Multi-environment automobile battery management method and system

Abstract

The present invention provides a multi-environment automobile battery management method and system, which relates to the field of battery management technology, including obtaining environmental information and real-time operating parameters of the automobile, and performing multi-scale decomposition processing to extract local impact data of all environmental information. Then, based on the mutual information maximization criterion, the nonlinear dependency between each environmental information and battery consumption is calculated to determine the key environmental influencing factors. A sparse Bayesian regression model is used to process these influencing factors and automobile operating parameters, a dynamic power consumption prediction model is constructed, and real-time dynamic power consumption optimization is performed. Through a deep deterministic policy gradient algorithm, the power consumption allocation strategy is continuously updated to achieve adaptive battery management. The present invention not only improves the operating efficiency of the battery, but also enhances the adaptability in a changing environment. It can optimize the power consumption allocation strategy of the automobile in real time to ensure the optimal performance of the vehicle under various conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a multi-environment automobile battery management method and system. Background Art

[0002] With the development of electric vehicles, the battery management system has become one of the key links that affect the vehicle's endurance and overall performance. Most of the existing battery management methods are based on a single environmental parameter or a power consumption model under static conditions, which has obvious limitations in actual use scenarios. For example, when a vehicle is driving under different terrain and climate conditions, the battery consumption rate will be affected by a variety of complex environmental factors, such as terrain slope, road friction coefficient, climate temperature, etc. However, the existing technology usually only considers part of these factors and predicts the battery power consumption through simple empirical formulas or static models, ignoring the nonlinear effects of complex conditions on the battery under multiple environments. This leads to low prediction accuracy and cannot fully adapt to the actual operation needs of the vehicle in a changing environment. In addition, when optimizing battery management, the existing technology usually relies on fixed strategies and lacks the ability to adjust dynamically in real time. It cannot automatically adjust the battery power consumption distribution according to the changes in real-time environment and vehicle parameters, resulting in low battery efficiency and exacerbating excessive battery consumption.

[0003] Some current solutions attempt to improve the accuracy of power consumption prediction by adding sensors and data acquisition modules to obtain more environmental information, but these methods often use traditional linear regression or simple polynomial regression methods and have limited ability to process complex multi-dimensional environmental information.

[0004] Therefore, there is an urgent need for a multi-environment automobile battery management method and system to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-environment automobile battery management method and system to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] In a first aspect, the present application provides a multi-environment automobile battery management method, comprising:

[0007] Acquiring environmental information and real-time operating parameter information of the vehicle, wherein the environmental information includes climate information, terrain slope and road friction coefficient information, and the real-time operating parameter information of the vehicle includes vehicle load information and battery consumption information;

[0008] Perform multi-scale decomposition processing based on the environmental information and parameter information of the real-time operation of the vehicle, wherein local impact data of all environmental information is obtained by separating and extracting time-frequency domain features;

[0009] Perform feature selection on the local impact data of all environmental information according to the mutual information maximization criterion, wherein the key environmental impact factors are determined by calculating the nonlinear dependency between each environmental information and battery consumption;

[0010] The key environmental influencing factors and the parameter information of the real-time operation of the vehicle are processed according to sparse Bayesian regression, wherein the prediction model of battery consumption is built based on the key environmental influencing factors and the parameter information of the real-time operation of the vehicle, and the optimal model is obtained by automatically adjusting the feature weights to obtain a dynamic power consumption prediction model;

[0011] The dynamic power consumption of the vehicle is determined based on the real-time environmental information of the vehicle and the dynamic power consumption prediction model, and real-time dynamic power consumption optimization is performed according to the deep deterministic policy gradient algorithm to obtain the optimal power consumption allocation strategy.

[0012] In a second aspect, the present application also provides a multi-environment vehicle battery management system, including:

[0013] An acquisition unit, used to acquire environmental information and real-time operating parameter information of the vehicle, wherein the environmental information includes climate information, terrain slope and road friction coefficient information, and the real-time operating parameter information of the vehicle includes vehicle load information and battery consumption information;

[0014] A processing unit, configured to perform multi-scale decomposition processing based on the environmental information and parameter information of the real-time operation of the vehicle, wherein local impact data of all environmental information is obtained by separating and extracting time-frequency domain features;

[0015] A selection unit is used to perform feature selection on the local impact data of all environmental information according to a mutual information maximization criterion, wherein the key environmental impact factors are determined by calculating the nonlinear dependency between each environmental information and the battery consumption;

[0016] A prediction unit, used for processing key environmental influencing factors and parameter information of real-time operation of the vehicle according to sparse Bayesian regression, wherein a prediction model of battery consumption is established through the key environmental influencing factors and parameter information of real-time operation of the vehicle, and an optimal model is obtained by automatically adjusting feature weights to obtain a dynamic power consumption prediction model;

[0017] The optimization unit is used to determine the dynamic power consumption of the vehicle based on the real-time acquired environmental information of the vehicle operation and the dynamic power consumption prediction model, and perform real-time dynamic power consumption optimization according to the deep deterministic policy gradient algorithm to obtain the optimal power consumption allocation strategy.

