New Energy Battery Current Monitoring and Abnormal Detection Method Based on Adaptive Control

Abstract

The present invention relates to the technical field of battery monitoring, and specifically to a new energy battery current monitoring and anomaly detection method based on adaptive control, comprising the following steps: S1. Obtain multi-dimensional vehicle state data and determine the driving state of the vehicle. The multi-dimensional vehicle state data includes, but is not limited to, battery discharge state, driving mode, and driving environment data; S2. Collect the vehicle driving state and battery current output data at a fixed frequency; S3. Perform filtering preprocessing on the collected original current data to remove noise and interference signals in the data. By monitoring the change value of the battery current in real time and accurately, and obtaining multi-source data during the vehicle driving process, combined with the micro-change of the battery current, analyze the reasons for the upcoming battery failure, capture the early signs of the battery failure in advance, and strive sufficient time for the driver to adjust the vehicle position, effectively avoiding emergencies caused by battery failures.

Description

Technical Field

[0001] The present invention relates to the technical field of battery monitoring, and specifically to a new energy battery current monitoring and anomaly detection method based on adaptive control. Background Technique

[0002] With the wide application of new energy vehicles, the battery, as a core component, its performance and safety are directly related to the operation status of the vehicle and the safety of the driver and passengers. During traditional battery anomaly detection, the battery has already failed, and when an anomaly is detected, the vehicle has already become inoperable. This results in the driver not being able to reserve enough time to adjust the vehicle's position before the battery fails, which may lead to serious consequences such as the vehicle suddenly breaking down during driving, bringing great inconvenience to the driver and passengers, and even endangering their lives.

[0003] In addition, the existing monitoring methods are not perfect in verifying and evaluating the prediction results. The multiple prediction results lack systematic analysis and verification, making it difficult to ensure the accuracy and reliability of the prediction. This makes it difficult for the driver to judge the credibility of the warning information when facing it and unable to make a scientific and reasonable decision. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides a new energy battery current monitoring and anomaly detection method based on adaptive control to solve the problems raised in the above background technique.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A new energy battery current monitoring and anomaly detection method based on adaptive control, including the following steps:

[0006] S1. Obtain multi-dimensional vehicle state data and determine the driving state of the vehicle. The multi-dimensional vehicle state data includes but is not limited to the battery discharge state, driving mode, and driving environment data;

[0007] S2. Collect the vehicle driving state and battery current output data at a fixed frequency;

[0008] S3. Perform filter preprocessing on the collected original current data to remove noise and interference signals in the data;

[0009] S4. Predict the battery discharge current state of the vehicle through model deduction based on the preprocessed data;

[0010] S5. According to the anomaly determination result, adaptively adjust the result weight to improve the accuracy of anomaly detection and the ability of early warning, and timely discover potential battery failure hazards;

[0011] S6. After adaptively adjusting the result weights, evaluate the abnormal results, calculate the deviation degree, further evaluate the abnormal results after weight adjustment, and divide the risk levels according to the evaluation results;

[0012] S7. Generate corresponding emergency measures based on the risk levels and abnormal results, and control the discharge current of the vehicle battery;

[0013] S8. Generate warning information including the abnormal results, risk levels, and emergency measures, and send it to the vehicle central control, user mobile APP, vehicle back-end management system, and blockchain to remind the driver to operate or initiate intelligent driving to take over the driver's operation. After the vehicle stops at a safe location, cut off the discharge of the vehicle battery.

[0014] Preferably, in step S3, the calculation formula for filtering the original current data is:

[0015] ;

[0016] In the formula, is the filtered current value, is the filter window size, is the current data at a certain moment , is from to the respective original current values at the moment.

[0017] Preferably, in step S4, predicting the state of the vehicle battery discharge current includes the following steps:

[0018] S41. According to the obtained multi-dimensional data, construct a battery current prediction model, and the calculation formula is:

[0019] ;

[0020] In the formula, is the output vector of the hidden layer, is the data matrix, is the weight matrix, is the bias vector, is the activation function;

[0021] S42. According to the prediction model of the battery current signal, adaptively and dynamically adjust the parameters, predict the state of the battery current, quantify the uncertainty of the prediction state, reflect the observation error, and correct the prediction result to extract the characteristics of the real current signal;

[0022] S43. Dynamically adjust the model parameters according to the real-time input data to perform high-precision prediction on the change of the vehicle battery current, and the calculation formula is:

[0023] ;

[0024] Wherein, is the model parameter value at the th iteration, is the model parameter value at the th iteration, is the learning rate, is the number of iterations, is the loss function at gradient;

[0025] S44. Combine the multi-dimensional data of the vehicle battery, evaluate the prediction results, and analyze the accuracy of the model prediction under different conditions;

[0026] S45. Classify the abnormal results according to the high-precision prediction and analysis results of the vehicle battery current change.

