A power supply control system for electric vehicles

Abstract

The present invention discloses a power supply management and control system for electric vehicles, specifically in the field of power supply operation risk identification, and is used to solve the problem of being unable to fully capture potential risk characteristic parameters. First, by performing eigenvalue decomposition on the correlation coefficient matrix, the number of retained factors is determined to evaluate the degree of relationship between observed variables and factors, and a factor loading matrix is ​​constructed. Subsequently, the maximum information coefficient and Gaussian kernel function are comprehensively used to construct a correlation coefficient matrix, and the key factors affecting power supply safety are identified. Using the LASSO regression algorithm and cross-validation method, the contribution of each observed variable to the factor is evaluated, and the key factors in power supply operation are identified. Finally, the data spectrum features are extracted by discrete Fourier transform, and standardized processing is performed to identify and warn of possible abnormal behavior or instability in the power supply system, providing an effective means for the timely discovery and handling of abnormal situations.

Description

Technical Field

[0001] The present invention relates to the field of power supply operation risk identification, and more specifically, to a power supply management and control system for electric vehicles. Background Art

[0002] The power source of an electric vehicle refers to the energy source that provides power to drive the electric vehicle, usually a battery pack. The power management of an electric vehicle refers to the process of effectively controlling and monitoring the power system of an electric vehicle, aiming to ensure the safety, performance stability and life of the battery, as well as optimize the energy efficiency of the vehicle. This involves the management and optimization of multiple aspects such as battery charge and discharge management, temperature control, voltage monitoring, and current control to improve the performance, endurance and safety of electric vehicles. The power management system may also include intelligent charging systems, energy recovery systems, and vehicle dynamic management systems to achieve comprehensive monitoring and optimization of the electric vehicle power system.

[0003] At present, there are some shortcomings in the existing electric vehicle power management. First, the existing methods are insufficient in identifying risk characteristic parameters and may not be able to fully capture all potential risk characteristic parameters, resulting in incomplete risk identification. In addition, it is difficult for existing technologies to fully understand the real risks of power management because the complex relationship between various potential factors and various abnormal situations that may occur in the actual operation of the electric vehicle power system may not be fully considered when identifying and assessing risks. Therefore, the existing methods may have limitations and cannot effectively identify and respond to potential risks in power management, which may affect the safety and performance stability of electric vehicles.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a power supply management and control system for electric vehicles. First, by performing eigenvalue decomposition on the correlation coefficient matrix, the number of retained factors is determined to evaluate the degree of relationship between the observed variables and the factors, and a factor loading matrix is ​​constructed. Subsequently, the maximum information coefficient and Gaussian kernel function are comprehensively used to construct the correlation coefficient matrix and identify the key factors affecting the safety of the power supply. Using the LASSO regression algorithm and cross-validation method, the contribution of each observed variable to the factor is evaluated, the key factors in the power supply operation are identified, and more accurate and comprehensive information is provided to support the risk management of the power supply system at various operation stages to ensure its safe and stable operation. Finally, the data spectrum features are extracted by discrete Fourier transform and standardized, and the abnormal behavior or instability that may exist in the power supply system is identified and warned, which provides an effective means for the timely discovery and handling of abnormal situations to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: comprising a perception storage module, a feature extraction module, a factor identification module, a factor rotation module, a contribution evaluation module and an anomaly detection module;

[0007] The perception storage module uses sensors to monitor various aspects of power supply data, records and stores them locally and in the cloud, and sends the data call interface to the feature extraction module and contribution evaluation module;

[0008] The feature extraction module retrieves data from the cloud server, maps variables to high-dimensional space using the Gaussian kernel function, calculates the maximum information coefficient, constructs a correlation coefficient matrix after normalization, and sends the correlation coefficient matrix to the factor identification module;

[0009] The factor identification module determines the number of retained factors through eigenvalue decomposition and constructs a factor loading matrix to identify the key factors affecting power supply security, and sends the identification results to the factor rotation module;

[0010] The factor rotation module uses the gradient descent optimization algorithm to perform maximum variance rotation on the factor loading matrix, optimizes the rotation parameters to minimize the difference between the rotated factor loading matrix and the original matrix, obtains the factor loading matrix, and sends the factor loading matrix to the contribution evaluation module;

[0011] Contribution evaluation module This module uses LASSO regression algorithm and cross-validation method to evaluate the contribution of each observed variable to the factor, identify the key factors in power supply operation, and send the key factors to the anomaly detection module;

[0012] Abnormality Detection Module 6 This module extracts the main frequency components through discrete Fourier transform, standardizes and evaluates the abnormality degree, identifies the abnormality of the power system and generates corresponding signals.