[0018] The beneficial effects of the present invention are:

[0019] The present invention aims at the shortcomings of the prior art in automobile battery management under multi-environment conditions, and proposes a method for automobile battery management based on multi-environment information, which can effectively improve the accuracy of battery power consumption prediction and realize dynamic optimization. The method comprehensively considers multiple environmental factors such as climate information, terrain slope, road friction coefficient, etc., and combines the parameter information of real-time operation of the automobile such as vehicle load, battery consumption, etc., and uses adaptive multi-scale decomposition technology to process environmental information and operating parameters, and extracts local impact data of different environments on battery power consumption. On this basis, the feature data is screened by the mutual information maximization criterion to find out the key environmental factors that have the greatest impact on battery consumption, and then sparse Bayesian regression is used to predict and model these key factors and automobile operating parameters, and the sparse processing of the model is realized by automatically adjusting the weights, and finally the optimal dynamic power consumption prediction model is constructed. Another innovation of the invention is that a deep deterministic policy gradient algorithm is introduced, and the power consumption allocation strategy is optimized in real time through a reinforcement learning framework. Combined with the changes in the actual environment, the power consumption management strategy of the battery is dynamically adjusted to maximize the battery utilization efficiency. Compared with traditional battery management technology, this method can better adapt to complex and changeable environmental conditions and has higher prediction accuracy and adaptability.

[0020] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A schematic flow chart of a multi-environment vehicle battery management method according to an embodiment of the present invention;

[0023] Figure 2 It is a schematic diagram of the structure of a multi-environment automobile battery management system described in an embodiment of the present invention.

[0024] In the figure: 701, acquisition unit; 702, processing unit; 703, selection unit; 704, prediction unit; 705, optimization unit; 7021, first processing subunit; 7022, second processing subunit; 7023, third processing subunit; 7031, first selection subunit; 7032, second selection subunit; 70311, third selection subunit; 70312, fourth selection subunit; 70313, fifth selection subunit; 7041, first prediction subunit; 7042, second prediction subunit; 7043, third prediction subunit; 7044, fourth prediction subunit; 7045, fifth prediction subunit; 7051, first optimization subunit; 7052, second optimization subunit; 7053, third optimization subunit. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, 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 part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in 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.

[0026] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0027] Example 1

[0028] This embodiment provides a multi-environment automobile battery management method.

[0029] See also Figure 1 , the figure shows that the method includes step S1, step S2, step S3, step S4 and step S5.

[0030] Step S1, obtaining environmental information and real-time operating parameter information of the vehicle, wherein the environmental information includes climate information, terrain slope and road friction coefficient information, and the real-time operating parameter information of the vehicle includes vehicle load information and battery consumption information;

[0031] It is understandable that the climate information in this step can come from meteorological sensors or third-party meteorological data, the terrain slope can be measured in real time based on the GPS geographic information system (GIS) or the vehicle-mounted tilt sensor, and the road friction coefficient is usually inferred through the vehicle dynamic sensor. The relationship between these external environmental factors and battery consumption is not linear, so it is necessary to consider the time-varying nature of this information and its comprehensive impact on the overall performance of the vehicle.

[0032] At the same time, vehicle load information and battery consumption information are acquired in real time through on-board sensors. This type of data fluctuates frequently and needs to be smoothed or noise filtered to ensure its accuracy. In this step, the focus is on the integrity and timeliness of data acquisition to ensure that this information can provide a basis for subsequent feature extraction and modeling. The technical effect of this step is to provide the necessary data support for multi-dimensional power consumption prediction and optimization, ensuring that the dynamic changes of environmental factors and vehicle operating status can be reflected in the model in real time, laying a solid foundation for subsequent processing steps.

[0033] Step S2, performing multi-scale decomposition processing based on the environmental information and parameter information of the real-time operation of the vehicle, wherein local impact data of all environmental information is obtained by separating and extracting time-frequency domain features;

[0034] It can be understood that this step decomposes the complex multi-dimensional environmental information and vehicle parameters into multiple independent feature signals, so that the subsequent feature extraction and modeling can make more accurate predictions for these local effects. This multi-scale decomposition greatly improves the model's ability to perceive battery consumption in a variable environment by capturing environmental changes at different scales. In this step, step S2 includes step S21, step S22, and step S23.

[0035] Step S21, performing adaptive multi-scale decomposition processing on the environmental information and the parameter information of the real-time operation of the vehicle, wherein the environmental information and the parameter information of the real-time operation of the vehicle are decomposed by an empirical mode decomposition algorithm to obtain at least two intrinsic mode functions;

[0036] It can be understood that this step constructs upper and lower envelopes by screening local extreme values ​​in environmental information and real-time operating parameters, and then calculates the mean of the upper and lower envelopes, subtracts the original signal, and obtains the first candidate intrinsic mode function. The candidate intrinsic mode function is then screened repeatedly until the stop condition defined by the intrinsic mode function is met, the first intrinsic mode function is extracted, and finally the intrinsic mode function is removed from the original signal, and the above steps are repeated for the remaining signal, and subsequent intrinsic mode functions are extracted in turn until the remaining signal cannot be further decomposed.

