificial intelligence-based new energy vehicle battery charging state management system provided by the present invention includes a data collection module, a data preprocessing module, a charging demand prediction module, a charging strategy optimization module and a charging safety management module;

[0136] The data collection module is used for data collection, and obtains a charging state management original data set through data collection, and sends the charging state management original data set to the data preprocessing module;

[0137] The data preprocessing module is used for data preprocessing, and obtains an optimized charging state management data set through data preprocessing, and sends the optimized charging state management data set to the charging demand prediction module and the charging strategy optimization module;

[0138] The charging demand prediction module is used for charging demand prediction, obtains charging demand prediction reference data through charging demand prediction, and sends the charging demand prediction reference data to the charging strategy optimization module and the charging safety management module;

[0139] The charging strategy optimization module is used for charging strategy optimization, obtains charging strategy optimization reference data through charging strategy optimization, and sends the charging strategy optimization reference data to the charging safety management module;

[0140] The charging safety management module is used for charging safety management, and obtains battery status management reference data through charging safety management.

[0141] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0142] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0143] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A new energy vehicle battery charging state management method based on artificial intelligence, characterized in that: The method comprises the following steps: Step S1: Data collection, through vehicle-mounted records and driving environment data records, obtain the charging state management original data set; Step S2: data preprocessing to obtain an optimized charging state management data set; Step S3: Charging demand prediction, used to provide long-term and short-term charging demand prediction, specifically, based on the optimized charging state management data set, using an integrated classification method combining modal decomposition time series convolution and long short-term memory to perform charging demand prediction and obtain charging demand prediction reference data; The integrated classification method combining modal decomposition temporal convolution and long short-term memory specifically includes an empirical mode decomposition operator, a temporal convolution subnet, a long short-term memory subnet and an integrated classification subnet; The step of using an integrated classification method combining modal decomposition time series convolution and long short-term memory to predict charging demand and obtain charging demand prediction reference data includes: Step S31: improving the ensemble empirical mode decomposition, specifically by introducing an initialization white noise with an expected value and a deviation of zero in the initialization step of the standard ensemble empirical mode decomposition process, optimizing the initialization data, and then performing ensemble empirical mode decomposition on the characteristic data in the optimized charging state management data set to obtain characteristic mode component data, wherein the characteristic mode component data specifically includes a mode component and a residual number; Step S32: constructing a temporal convolutional subnet for extracting long-term and short-term time step features, specifically constructing a standard temporal convolutional network as the temporal convolutional subnet, and extracting features of long-term and short-term time steps by introducing an extended convolutional layer to obtain temporal convolutional feature data; Step S33: constructing a long short-term memory subnet to optimize the processing of long time series data, specifically constructing a standard long short-term memory neural network including a forget gate, an input gate and an output gate as a long short-term memory subnet, and performing time dependency decomposition according to the long time series feature data in the time series convolution feature data to obtain long time series optimized feature data; Step S34: construct an integrated classification subnet, specifically a three-structure classification subnet using a support vector machine, a random forest model, and an extreme gradient boosting model, and obtain integrated classification output data through mean prediction. The calculation formula is: In the formula, f y (x′) is the integrated classification output data, x′ is the input data of the integrated classification subnet, and is used to represent the feature connection data of the long time series optimization feature data and the time series convolution feature data, argmax n is the summation function of the classification subnet prediction results n, Q is the total number of classification subnets, and the specific value here is 3. q is the classification subnet index. When q is 1, it indicates the support vector machine subnet, when q is 2, it indicates the random forest model subnet, and when q is 3, it indicates the extreme gradient boosting model subnet. I(·) is the subnet output normalization function, and f q (·) is the prediction function of the qth subnetwork; Step S35: training the charging demand prediction model, specifically, training the charging demand prediction model by improving the set empirical mode decomposition, building the time series convolution subnet, building the long short-term memory subnet and building the integrated classification subnet, and applying the model loss function to optimize the training process to obtain the charging demand prediction model Model OP ; Step S36: Charging demand prediction, specifically using the charging demand prediction model Model OP , performing charging demand prediction based on the optimized charging state management data set to obtain charging demand prediction reference data; The charging demand forecast reference data specifically includes a short-term charging demand forecast reference and a long-term charging demand trend reference; The short-term charging demand forecast reference specifically refers to the real-time predicted charging demand time and charging demand mode; the charging demand time specifically refers to the predicted time when charging is required next time, and the charging demand mode specifically includes a fast charging mode and a slow charging mode; The long-term charging demand trend reference specifically refers to the charging demand trend in the future, which is used for battery power consumption warning and prompting; Step S4: Charging strategy optimization, which is used to generate and optimize a multi-objective charging strategy. Specifically, based on the optimized charging state management data set and the charging demand prediction reference data, a reinforcement learning method for multi-objective optimization of the charging strategy is used to optimize the charging strategy and obtain charging strategy optimization reference data. Specifically, the following steps are included: Step S41: multi-objective demand modeling of charging strategy, specifically constructing a multi-objective optimization demand model for charging measurement and recording optimization, and designing a multi-objective reward function based on the multi-objective optimization demand model, wherein the multi-objective optimization demand model specifically includes the objectives of maximizing charging speed, extending battery life, and minimizing charging cost; Step S42: defining initial parameters of reinforcement learning, specifically constructing a reinforcement learning parameter combination and a multi-objective reward function, wherein the reinforcement learning parameter combination specifically includes state parameters, action parameters and reward parameters; The state parameters specifically refer to the current state parameters of the vehicle, including battery power, remaining charge, current charging power and battery temperature; The action parameters specifically refer to the current charging strategy of the vehicle, including fast charging action, slow charging action and charging power adjustment action; The reward parameter specifically refers to the output of the multi-objective reward function; Step S43: constructing a basic network structure for reinforcement learning, specifically, sequentially using an actor network and a critic network as the basic network structure for reinforcement learning, wherein the actor network is used to select a charging strategy, specifically, outputting action parameters according to state parameters, and the critic network is used to evaluate the value of the current charging strategy, specifically, calculating a state value function and an advantage function according to state parameters and action parameters, and obtaining a basic network structure for reinforcement learning and an optimized output of a charging strategy by performing strategy updates and value evaluations; Step S44: updating the dynamic weights of multiple objectives of the charging strategy, specifically constructing a multi-objective reward weight update rule, and dynamically adjusting the weights of multiple objectives according to the multi-objective reward weight update rule, to obtain a dynamic weight update output of the charging strategy; Step S45: strategy evaluation calculation, specifically, performing strategy evaluation after each round of reinforcement learning training, and combining the Pareto optimal evaluation method to calculate the optimal balance point to obtain the optimal strategy parameter combination; Step S46: Charging strategy optimization model training, specifically, through the charging strategy multi-objective demand modeling, the reinforcement learning basic network structure, the charging strategy multi-objective dynamic weight update and the strategy evaluation calculation, the charging strategy optimization model training is performed to obtain the charging strategy optimization model Model OC ; Step S47: Charging strategy optimization, specifically using the charging strategy optimization model Model OC , optimizing the charging strategy according to the optimized charging state management data set and the charging demand prediction reference data, and obtaining charging strategy optimization reference data; Step S5: Charging safety management, combining the charging demand prediction reference data and the charging strategy optimization reference data to perform charging safety management, and by setting the battery status threshold, perform real-time battery charging status management and charging status monitoring to obtain battery status management reference data.