[0013] In a preferred embodiment, the feature extraction module includes the following contents:

[0014] Step 1, retrieve various observed variable data during the battery operation process stored in the cloud server;

[0015] Step 2: For each pair of variables, use the Gaussian kernel function to map the original variables into a high-dimensional feature space;

[0016] Step 3: Calculate the maximum information coefficient between variables in the high-dimensional feature space;

[0017] Step 4: normalize the calculated mutual information value so that it falls within the range of [0,1];

[0018] Step 5: Construct the normalized mutual information values ​​into a correlation coefficient matrix. The rows and columns of the matrix correspond to the indexes of the variables, and each element in the matrix represents the degree of correlation between the corresponding variables.

[0019] In a preferred embodiment, the factor identification module includes the following contents:

[0020] Step 1, perform eigenvalue decomposition on the correlation coefficient matrix to obtain the eigenvalues ​​and corresponding eigenvectors of the matrix;

[0021] Step 3: Calculate the eigenvalue of each factor and arrange them in order from largest to smallest.

[0022] Draw a graph of the relationship between the eigenvalue and the number of factors, where the horizontal axis is the number of factors and the vertical axis is the eigenvalue;

[0023] Find the number of factors corresponding to the inflection point;

[0024] The factors before the inflection point are taken as the number of factors to be retained;

[0025] Step 4: Perform eigenvalue decomposition on the correlation coefficient matrix to obtain eigenvalues ​​λ1,λ2,...,λ m and the corresponding eigenvectors v1,v2,…,v m , where m is the number of factors retained;

[0026] Select the m eigenvectors with the largest eigenvalues ​​as the column vectors of the factor loading matrix;

[0027] Arrange the selected eigenvectors into a p×m matrix L, i.e., the factor loading matrix;

[0028] The formula for constructing the factor loading matrix is ​​as follows: L = [v1, v2, ..., v m ];

[0029] Among them, v1, v2, …, v m They represent the m eigenvectors with the largest eigenvalues, · represents the matrix arranged in columns. Each column of the factor loading matrix L corresponds to a factor, and each row corresponds to an observed variable. The elements in the matrix represent the degree of relationship between the observed variable and the factor, that is, the loading value.

[0030] In a preferred embodiment, the factor rotation module includes the following contents:

[0031] Step 1: Use the factor loading matrix as the initial loading matrix;

[0032] Step 2: In the iterative process, the rotation parameters are optimized by minimizing the difference between the rotated factor loading matrix and the original factor loading matrix, and the rotation parameters are continuously updated until the convergence condition is reached;

[0033] Step three: Use the optimal rotation parameters to rotate the initial factor loading matrix to obtain the rotated factor loading matrix.

[0034] In a preferred embodiment, the contribution assessment module includes the following contents:

[0035] The observed variable data is represented as a matrix U, where each row represents a sample, each column represents an observed variable, and the factor score is used as the target variable y;

[0036] Construct a regression model, the goal is to minimize the sum of the loss function and the penalty term:

[0037]

[0038] Among them, β0 is the intercept, β is the coefficient vector, ||·||2 represents the L2 norm, ||·||1 represents the L1 norm, and ε is the regularization parameter used to control sparsity;

[0039] Use geometric progression to select a set of candidate regularization parameters ε;

[0040] The dataset is divided into K mutually exclusive subsets, of which K-1 subsets are used to train the model and the remaining subset is used to verify the model performance. This process is repeated K times until each subset is used as a validation set once.

[0041] For each regularization parameter ε, the regression model is entropy fitted on the training set of each cross-validation, and the model performance is evaluated on the corresponding validation set;

[0042] According to the performance evaluation index of the model on the validation set, select the regularization parameter ε that can achieve the best model performance;

[0043] Use the selected optimal parameter ε to fit the LASSO regression model and obtain the final coefficient vector β;

[0044] For each observed variable u i , and its corresponding coefficient β i The absolute value of indicates its importance in the regression model, that is, its contribution to the factor, and it is marked as contribution degree.

[0045] In a preferred embodiment, the anomaly detection module includes the following contents:

[0046] Collect data on the required backup features in the form of time series;

[0047] Performing discrete Fourier transform on the collected data of the backup features;

[0048] The spectral power density spectrum is obtained by normalizing the square of the modulus value of the discrete Fourier transform result;

[0049] According to the spectrum power density spectrum, the frequency component ranked first is selected as the main frequency component;

[0050] The amplitude of the main frequency component was used as the regularity evaluation value;

[0051] The regularity evaluation values ​​calculated in different unit times are combined into a regularity evaluation data set, and each data point represents a regularity evaluation value in a time period;

[0052] The regularity evaluation data set is standardized and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1. The latest data point is compared with the mean of the entire data set, and then the difference is divided by the standard deviation of the entire data set to obtain a standardized score. The standardized score is compared with the warning threshold. If the standardized score is greater than or equal to the warning threshold, an active power-off signal is generated.