[0037] This step can decompose the complex, multi-scale environment and operation data into several physically meaningful modal functions, making the subsequent feature analysis and modeling more accurate. Since the empirical mode decomposition algorithm is based on data adaptive decomposition and does not require pre-set assumptions, it can dynamically adjust the decomposition scale for the complex and changeable automobile operating environment to ensure that the decomposition results are adapted to the actual situation.

[0038] Step S22: performing instantaneous characteristic analysis on each intrinsic mode function based on Hilbert-Huang transform, and calculating the instantaneous frequency and instantaneous energy distribution information of each intrinsic mode function;

[0039] It can be understood that this step performs instantaneous feature analysis on each intrinsic mode function based on the Hilbert-Huang transform, wherein the intrinsic mode function is converted into an analytical signal in a complex form after the Hilbert-Huang transform, and the conversion formula of the analytical signal is as follows:

[0040]

[0041] in, Represents the analytical signal, represents the intrinsic mode function, represents the Hilbert change of the intrinsic mode function, Is an imaginary unit.

[0042] Step S23: taking the instantaneous frequency and instantaneous energy distribution information of each intrinsic mode function as the local impact data of all environmental information.

[0043] The calculation formulas of the instantaneous frequency and instantaneous energy are as follows:

[0044]

[0045]

[0046] in, represents the instantaneous frequency of the signal, represents the phase of the analytical signal, represents the original signal, Represents the analytical signal, represents the intrinsic mode function, represents the Hilbert change of the intrinsic mode function, Represents the instantaneous energy of the signal.

[0047] Step S3: performing feature selection on the local impact data of all environmental information according to the mutual information maximization criterion, wherein the key environmental impact factors are determined by calculating the nonlinear dependency between each environmental information and the battery consumption;

[0048] It can be understood that this step successfully filters out the most influential environmental factors by maximizing the criterion, reduces the interference of redundant information, makes the power consumption prediction more accurate, and can adapt to complex, multi-dimensional environmental data scenarios. The advantage of this method is that it can handle nonlinear dependencies. In this step, step S3 includes step S31 and step S32.

[0049] Step S31, performing mutual information calculation on the local impact data of all environmental information and the battery consumption information to obtain a calculation result, wherein the calculation result is the mutual information between the local impact data of each environmental information and the battery consumption information;

[0050] It can be understood that this step can handle the complex nonlinear dependencies between environmental variables and battery consumption through mutual information calculation, significantly improving the accuracy of feature selection. Through this calculation, we can filter out the truly important variables from a large amount of environmental information, thereby providing an efficient and accurate data basis for subsequent power consumption prediction and optimization. This method has advantages over traditional linear correlation analysis, especially when faced with highly nonlinear data, the calculation of mutual information can better capture the deep relationship between variables. In this step, step S31 includes step S311, step S312 and step S313.

[0051] Step S311: discretizing the local impact data of the environmental information and the battery consumption information based on a preset adaptive binning method, wherein the binning intervals of the local impact data of the environmental information and the battery consumption information are automatically selected by using an entropy maximization criterion to obtain discretized data, wherein the discretized data includes the discretized local impact data and the battery consumption information;

[0052] It can be understood that this step not only improves the accuracy of mutual information calculation, but also can flexibly respond to the distribution characteristics of different environmental data, thereby enhancing the applicability and reliability of the model in practical applications. The creativity of this method lies in its ability to automatically adjust the binning rules to adapt to dynamically changing environmental information, ensuring that the model is always analyzed based on the latest and most relevant data. Among them, the formula for the entropy maximization criterion is as follows:

[0053]

[0054] in, is a variable The entropy of Indicates that there is intervals, Representation variables The value is taken in The probability in the interval.

[0055] Step S312: Calculate the probability density of the bin intervals in the discretized data based on the adaptive kernel function to obtain the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval;

[0056] It can be understood that the calculation formula for calculating the probability density of the bin intervals in the discretized data based on the adaptive kernel function in this step is as follows:

[0057]

[0058] in, Represents a random variable The probability density of represents the total number of samples, represents the sample index, represents the total number of samples in the dataset, represents the kernel function, represents the bandwidth parameter, Indicates The value of the sample points, represents a random variable.

[0059] Step S313: Based on a preset mutual information calculation formula, mutual information is calculated on the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval to obtain the local impact data of each environmental information and the mutual information of the battery consumption information.

[0060] It can be understood that through mutual information calculation, the mutual information between the local impact data of each environmental information and the battery consumption information can be obtained. This mutual information value reflects the degree of influence of environmental factors on battery consumption. The higher the value, the stronger the nonlinear dependency between the two variables, which can effectively indicate which environmental factors play an important role in dynamic power consumption prediction. Therefore, this step has an important technical effect on feature selection and model optimization. By extracting these key environmental influencing factors, the accuracy and efficiency of the dynamic power consumption prediction model can be further improved, thereby achieving a more efficient battery management strategy.