2. The method for managing the battery charging status of a new energy vehicle based on artificial intelligence according to claim 1, characterized in that: In step S1, the data collection is used to collect the original data set required for the new energy vehicle battery charging state management, specifically, the original data set for charging state management is obtained through vehicle-mounted recording and driving environment data recording and collection; The charging status management raw data set specifically includes vehicle raw data, driving raw data, environmental raw data, vehicle owner behavior data and historical charging data.

3. The method for managing the charging status of a new energy vehicle battery based on artificial intelligence according to claim 2 is characterized in that: In step S2, the data preprocessing is used to optimize and process the original data set, specifically, performing outlier processing, data cleaning, feature extraction, data normalization and time series processing operations on the charging state management original data set in sequence to obtain an optimized charging state management data set; The optimized charging state management data set specifically includes charging demand prediction time series characteristic data, charging strategy optimization characteristic data and cleaning optimization original data.

4. The method for managing the battery charging status of a new energy vehicle based on artificial intelligence according to claim 3 is characterized in that: In step S44, the multi-objective reward weight update rule specifically includes a fast charging weight sequence, a battery protection weight sequence, and a cost saving weight sequence; The calculation formula of the fast charging weight sequence is: λ1>λ2>λ3; Where λ1 is the target reward weight for maximizing charging speed; The calculation formula of the battery protection weight sequence is: λ2>λ1>λ3; Where λ2 is the target reward weight for extending battery life; The calculation formula of the cost saving weight sequence is: λ3>λ2>λ1; Where λ3 is the objective reward weight for minimizing charging cost.

5. The method for managing the charging status of a new energy vehicle battery based on artificial intelligence according to claim 4 is characterized in that: In step S5, the charging safety management is used to perform comprehensive charging safety management in combination with the results of long-term and short-term charging demand forecasts and the results of multi-objective charging strategy optimization. Specifically, charging safety management is performed in combination with the charging demand forecast reference data and the charging strategy optimization reference data, and real-time battery charging status management and charging status monitoring are performed by setting the battery status threshold to obtain battery status management reference data.

6. A new energy vehicle battery charging state management system based on artificial intelligence, used to implement the new energy vehicle battery charging state management method based on artificial intelligence as described in any one of claims 1 to 5, characterized in that: It includes data collection module, data preprocessing module, charging demand prediction module, charging strategy optimization module and charging safety management module.

7. The artificial intelligence-based new energy vehicle battery charging state management system according to claim 6 is characterized in that: The data collection module is used for data collection, and obtains a charging state management original data set through data collection, and sends the charging state management original data set to the data preprocessing module; The data preprocessing module is used for data preprocessing, and obtains an optimized charging state management data set through data preprocessing, and sends the optimized charging state management data set to the charging demand prediction module and the charging strategy optimization module; The charging demand prediction module is used for charging demand prediction, obtains charging demand prediction reference data through charging demand prediction, and sends the charging demand prediction reference data to the charging strategy optimization module and the charging safety management module; The charging strategy optimization module is used for charging strategy optimization, obtains charging strategy optimization reference data through charging strategy optimization, and sends the charging strategy optimization reference data to the charging safety management module; The charging safety management module is used for charging safety management, and obtains battery status management reference data through charging safety management.

Images