[0053] The technical effects and advantages of the power supply control system for electric vehicles of the present invention are as follows:

[0054] 1. The present invention comprehensively considers and analyzes the linear and nonlinear relationships between variables in the battery operation process by combining the Gaussian kernel function and the maximum information coefficient method, so as to more comprehensively and accurately evaluate the correlation between them. The Gaussian kernel function maps the original variables to a high-dimensional feature space, so that nonlinear relationships can be considered; while the maximum information coefficient calculates the correlation between variables in this space, taking into account nonlinear relationships, making full use of their respective advantages, and solving the problem of linear and nonlinear variable correlation analysis at the same time, providing a reliable basis for further analysis and modeling of battery operation data, helping to discover hidden laws, and providing more reliable decision support.

[0055] 2. The present invention first performs eigenvalue decomposition on the correlation coefficient matrix to obtain eigenvalues ​​and corresponding eigenvectors, and then calculates the eigenvalues ​​of each factor, and determines the number of factors to be retained by drawing a relationship diagram between the eigenvalue and the number of factors, and finding the number of factors corresponding to the inflection point. Subsequently, the maximum information coefficient and Gaussian kernel function are used to comprehensively consider and analyze the linear and nonlinear relationships between the variables during the battery operation process to construct a correlation coefficient matrix. The matrix represents the degree of correlation between the variables and can be used to identify the key factors affecting the safety of the power supply. The correlation coefficient matrix is ​​decomposed by eigenvalues, and the eigenvector with the largest eigenvalue is selected to construct a factor loading matrix to evaluate the degree of relationship between the observed variables and the factors. The LASSO regression algorithm and cross-validation method are used to evaluate the contribution of each observed variable to the factor, and identify the key factors in the power supply operation. These key factors can provide more accurate and comprehensive information, which helps to better understand the state and performance characteristics of the power supply operation. It can then provide important support for the risk management and control of the power supply system at each operation stage to ensure its safe and stable operation.

[0056] 3. The present invention utilizes discrete Fourier transform to extract data spectrum features, and compares the degree of deviation of the data with the warning threshold through standardization processing, so as to identify and warn of possible abnormal behavior or instability of the power supply system. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The present invention is a schematic structural diagram of a power supply control system for an electric vehicle. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example 1

[0060] Figure 1 The present invention provides a power supply management and control system for electric vehicles, comprising: a perception storage module, a feature extraction module, a factor identification module, a factor rotation module, a contribution evaluation module and an anomaly detection module;

[0061] The perception storage module uses sensors to monitor various aspects of power supply data, records and stores them locally and in the cloud, and sends the data call interface to the feature extraction module and contribution evaluation module;

[0062] The feature extraction module retrieves data from the cloud server, maps variables to high-dimensional space using the Gaussian kernel function, calculates the maximum information coefficient, constructs a correlation coefficient matrix after normalization, and sends the correlation coefficient matrix to the factor identification module;

[0063] The factor identification module determines the number of retained factors through eigenvalue decomposition and constructs a factor loading matrix to identify the key factors affecting power supply security, and sends the identification results to the factor rotation module;

[0064] The factor rotation module uses the gradient descent optimization algorithm to perform maximum variance rotation on the factor loading matrix, optimizes the rotation parameters to minimize the difference between the rotated factor loading matrix and the original matrix, obtains the factor loading matrix, and sends the factor loading matrix to the contribution evaluation module;

[0065] Contribution evaluation module This module uses LASSO regression algorithm and cross-validation method to evaluate the contribution of each observed variable to the factor, identify the key factors in power supply operation, and send the key factors to the anomaly detection module;

[0066] Abnormality Detection Module 6 This module extracts the main frequency components through discrete Fourier transform, standardizes and evaluates the abnormality degree, identifies the abnormality of the power system and generates corresponding signals.

[0067] The perception storage module includes the following:

[0068] Step 1: List the observed variables that need to be monitored. These variables cover all aspects of power supply operation, including voltage, current, temperature, capacity, charge and discharge rate, etc.

[0069] Step 2: Collect data based on sensors pre-installed in various parts of the power supply. For example, voltage and current can be monitored using voltage sensors and current sensors, temperature can be monitored using temperature sensors, and capacity can be calculated by current integration.

[0070] Step 3: Collect data and monitor various observed variables in the battery operation process in real time through sensors. You can choose different data collection frequencies, such as collecting data once per second, once per minute, or once per hour, and determine the collection frequency according to your needs;

[0071] Step 4: Record the collected data, mark and classify the data for subsequent analysis and processing. Ensure the accuracy and completeness of the data records, including timestamps, variable values ​​and other information;

[0072] Step 5: Store the collected data locally and in cloud service media. The local media is used to store temporary data, and the cloud service is used to store permanent data.