[0061] Among them, the preset mutual information calculation formula is as follows:

[0062]

[0063] in, Representation variables and variables The mutual information between Representation variables and variables The joint probability distribution of Representation variables The marginal probability distribution of Representation variables The marginal probability distribution of .

[0064] Step S32: sort the local impact data of each environmental information by the calculation results obtained by mutual information calculation, and retain the first preset number of variables with mutual information values ​​from high to low as key environmental impact factors.

[0065] It can be understood that this step not only achieves effective compression of data dimensions, but also ensures that the selected features contribute significantly to the model's predictive ability. This will lay the foundation for the subsequent construction of dynamic power consumption prediction models, improve the performance and accuracy of the model, and reduce computational complexity and potential overfitting risks.

[0066] Step S4, processing the key environmental influencing factors and the parameter information of the real-time operation of the vehicle according to sparse Bayesian regression, wherein the prediction model of battery consumption is established through the key environmental influencing factors and the parameter information of the real-time operation of the vehicle, and the optimal model is obtained by automatically adjusting the feature weights to obtain a dynamic power consumption prediction model;

[0067] It can be understood that the dynamic power consumption prediction model finally obtained through the processing steps in this step can reflect the changes in battery consumption in real time, improve the intelligence level of battery management, and ensure the efficient operation of the vehicle under different environmental conditions. This process demonstrates the unique advantages of sparse Bayesian regression in feature selection and model optimization, and effectively supports the formulation of battery management strategies. In this step, step S4 includes step S41, step S42, step S43, step S44 and step S45.

[0068] Step S41, constructing a multi-dimensional feature matrix based on preset historical key environmental influencing factors and parameter information of historical vehicle operation;

[0069] It can be understood that each row of the multidimensional feature matrix in this step represents the comprehensive data of a time point, including a combination of historical key environmental influencing factors and historical operating parameter information. For example, for each specific time point, the feature matrix may contain the following columns: temperature, humidity, terrain slope, road friction coefficient, vehicle load, and battery consumption. This step not only enhances the availability of data, but also provides rich contextual information for subsequent model training and prediction, thereby improving the model's responsiveness to changes in battery consumption and prediction accuracy.

[0070] Step S42: performing regression modeling on the multidimensional feature matrix based on a preset Bayesian regression model to construct a regression model of a Bayesian framework;

[0071] It can be understood that this step uses the constructed multi-dimensional feature matrix as input, and the Bayesian regression model will associate these features with the battery consumption information. When faced with high-dimensional data, the sparsity characteristics of the Bayesian regression model can automatically remove redundant features, thereby improving the model's interpretability and predictive performance.

[0072] In the regression modeling process, the Bayesian framework allows joint inference of parameters and hyperparameters in the model, which is crucial for dynamic power consumption prediction. By modeling the uncertainty of the model, it is better able to cope with the noise and variability of the input features. This feature is of great significance in practical applications because the operation and battery consumption patterns of vehicles under different environmental conditions are often highly nonlinear and complex.

[0073] Finally, by modeling the multi-dimensional feature matrix, the Bayesian framework regression model finally constructed can not only effectively reflect the relationship between key environmental factors and battery consumption, but also continuously improve the prediction accuracy through subsequent parameter updates and model evaluation. It provides a reliable statistical basis for dynamic power consumption prediction, can adapt to environmental changes, and thus achieve more accurate battery management strategies to ensure that the vehicle can achieve optimal energy efficiency under various operating conditions.

[0074] Step S43, iteratively updating the weights and hyperparameters of the regression model of the Bayesian framework through the expectation maximization algorithm in the Bayesian regression model to obtain a sparse regression model;

[0075] It is understandable that this step uses the posterior distribution calculated in the previous step to optimize the weights and hyperparameters of the model. In practical applications, the optimization of hyperparameters is particularly important because they can significantly affect the predictive power and complexity of the model. By setting a suitable prior distribution and combining the expectation maximization algorithm, the hyperparameters can be adaptively adjusted to achieve the effect of sparsification, that is, only the features that contribute significantly to the prediction of battery consumption are retained, and redundant or noise features are eliminated.

[0076] Step S44, sending the key environmental influencing factors and the parameter information of the real-time operation of the vehicle to the sparse regression model for battery consumption prediction, wherein the predicted mean and variance are calculated by the Bayesian regression model to generate the battery consumption prediction value and its corresponding uncertainty interval;

[0077] It can be understood that this step uses the posterior distribution obtained during the training process to calculate the mean and variance of the prediction. The mean represents the expected battery consumption under the current environment and operating conditions; while the variance represents the uncertainty of the model about this predicted value, thereby generating an uncertainty interval. This uncertainty interval can help engineers and decision makers understand the possible fluctuation range of battery consumption under different environmental changes in practical applications.

[0078] Step S45 , selecting the optimal hyperparameter combination by cross-validating the battery consumption prediction value and its corresponding uncertainty interval, and then obtaining the optimal dynamic power consumption prediction model.