[0073] In actual data analysis, such as various variables in battery operation, the relationships between variables are often complex and diverse. Some variables have linear relationships, such as a direct linear dependency between voltage and current. However, other variables may have nonlinear relationships, such as a nonlinear relationship between battery temperature and capacity that changes with temperature. Therefore, only considering linear or nonlinear relationships often cannot fully understand the inherent structure of the data. In order to fully understand the relevance of the data and accurately evaluate the correlation between variables, it is necessary to comprehensively consider and analyze linear and nonlinear relationships. By combining the Gaussian kernel function and the maximum information coefficient method, linear and nonlinear relationships can be captured simultaneously, thereby more comprehensively and accurately evaluating the correlation between variables. This comprehensive analysis method provides a reliable basis for further data analysis and modeling, helps to discover the laws and trends hidden behind the data, and provides more reliable support for decision-making.

[0074] The feature extraction module includes the following:

[0075] Step 1, retrieve various observed variable data during the battery operation process stored in the cloud server;

[0076] Step 2: For each pair of variables, use the Gaussian kernel function to map the original variables into a high-dimensional feature space;

[0077] The Gaussian kernel function is defined as follows:

[0078] Among them, x and y are the original variables, σ is the bandwidth parameter of the Gaussian kernel function, which controls the distribution of the data after being mapped to the high-dimensional space;

[0079] Step 3: Calculate the maximum information coefficient between variables in the high-dimensional feature space;

[0080] The calculation formula of the maximum information coefficient is as follows:

[0081] Among them, X and Y are two original variables, n is the feature of spatial dimension, and MIC i (X,Y) is the maximum information coefficient calculated on the i-th feature dimension.

[0082] Step 4: normalize the calculated mutual information value so that it falls within the range of [0,1];

[0083] The maximum information coefficient is used to evaluate the correlation between two variables. During the calculation process, the maximum information coefficient will find the optimal feature subset that can maximize the mutual information between the variables, thereby quantifying the degree of correlation between the variables. Among them, "maximum" means that in the process of searching for feature subsets, the subset that can maximize the mutual information is selected, and "minimum" means that the minimum mutual information value is selected from all possible feature subsets as the final evaluation result.

[0084] Step 5: Construct the normalized mutual information values ​​into a correlation coefficient matrix. The rows and columns of the matrix correspond to the indexes of the variables, and each element in the matrix represents the degree of correlation between the corresponding variables.

[0085] The significance of combining the Gaussian kernel function and the maximum information coefficient to construct the correlation coefficient matrix is ​​that it can make full use of their respective advantages and solve the correlation analysis problem between linear and nonlinear variables at the same time. As a nonlinear kernel function, the Gaussian kernel function can map the original variables to a high-dimensional feature space, so that linearly inseparable data becomes linearly separable in the high-dimensional feature space; while the maximum information coefficient can calculate the correlation between variables in this high-dimensional feature space, taking into account the nonlinear relationship. Therefore, the method combining the Gaussian kernel function and the maximum information coefficient can not only evaluate the correlation between variables more comprehensively and accurately, but also solve the correlation analysis problem between linear variables and nonlinear variables at the same time, and has high applicability and flexibility.

[0086] The present invention combines the Gaussian kernel function and the maximum information coefficient method to comprehensively consider and analyze the linear and nonlinear relationships between the variables in the battery operation process, so as to more comprehensively and accurately evaluate the correlation between them. The Gaussian kernel function maps the original variables to a high-dimensional feature space, so that nonlinear relationships can be considered; while the maximum information coefficient calculates the correlation between variables in this space, taking into account nonlinear relationships, making full use of their respective advantages, and solving the problem of linear and nonlinear variable correlation analysis at the same time, providing a reliable basis for further analysis and modeling of battery operation data, helping to discover hidden laws, and providing more reliable decision support.

[0087] The factor identification module includes the following:

[0088] Step 1, perform eigenvalue decomposition on the correlation coefficient matrix to obtain the eigenvalues ​​and corresponding eigenvectors of the matrix;

[0089] The eigenvalue decomposition formula of the correlation coefficient matrix is: R = QΛQ T ;

[0090] Where Q is the eigenvector matrix, Λ is a diagonal matrix, and the elements on the diagonal are the eigenvalues;

[0091] Step 3: Calculate the eigenvalue of each factor and arrange them in order from largest to smallest.

[0092] Draw a graph of the relationship between the eigenvalue and the number of factors, where the horizontal axis is the number of factors and the vertical axis is the eigenvalue;

[0093] Find the number of factors corresponding to the inflection point;

[0094] The factors before the inflection point are taken as the number of factors to be retained.