[0079] It can be understood that this step selects the optimal hyperparameter combination based on cross-validation of the battery consumption prediction value and its uncertainty interval, and can build a more accurate and stable dynamic power consumption prediction model. This model can accurately reflect the impact of real-time environmental changes on battery consumption and provide a more scientific decision-making basis for automotive battery management. This process not only improves the overall performance of the model, but also lays a solid foundation for the formulation of subsequent battery energy management strategies.

[0080] Step S5: determine the dynamic power consumption of the vehicle based on the real-time acquired vehicle operation environment information and the dynamic power consumption prediction model, and perform real-time dynamic power consumption optimization according to the deep deterministic policy gradient algorithm to obtain the optimal power consumption allocation strategy.

[0081] It can be understood that by comprehensively considering real-time environmental information and dynamic power consumption prediction, combined with the advantages of deep deterministic policy gradient algorithm, this step realizes accurate identification and optimal allocation of dynamic power consumption of automobiles. It not only improves the intelligent level of battery management, but also lays a solid foundation for energy efficiency management of future electric vehicles in complex environments. In this step, step S5 includes step S51, step S52 and step S53.

[0082] Step S51, sending the real-time acquired vehicle operation environment information to the dynamic power consumption prediction model for processing to obtain the power consumption information of the vehicle in the current environment;

[0083] It can be understood that this step achieves accurate calculation of the power consumption information of the car in the current environment by passing the real-time environmental information to the dynamic power consumption prediction model. The successful implementation of this process lays the foundation for efficient battery management and optimized power consumption allocation strategy, thereby improving the energy efficiency performance of the vehicle under different driving conditions.

[0084] Step S52: constructing a reinforcement learning framework based on a deep deterministic policy gradient algorithm based on the power consumption information of the car in the current environment and the real-time acquired environment information of the car's operation, wherein the policy network of the reinforcement learning framework outputs a power consumption allocation strategy by inputting the power consumption information in the current environment and the real-time acquired environment information of the car's operation, and evaluates the current strategy through the value network and optimizes the power consumption allocation strategy through the policy gradient update formula;

[0085] It can be understood that this step optimizes the vehicle power consumption allocation strategy by constructing a reinforcement learning framework based on a deep deterministic policy gradient algorithm. By combining real-time environmental information and current power consumption data, the framework can continuously learn and adapt, significantly improving the intelligence level and energy efficiency performance of the battery management system. This method not only improves the efficiency of battery use, but also provides strong support for more efficient vehicle energy consumption management. Among them, the policy gradient update formula is as follows:

[0086]

[0087] in, represents the new policy parameters, Indicates the current policy parameters, represents the learning rate, The objective function representing the performance of the strategy, represents the gradient of the policy parameters.

[0088] Step S53: continuously input the real-time vehicle operation environment information into the reinforcement learning framework for reinforcement learning until the number of times the policy gradient update formula optimizes the power consumption allocation strategy is less than a preset threshold, thereby obtaining the optimal power consumption allocation strategy solution.

[0089] It can be understood that through real-time learning and adjustment, the reinforcement learning framework can quickly adapt to different driving environments, thereby optimizing dynamic power consumption. This method significantly improves the flexibility and reliability of the battery management system, ensuring that the vehicle can achieve the best energy efficiency in various environments.

[0090] Example 2

[0091] like Figure 2 As shown, this embodiment provides a multi-environment vehicle battery management system, see Figure 2 The system includes an acquisition unit 701 , a processing unit 702 , a selection unit 703 , a prediction unit 704 and an optimization unit 705 .

[0092] The acquisition unit 701 is used to acquire environmental information and real-time operating parameter information of the vehicle, wherein the environmental information includes climate information, terrain slope and road friction coefficient information, and the real-time operating parameter information of the vehicle includes vehicle load information and battery consumption information;

[0093] The processing unit 702 is used to perform multi-scale decomposition processing based on the environmental information and the parameter information of the real-time operation of the vehicle, wherein the local impact data of all environmental information is obtained by separating and extracting the time-frequency domain features;

[0094] The processing unit 702 includes a first processing sub-unit 7021 , a second processing sub-unit 7022 and a third processing sub-unit 7023 .

[0095] The first processing subunit 7021 is used to perform adaptive multi-scale decomposition processing on the environmental information and the parameter information of the real-time operation of the vehicle, wherein the environmental information and the parameter information of the real-time operation of the vehicle are decomposed by an empirical mode decomposition algorithm to obtain at least two intrinsic mode functions;

[0096] The second processing subunit 7022 is used to perform instantaneous characteristic analysis on each intrinsic mode function based on Hilbert-Huang transform, and calculate the instantaneous frequency and instantaneous energy distribution information of each intrinsic mode function;

[0097] The third processing subunit 7023 is used to use the instantaneous frequency and instantaneous energy distribution information of each intrinsic mode function as local impact data of all environmental information.

[0098] A selection unit 703 is used to perform feature selection on the local impact data of all environmental information according to a mutual information maximization criterion, wherein the key environmental impact factors are determined by calculating the nonlinear dependency between each environmental information and the battery consumption;

[0099] The selection unit 703 includes a first selection subunit 7031 and a second selection subunit 7032 .

[0100] A first selection subunit 7031 is used to calculate mutual information between the local impact data of all environmental information and the battery consumption information to obtain a calculation result, wherein the calculation result is the mutual information between the local impact data of each environmental information and the battery consumption information;

[0101] Among them, the first selection subunit 7031 includes a third selection subunit 70311, a fourth selection subunit 70312 and a fifth selection subunit 70313.

[0102] The third selection subunit 70311 is used to discretize the local impact data of the environmental information and the battery consumption information based on a preset adaptive binning method, wherein the binning intervals of the local impact data of the environmental information and the battery consumption information are automatically selected by using an entropy maximization criterion to obtain discretized data, wherein the discretized data includes the discretized local impact data and the battery consumption information;

[0103] The fourth selection subunit 70312 is used to calculate the probability density of the bin intervals in the discretized data based on the adaptive kernel function, and obtain the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval;

[0104] The fifth selection subunit 70313 is used to calculate the mutual information of the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval based on a preset mutual information calculation formula to obtain the local impact data of each environmental information and the mutual information of the battery consumption information.

[0105] The second selection subunit 7032 is used to sort the local impact data of each environmental information according to the calculation results obtained by mutual information calculation, and retain the first preset number of variables with mutual information values ​​from high to low as key environmental impact factors.

[0106] The prediction unit 704 is used to process the key environmental influencing factors and the parameter information of the real-time operation of the vehicle according to the sparse Bayesian regression, wherein the prediction model of the battery consumption is established by the key environmental influencing factors and the parameter information of the real-time operation of the vehicle, and the optimal model is obtained by automatically adjusting the feature weights to obtain the dynamic power consumption prediction model;

[0107] The prediction unit 704 includes a first prediction subunit 7041 , a second prediction subunit 7042 , a third prediction subunit 7043 , a fourth prediction subunit 7044 and a fifth prediction subunit 7045 .

[0108] The first prediction subunit 7041 is used to construct a multi-dimensional feature matrix based on preset historical key environmental influencing factors and parameter information of historical operation of the vehicle;

[0109] The second prediction subunit 7042 is used to perform regression modeling on the multi-dimensional feature matrix based on a preset Bayesian regression model to construct a regression model of a Bayesian framework;

[0110] The third prediction subunit 7043 is used to iteratively update the weights and hyperparameters of the regression model of the Bayesian framework through the expectation maximization algorithm in the Bayesian regression model to obtain a sparse regression model;

[0111] The fourth prediction subunit 7044 is used to send the key environmental influencing factors and the parameter information of the real-time operation of the vehicle to the sparse regression model for battery consumption prediction, wherein the predicted mean and variance are calculated by the Bayesian regression model to generate the battery consumption prediction value and its corresponding uncertainty interval;

[0112] The fifth prediction subunit 7045 is used to select the optimal hyperparameter combination by cross-validating the battery consumption prediction value and its corresponding uncertainty interval, so as to obtain the optimal dynamic power consumption prediction model.

[0113] The optimization unit 705 is used to determine the dynamic power consumption of the vehicle based on the real-time acquired vehicle operation environment information and the dynamic power consumption prediction model, and perform real-time dynamic power consumption optimization according to the deep deterministic policy gradient algorithm to obtain the optimal power consumption allocation strategy solution.

[0114] The optimization unit 705 includes a first optimization subunit 7051 , a second optimization subunit 7052 and a third optimization subunit 7053 .

[0115] The first optimization subunit 7051 is used to send the real-time acquired vehicle operation environment information to the dynamic power consumption prediction model for processing to obtain the power consumption information of the vehicle in the current environment;

[0116] The second optimization subunit 7052 is used to build a reinforcement learning framework based on a deep deterministic policy gradient algorithm based on the power consumption information of the automobile in the current environment and the real-time acquired environment information of the automobile operation, wherein the policy network of the reinforcement learning framework outputs a power consumption allocation strategy by inputting the power consumption information in the current environment and the real-time acquired environment information of the automobile operation, and evaluates the current strategy through the value network and optimizes the power consumption allocation strategy through the policy gradient update formula;

[0117] The third optimization subunit 7053 is used to continuously input the real-time environmental information of the vehicle operation into the reinforcement learning framework for reinforcement learning until the number of times the policy gradient update formula optimizes the power consumption allocation strategy is less than a preset threshold, thereby obtaining the optimal power consumption allocation strategy solution.