[0095] Step 4: Perform eigenvalue decomposition on the correlation coefficient matrix to obtain eigenvalues ​​λ1,λ2,...,λ m and the corresponding eigenvectors v1,v2,…,v m , where m is the number of factors retained;

[0096] Select the m eigenvectors with the largest eigenvalues ​​as the column vectors of the factor loading matrix;

[0097] Arrange the selected eigenvectors into a p×m matrix L, namely the factor loading matrix.

[0098] The formula for constructing the factor loading matrix is ​​as follows: L = [v1, v2, ..., v m ];

[0099] Among them, v1,v2,...,v m Respectively represent the m eigenvectors with the largest eigenvalues, and · represents the arrangement of columns into a matrix. Each column of the factor loading matrix L corresponds to a factor, and each row corresponds to an observed variable. The elements in the matrix represent the degree of relationship between the observed variable and the factor, that is, the loading value.

[0100] The factor loading value is used to indicate the degree of relationship between the observed variable and the factor. The larger the absolute value of the loading value, the higher the degree of association between the observed variable and the factor. A positive loading value indicates that there is a positive correlation between the observed variable and the factor, that is, when the factor score increases, the score of the observed variable will also increase; a negative loading value indicates that there is a negative correlation between the observed variable and the factor, that is, when the factor score increases, the score of the observed variable will decrease. Therefore, the size and positive and negative direction of the loading value both provide information about the relationship between the observed variable and the factor, which can help explain the meaning and impact of the factor.

[0101] In the risk management and control of power supply, factor analysis can help identify the key factors that affect the stability and safety of power supply operation. By observing the factor loading matrix, it is possible to determine which observed variables correspond to key factors and evaluate their relationship with the risk management objectives. For example, if a factor is closely related to the risk of power supply operation and its loading value is large, it indicates that the factor may have an important impact on the safety of the power supply. Therefore, in the risk management and control strategy, the observed variables related to these key factors can be monitored and managed to reduce potential risks. Therefore, the results of factor analysis are closely related to the risk management and control of power supply, which helps to identify key risk factors and formulate effective control measures.

[0102] The factor rotation module includes the following:

[0103] Step 1: Use the factor loading matrix as the initial loading matrix;

[0104] Step 2: In the iterative process, the rotation parameters are optimized by minimizing the difference between the rotated factor loading matrix and the original factor loading matrix, and the rotation parameters are continuously updated until the convergence condition is reached;

[0105] For example: Assuming there is an initial factor loading matrix L0, a new factor loading matrix L1 is found through rotation to make it more in line with the requirements.

[0106] First, an initial rotation matrix R0 is selected, which can be a unit matrix or a randomly generated initial rotation matrix. Then, the initial rotation matrix is ​​applied to the initial factor loading matrix to obtain the rotated factor loading matrix L1=L0R0.

[0107] Next, the difference between the rotated factor loading matrix L1 and the original factor loading matrix L0 is calculated. Some loss function can be used to measure this difference, such as squared difference loss. Assume that the loss function is J(R), where R represents the rotation matrix.

[0108] Then, an optimization algorithm such as gradient descent is used to minimize the loss function J(R) by adjusting the rotation matrix R. The rotation parameters are continuously updated until convergence conditions are met, such as the change in the loss function is less than a certain threshold or the preset number of iterations is reached.

[0109] The specific iteration process is as follows:

[0110] 1. Initialize the rotation matrix R0;

[0111] 2. Calculate the rotated factor loading matrix L1=L0R0;

[0112] 3. Calculate the loss function J(R) to measure the difference between L1 and L0;

[0113] 4. Use the optimization algorithm to adjust the rotation matrix R to minimize the loss function J(R);

[0114] 5. Check the convergence condition. If it is satisfied, stop the iteration. Otherwise, return to step 2 and continue to iterate and update the rotation parameters.

[0115] Step three: Use the optimal rotation parameters to rotate the initial factor loading matrix to obtain the rotated factor loading matrix.

[0116] The factor loading matrix after rotation is L (k) It can be calculated by the following formula: (k) =L (k-1) R (k) ;

[0117] Among them, L (k-1) is the factor loading matrix of the previous iteration, R (k) Is the rotation matrix, which represents the rotation parameters of this iteration.

[0118] The rotation parameters are solved by minimizing a certain loss function between the rotated factor loading matrix and the original factor loading matrix based on the gradient descent optimization algorithm. A commonly used loss function is to minimize the following objective function:

[0119] in, is the element of the factor loading matrix in the kth iteration, is the element of the initial factor loading matrix, i represents the observed variable, and j represents the factor. By continuously iteratively optimizing the rotation parameters, the optimal rotated factor loading matrix can be obtained.