[0118] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0120] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A multi-environment automobile battery management method, characterized in that: include: Acquiring environmental information and real-time operating parameter information of the vehicle, wherein the environmental information includes climate information, terrain slope and road friction coefficient information, and the real-time operating parameter information of the vehicle includes vehicle load information and battery consumption information; Perform multi-scale decomposition processing based on the environmental information and parameter information of the real-time operation of the vehicle, wherein local impact data of all environmental information is obtained by separating and extracting time-frequency domain features; Perform feature selection on the local impact data of all environmental information according to the mutual information maximization criterion, wherein the key environmental impact factors are determined by calculating the nonlinear dependency between each environmental information and battery consumption; The key environmental influencing factors and the parameter information of the real-time operation of the vehicle are processed according to sparse Bayesian regression, wherein the prediction model of battery consumption is built based on the key environmental influencing factors and the parameter information of the real-time operation of the vehicle, and the optimal model is obtained by automatically adjusting the feature weights to obtain a dynamic power consumption prediction model; Determine the dynamic power consumption of the vehicle based on the real-time environmental information of the vehicle and the dynamic power consumption prediction model, and perform real-time dynamic power consumption optimization based on the deep deterministic policy gradient algorithm to obtain the optimal power consumption allocation strategy. The dynamic power consumption of the vehicle is determined based on the real-time environmental information of the vehicle and the dynamic power consumption prediction model, and the real-time dynamic power consumption optimization is performed according to the deep deterministic policy gradient algorithm to obtain the optimal power consumption allocation strategy, including: The real-time acquired environment information of the vehicle operation is sent to the dynamic power consumption prediction model for processing to obtain the power consumption information of the vehicle in the current environment; Based on the power consumption information of the car in the current environment and the real-time information of the car's operation environment, a reinforcement learning framework based on a deep deterministic policy gradient algorithm is constructed, wherein the policy network of the reinforcement learning framework outputs a power consumption allocation strategy by inputting the power consumption information in the current environment and the real-time information of the car's operation environment, and evaluates the current strategy through the value network and optimizes the power consumption allocation strategy through the policy gradient update formula; The real-time environmental information of the vehicle operation is continuously input into the reinforcement learning framework for reinforcement learning until the number of times the policy gradient update formula optimizes the power consumption allocation strategy is less than a preset threshold, thereby obtaining the optimal power consumption allocation strategy solution.

2. The multi-environment automobile battery management method according to claim 1, characterized in that , based on the environmental information and the parameter information of the real-time operation of the vehicle, multi-scale decomposition processing is performed, including: The environmental information and the parameter information of the real-time operation of the vehicle are subjected to adaptive multi-scale decomposition processing, wherein the environmental information and the parameter information of the real-time operation of the vehicle are decomposed by an empirical mode decomposition algorithm to obtain at least two intrinsic mode functions; Based on the Hilbert-Huang transform, the instantaneous characteristic analysis of each intrinsic mode function is performed, and the instantaneous frequency and instantaneous energy distribution information of each intrinsic mode function are calculated; The instantaneous frequency and instantaneous energy distribution information of each natural mode function are used as the local impact data of all environmental information.

3. The multi-environment automobile battery management method according to claim 1, characterized in that ,According to the mutual information maximization criterion, feature selection is performed on the local impact data of all environmental information, including: Perform mutual information calculation on the local impact data of all environmental information and the battery consumption information to obtain a calculation result, wherein the calculation result is the mutual information between the local impact data of each environmental information and the battery consumption information; The calculation results obtained through mutual information calculation are used to sort the local impact data of each environmental information, and the first preset number of variables with mutual information values ​​from high to low are retained as key environmental influencing factors.

4. The multi-environment automobile battery management method according to claim 3, characterized in that , calculate the mutual information of the local impact data of all environmental information and the battery consumption information, including: The local impact data of the environmental information and the battery consumption information are discretized based on a preset adaptive binning method, wherein the binning intervals of the local impact data of the environmental information and the battery consumption information are automatically selected by using an entropy maximization criterion to obtain discretized data, wherein the discretized data includes the discretized local impact data and the battery consumption information; Based on the adaptive kernel function, the probability density of the bin intervals in the discretized data is calculated respectively, and the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval are obtained; Based on a preset mutual information calculation formula, mutual information is calculated for the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval to obtain the local impact data of each environmental information and the mutual information of the battery consumption information.

5. The multi-environment automobile battery management method according to claim 1, characterized in that ,According to sparse Bayesian regression, the key environmental factors and the parameter information of the real-time operation of the car are processed, including: Construct a multi-dimensional feature matrix based on preset historical key environmental influencing factors and parameter information of historical vehicle operation; Based on a preset Bayesian regression model, regression modeling is performed on the multidimensional feature matrix to construct a regression model of a Bayesian framework; The weights and hyperparameters of the regression model in the Bayesian framework are iteratively updated through the expectation maximization algorithm in the Bayesian regression model to obtain a sparse regression model. Send the key environmental factors and the real-time operating parameter information of the vehicle to the sparse regression model for battery consumption prediction, wherein the predicted mean and variance are calculated by the Bayesian regression model to generate the battery consumption prediction value and its corresponding uncertainty interval; The optimal hyperparameter combination is selected by cross-validating the battery consumption prediction value and its corresponding uncertainty interval, and then the optimal dynamic power consumption prediction model is obtained.