[0120] The factor loading matrix is ​​used to represent the degree of relationship between the observed variables and the factors. Each element represents the loading value of an observed variable corresponding to a factor, that is, the projection or contribution of the observed variable on the factor. The larger the absolute value of the loading value, the closer the relationship between the observed variable and the factor, that is, the stronger the explanatory power of the observed variable for the factor. The positive and negative signs indicate the positive or negative correlation between the observed variable and the factor. A positive loading value indicates a positive correlation, and a negative loading value indicates a negative correlation. Therefore, a larger loading value and a smaller negative loading value represent a stronger positive correlation or negative correlation between the observed variable and the factor, respectively.

[0121] Factors in factor analysis represent latent concepts or variables that cannot be directly observed. They are constructed through the common variability between observed variables. Factor analysis aims to transform the original observed variable data into fewer and more explanatory factors by finding common patterns or structures between observed variables. These factors can be regarded as a potential representation of the observed variables, where each factor captures the common variability of a set of related observed variables. Therefore, factors are closely related to the observed variable data collected at the beginning. They are extracted from the original data, and through the interpretation of the factor loading matrix, the relationship between each factor and the original observed variables can be understood, so as to better understand the structure and characteristics of the data.

[0122] The contribution assessment module includes the following:

[0123] The observed variable data is represented as a matrix U, where each row represents a sample, each column represents an observed variable, and the factor score is used as the target variable y;

[0124] Construct a regression model, the goal is to minimize the sum of the loss function and the penalty term:

[0125]

[0126] Among them, β0 is the intercept, β is the coefficient vector, ||·||2 represents the L2 norm, ||·||1 represents the L1 norm, and ε is the regularization parameter used to control sparsity.

[0127] Use geometric progression to select a set of candidate regularization parameters ε;

[0128] The dataset is divided into K mutually exclusive subsets (usually K is 5 or 10), of which K-1 subsets are used to train the model and the remaining subset is used to verify the model performance. This process is repeated K times until each subset is used as a validation set once.

[0129] For each regularization parameter ε, the regression model is entropy-fitted on the training set of each cross-validation, and the model performance is evaluated on the corresponding validation set. Common performance evaluation indicators include mean square error (MSE), root mean square error (RMSE), R-squared, etc.

[0130] According to the performance evaluation index of the model on the validation set, select the regularization parameter ε that can achieve the best model performance;

[0131] Use the selected optimal parameter ε to fit the LASSO regression model and obtain the final coefficient vector β.

[0132] Through the above steps, the cross-validation method can be used to select the optimal regularization parameter ε to optimize the performance of the regression model, thereby more accurately evaluating the contribution of each observed variable to the factor.

[0133] The following is an example to illustrate the process:

[0134] First, a set of candidate values ​​for the regularization parameter ε is selected, such as {0.001, 0.01, 0.1, 1, 10}.

[0135] For each ε value, cross validation is performed. Assume that 5-fold cross validation is selected, that is, the data set is divided into 5 subsets.

[0136] On each cross-validation training set, the regression model is fitted using the selected ε value;

[0137] Use the fitted model to make predictions on the corresponding validation set and calculate the square error between the predicted value and the actual observed value;

[0138] For each ε value, the above process is repeated 5 times (i.e., each subset serves as a validation set once), and 5 MSE values ​​are obtained;

[0139] For each ε value, the average of 5 MSE values ​​is calculated as the performance evaluation indicator of the model under this parameter;

[0140] Based on the calculated MSE value, select the regularization parameter ε that can achieve the best model performance. Usually, select the ε that minimizes the MSE value.

[0141] For each observed variable u i , and its corresponding coefficient β i The absolute value of indicates its importance in the regression model, that is, its contribution to the factor, and it is marked as contribution degree.

[0142] Compare the contribution value with the contribution threshold. If the contribution value is greater than or equal to the contribution threshold, it is marked as a spare feature, indicating that the observed variable has a high importance and contribution in the factor analysis. Specifically, this means that the relationship between the observed variable and the factor is significant, and the explanatory power of the factor is strong, so it is of great significance when monitoring the power supply. Observed variables with high contribution can provide more accurate and comprehensive information, which helps to better understand the status and performance characteristics of power supply operation. These observed variables can more accurately reflect the key factors and correlations in the power supply operation process, including battery health status, charge and discharge efficiency, temperature changes, etc. Therefore, monitoring these observed variables with high contribution can timely discover abnormal conditions and potential problems in power supply operation, help risk assessment and management, and thus ensure the safe and stable operation of the power supply system.