6. A multi-environment automobile battery management system, characterized in that: include: An acquisition unit, used to acquire environmental information and real-time operating parameter information of the vehicle, wherein the environmental information includes climate information, terrain slope and road friction coefficient information, and the real-time operating parameter information of the vehicle includes vehicle load information and battery consumption information; A processing unit, configured to perform multi-scale decomposition processing based on the environmental information and parameter information of the real-time operation of the vehicle, wherein local impact data of all environmental information is obtained by separating and extracting time-frequency domain features; A selection unit is used to perform feature selection on the local impact data of all environmental information according to a mutual information maximization criterion, wherein the key environmental impact factors are determined by calculating the nonlinear dependency between each environmental information and the battery consumption; A prediction unit, used for processing key environmental influencing factors and parameter information of real-time operation of the vehicle according to sparse Bayesian regression, wherein a prediction model of battery consumption is established through the key environmental influencing factors and parameter information of real-time operation of the vehicle, and an optimal model is obtained by automatically adjusting feature weights to obtain a dynamic power consumption prediction model; An optimization unit is used to determine the dynamic power consumption of the vehicle based on the real-time acquired vehicle operation environment information and the dynamic power consumption prediction model, and to perform real-time dynamic power consumption optimization according to a deep deterministic policy gradient algorithm to obtain an optimal power consumption allocation strategy; Among them, the optimization unit includes: A first optimization subunit is used to send the real-time acquired vehicle operation environment information to the dynamic power consumption prediction model for processing to obtain the power consumption information of the vehicle in the current environment; A second optimization subunit is used to construct a reinforcement learning framework based on a deep deterministic policy gradient algorithm based on the power consumption information of the car in the current environment and the real-time acquired environment information of the car's operation, wherein the policy network of the reinforcement learning framework outputs a power consumption allocation strategy by inputting the power consumption information in the current environment and the real-time acquired environment information of the car's operation, and evaluates the current strategy through the value network and optimizes the power consumption allocation strategy through the policy gradient update formula; The third optimization subunit is used to continuously input the real-time environmental information of the vehicle operation into the reinforcement learning framework for reinforcement learning until the number of times the policy gradient update formula optimizes the power consumption allocation strategy is less than a preset threshold, thereby obtaining the optimal power consumption allocation strategy solution.

7. The multi-environment vehicle battery management system according to claim 6, characterized in that: The processing unit comprises: A first processing subunit is used to perform adaptive multi-scale decomposition processing on the environmental information and the parameter information of the real-time operation of the vehicle, wherein the environmental information and the parameter information of the real-time operation of the vehicle are decomposed by an empirical mode decomposition algorithm to obtain at least two intrinsic mode functions; The second processing subunit is used to perform instantaneous characteristic analysis on each intrinsic mode function based on Hilbert-Huang transform, and calculate the instantaneous frequency and instantaneous energy distribution information of each intrinsic mode function; The third processing subunit is used to use the instantaneous frequency and instantaneous energy distribution information of each inherent mode function as local impact data of all environmental information.

8. The multi-environment vehicle battery management system according to claim 6, characterized in that: The selection unit comprises: A first selection subunit, configured to perform mutual information calculation on the local impact data of all environmental information and the battery consumption information to obtain a calculation result, wherein the calculation result is the mutual information between the local impact data of each environmental information and the battery consumption information; The second selection subunit is used to sort the local impact data of each environmental information according to the calculation results obtained by mutual information calculation, and retain the first preset number of variables with mutual information values ​​from high to low as key environmental impact factors.

9. The multi-environment vehicle battery management system according to claim 8, characterized in that: The first selection subunit comprises: a third selection subunit, configured to discretize the local impact data of the environmental information and the battery consumption information based on a preset adaptive binning method, wherein the binning intervals of the local impact data of the environmental information and the battery consumption information are automatically selected by using an entropy maximization criterion to obtain discretized data, wherein the discretized data includes the discretized local impact data and the battery consumption information; The fourth selection subunit is used to calculate the probability density of the bin intervals in the discretized data based on the adaptive kernel function, and obtain the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval; The fifth selection subunit is used to calculate the mutual information of the joint probability distribution and marginal probability distribution of the discretized data in each discrete interval based on a preset mutual information calculation formula to obtain the local impact data of each environmental information and the mutual information of the battery consumption information.

10. The multi-environment vehicle battery management system according to claim 6, characterized in that: The prediction unit comprises: A first prediction subunit is used to construct a multi-dimensional feature matrix based on preset historical key environmental influencing factors and parameter information of historical operation of the vehicle; A second prediction subunit is used to perform regression modeling on the multidimensional feature matrix based on a preset Bayesian regression model to construct a regression model of a Bayesian framework; The third prediction subunit is used to iteratively update the weights and hyperparameters of the regression model of the Bayesian framework through the expectation maximization algorithm in the Bayesian regression model to obtain a sparse regression model; The fourth prediction subunit is used to send key environmental influencing factors and parameter information of the real-time operation of the vehicle to the sparse regression model for battery consumption prediction, wherein the predicted mean and variance are calculated by the Bayesian regression model to generate the battery consumption prediction value and its corresponding uncertainty interval; the fifth prediction subunit is used to select the optimal hyperparameter combination by cross-validation of the battery consumption prediction value and its corresponding uncertainty interval, and then obtain the optimal dynamic power consumption prediction model.

Images