[0143] The present invention first performs eigenvalue decomposition on the correlation coefficient matrix to obtain eigenvalues ​​and corresponding eigenvectors, and then calculates the eigenvalues ​​of each factor, and by drawing a relationship diagram between the eigenvalue and the number of factors, and finding the number of factors corresponding to the inflection point, the number of factors to be retained is determined. Subsequently, the maximum information coefficient and Gaussian kernel function are used to comprehensively consider and analyze the linear and nonlinear relationships between the variables during the battery operation process to construct a correlation coefficient matrix. The matrix represents the degree of correlation between the variables and can be used to identify the key factors affecting the safety of the power supply. By performing eigenvalue decomposition on the correlation coefficient matrix and selecting the eigenvector with the largest eigenvalue to construct a factor loading matrix, the degree of relationship between the observed variables and the factors is evaluated. By using the LASSO regression algorithm and the cross-validation method, the contribution of each observed variable to the factor is evaluated, and the key factors in the power supply operation are identified. These key factors can provide more accurate and comprehensive information, which helps to better understand the state and performance characteristics of the power supply operation. It can then provide important support for the risk management and control of the power supply system at each operation stage to ensure its safe and stable operation.

[0144] The anomaly detection module includes the following:

[0145] Collect data on the required backup features in the form of time series;

[0146] The collected data of the backup features are subjected to discrete Fourier transform. The discrete Fourier transform can be expressed by the following formula:

[0147] Where X(k) is a complex value in the frequency domain, x(m) is an observation in the time domain, M is the number of observations, and k is the frequency index.

[0148] By normalizing the square of the modulus of the discrete Fourier transform result X(k), we get the spectral power density spectrum. The spectral power density represents the energy distribution of the observed value at different frequencies.

[0149] According to the spectral power density spectrum, the frequency components ranked first are selected as the main frequency components. These frequency components may correspond to periodic changes or regularities in the observed values.

[0150] The amplitude of the main frequency component is used as the regularity evaluation value.

[0151] The regularity evaluation values ​​calculated in different unit times are combined into a regularity evaluation data set, and each data point represents a regularity evaluation value in a time period.

[0152] The regularity evaluation data set is standardized and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1. The latest data point is compared with the mean of the entire data set, and then the difference is divided by the standard deviation of the entire data set to obtain a standardized score. This score represents the position of the data point in the entire data set. If this score is far away from the mean, it means that the data point is very different from the entire data set. Therefore, the greater the deviation of the data point, the larger the standardized score obtained, indicating a higher degree of abnormality, which also implies abnormal behavior or instability of the power system in certain time periods. This may be caused by factors such as battery performance degradation, abnormal charging process, and uneven discharge, which need to be taken seriously to ensure the normal operation and safety of the power system.

[0153] Compare the standardized score with the warning threshold. If the standardized score is greater than or equal to the warning threshold, it means that the data point is very different from the entire data set, the degree of deviation is high, and there is abnormal behavior or instability in the power system during this time period. This situation needs to be taken seriously and may require further monitoring and analysis to eliminate potential problems, ensure the normal operation and safety of the power system, generate active power-off signals, and actively shut down the power supply to prevent hidden dangers from continuing to occur and further deterioration. Conversely, if the standardized score is lower than the warning threshold, it means that the data point is less different from the entire data set, the degree of deviation is low, and the power system may be in normal operation during this time period and no signal is generated.

[0154] The present invention utilizes discrete Fourier transform to extract data spectrum features, and compares the degree of deviation of the data with a warning threshold through standardization processing, so as to identify and warn of possible abnormal behavior or instability of the power supply system.

[0155] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0156] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0157] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0158] In the several embodiments provided in the present application, it should be understood that the disclosed systems and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0159] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0160] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0161] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0162] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0163] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power supply control system for an electric vehicle, characterized in that: It includes a perception storage module, a feature extraction module, a factor identification module, a factor rotation module, a contribution evaluation module and an anomaly detection module; The perception storage module uses sensors to monitor various aspects of power supply data, records and stores them locally and in the cloud, and sends the data call interface to the feature extraction module and contribution evaluation module; The feature extraction module retrieves data from the cloud server, maps variables to high-dimensional space using the Gaussian kernel function, calculates the maximum information coefficient, constructs a correlation coefficient matrix after normalization, and sends the correlation coefficient matrix to the factor identification module; The factor identification module determines the number of retained factors through eigenvalue decomposition and constructs a factor loading matrix to identify the key factors affecting power supply security, and sends the identification results to the factor rotation module; The factor rotation module uses the gradient descent optimization algorithm to perform maximum variance rotation on the factor loading matrix, optimizes the rotation parameters to minimize the difference between the rotated factor loading matrix and the original matrix, obtains the factor loading matrix, and sends the factor loading matrix to the contribution evaluation module; Contribution evaluation module This module uses LASSO regression algorithm and cross-validation method to evaluate the contribution of each observed variable to the factor, identify the key factors in power supply operation, and send the key factors to the anomaly detection module; Abnormality detection module This module extracts the main frequency components through discrete Fourier transform, standardizes and evaluates the abnormality degree, identifies the abnormality of the power system and generates corresponding signals; The anomaly detection module includes the following: Collect data on the required backup features in the form of time series; Performing discrete Fourier transform on the collected data of the backup features; The spectral power density spectrum is obtained by normalizing the square of the modulus value of the discrete Fourier transform result; According to the spectrum power density spectrum, the frequency component ranked first is selected as the main frequency component; The amplitude of the main frequency component is used as the regularity evaluation value; The regularity evaluation values ​​calculated in different unit times are combined into a regularity evaluation data set, and each data point represents a regularity evaluation value in a time period; The regularity evaluation data set is standardized and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1. The latest data point is compared with the mean of the entire data set, and then the difference is divided by the standard deviation of the entire data set to obtain a standardized score. The standardized score is compared with the warning threshold. If the standardized score is greater than or equal to the warning threshold, an active power-off signal is generated.

2. A power management and control system for electric vehicles according to claim 1, characterized in that: The feature extraction module includes the following: Step 1, retrieve various observed variable data during the battery operation process stored in the cloud server; Step 2: For each pair of variables, use the Gaussian kernel function to map the original variables into a high-dimensional feature space; Step 3: Calculate the maximum information coefficient between variables in the high-dimensional feature space; Step 4: Normalize the calculated mutual information value so that it falls within the range of [0,1]; Step 5: Construct the normalized mutual information values ​​into a correlation coefficient matrix. The rows and columns of the matrix correspond to the indexes of the variables, and each element in the matrix represents the degree of correlation between the corresponding variables.

3. A power management and control system for electric vehicles according to claim 2, characterized in that: The factor identification module includes the following: Step 1: Calculate the eigenvalue of each factor and arrange them in order from largest to smallest; Draw a graph of the relationship between the eigenvalue and the number of factors, where the horizontal axis is the number of factors and the vertical axis is the eigenvalue; Find the number of factors corresponding to the inflection point; The factors before the inflection point are taken as the number of factors to be retained; Step 2: Perform eigenvalue decomposition on the correlation coefficient matrix to obtain eigenvalues ​​λ1,λ2,…,λ m and the corresponding eigenvectors v1,v2,…,v m , where m is the number of factors retained; Select the m eigenvectors with the largest eigenvalues ​​as the column vectors of the factor loading matrix; Arrange the selected eigenvectors into a p×m matrix L, i.e., the factor loading matrix; The formula for constructing the factor loading matrix is ​​as follows: L = [v1, v2, …, v m ]; Among them, v1, v2, …, v m They represent the m eigenvectors with the largest eigenvalues, · represents the matrix arranged in columns. Each column of the factor loading matrix L corresponds to a factor, and each row corresponds to an observed variable. The elements in the matrix represent the degree of relationship between the observed variable and the factor, that is, the loading value.

4. A power management and control system for electric vehicles according to claim 3, characterized in that: The factor rotation module includes the following: Step 1: Use the factor loading matrix as the initial loading matrix; Step 2: In the iterative process, the rotation parameters are optimized by minimizing the difference between the rotated factor loading matrix and the original factor loading matrix, and the rotation parameters are continuously updated until the convergence condition is reached; Step three: Use the optimal rotation parameters to rotate the initial factor loading matrix to obtain the rotated factor loading matrix.

5. A power management and control system for electric vehicles according to claim 4, characterized in that: The contribution assessment module includes the following: The observed variable data is represented as a matrix U, where each row represents a sample, each column represents an observed variable, and the factor score is used as the target variable y; Construct a regression model with the goal of minimizing the sum of the loss function and the penalty term: Among them, β0 is the intercept, β is the coefficient vector, ||·||2 represents the L2 norm, ||·||1 represents the L1 norm, and ε is the regularization parameter used to control sparsity; Use geometric progression to select a set of candidate regularization parameters ε; The dataset is divided into K mutually exclusive subsets, of which K-1 subsets are used to train the model and the remaining subset is used to verify the model performance. This process is repeated K times until each subset is used as a validation set once. For each regularization parameter ε, the regression model is entropy fitted on the training set of each cross-validation, and the model performance is evaluated on the corresponding validation set; According to the performance evaluation index of the model on the validation set, select the regularization parameter ε that can achieve the best model performance; Use the selected optimal parameter ε to fit the LASSO regression model and obtain the final coefficient vector β; For each observed variable u i , and its corresponding coefficient β i The absolute value of indicates its importance in the regression model, that is, its contribution to the factor, and it is marked as contribution degree.

Images