[0027] Preferably, in the step S42, the battery current state prediction is based on the dynamic model of the system to make a preliminary estimate of the future state of the battery current, providing a basis for subsequent precise adjustment. The calculation formula is:

[0028] ;

[0029] Wherein, is the prior prediction value of the system state based on the information at time step at moment, is the transition matrix, is the posterior prediction value of the system state based on its own information at time step at moment, is the control input matrix, is the control input amount applied to the system at time step ;

[0030] Quantifying the uncertainty of the predicted state considers the unpredictable noise factors in the system and provides important parameters for subsequent calculations. The calculation formula is:

[0031] ;

[0032] Wherein, is the posterior covariance matrix based on the information at time step at moment, is the posterior covariance matrix after fusing the observation information at time step at moment, is the matrix transpose, is the process noise covariance matrix;

[0033] reflects the error of the observation and reflects the noise level in the measurement process. The calculation formula is:

[0034] ;

[0035] In the formula, is the Kalman gain, is the observation noise covariance, is the observation matrix;

[0036] is used to correct the prediction result to combine the actual measurement data with the prediction result, making the estimated value closer to the true battery current state. The calculation formula is:

[0037] ;

[0038] In the formula, is the observation value at time step ;

[0039] Extract the characteristics of the true current signal by adjusting the covariance, providing a reliable data basis for subsequent battery current monitoring and anomaly detection tasks. The calculation formula is:

[0040] ;

[0041] In the formula, is the identity matrix.

[0042] Preferably, the evaluation of the battery current prediction result in step S44 includes the following steps:

[0043] S441. Obtain data on current amplitude, change rate, current spectrum, time-frequency domain, and correlations with voltage, temperature, driving, and operation of in-vehicle electrical equipment;

[0044] S442. Calculate the current change rate at present. The calculation formula is:

[0045] ;

[0046] In the formula, is the change rate of current with time, is the current value at time is the current value at time is the sampling time interval;

[0047] S443. Calculate the current spectrum characteristics at present. The calculation formula is:

[0048] ;

[0049] Wherein, is the Fourier transform result of the signal , that is, a function of the frequency ; is the current signal varying with time ; is the imaginary unit is the time calculus;

[0050] S444. Calculate the time-frequency domain characteristics of the current, and the calculation formula is:

[0051] ;

[0052] Wherein, is the short-time Fourier transform result of the signal at time and frequency ; is the original signal varying with the time variable ; is the window function used to intercept the local segment of the signal ; is the time variable calculus;

[0053] S445. Calculate the correlation characteristics between the current and the voltage, and the calculation formula is:

[0054] ;

[0055] Wherein, is the Pearson correlation coefficient between the current and the voltage ; is the th current sampling value is the mean value of the current sampling is the th voltage sampling value is the mean value of the voltage sampling is the number of samplings;

[0056] S446. Adopt a feature selection method based on the quantum genetic algorithm, and use the superposition state of quantum bits and quantum gate operations to quickly search for the optimal feature combination.

[0057] Preferably, in the step S5, the adaptive adjustment of the result weight includes the following steps:

[0058] S51. Collect parameters of the abnormal data of the determination result and the normal data of the battery current;

[0059] S52. Determine the reliability of the abnormal data of the determination result. The calculation formula is:

[0060] ;

[0061] In the formula, is the abnormal data of the determination result, is the abnormal data of the determination result The reliability of is Under the condition of, the probability of observing the data is is the data set The th sample value, and and are respectively the abnormal data of the determination result The reliability is When, the data The mean, variance and standard deviation of the normal distribution;

[0062] S53. Measure the correlation between the normal data of the battery current and the abnormal data of the determination result. The calculation formula is:

[0063] ;

[0064] In the formula, is the normal data of the battery current, is the normal data of the battery current And the abnormal data of the determination result The correlation between them is is the normal data of the battery current And the abnormal data of the determination result The joint probability distribution of is and Are respectively the normal data of the battery current And the abnormal data of the determination result The marginal probability distribution of;

[0065] S54. Train the neural network using the backpropagation algorithm. The loss function is defined as the prediction weight. The calculation formula is:

[0066] ;

[0067] In the formula, is the th monitoring data, is the abnormal data of the determination result And the th The correlation between monitoring data is the reliability of the th data source;

[0068] The mean square error between the calculated theoretical weights:

[0069] ;

[0070] In the formula, is the calculation result of the mean square error between theoretical weights, is the weight predicted by the neural network, is the weight calculated theoretically;

[0071] S55. Use the fused data for battery current anomaly detection, calculate the accuracy of the anomaly detection result, and verify the error. The calculation formula is:

[0072] ;

[0073] In the formula, is the number of actual abnormal samples, the number of abnormal samples correctly detected, is the accuracy;

[0074] S56. When the error verification flag is in the error state, re-analyze the data features and reliability evaluation results, and adjust the parameters of the weight calculation model according to the re-analysis results.

[0075] Preferably, in step S6, evaluating the abnormal result and dividing the risk level includes the following steps:

[0076] S61. Obtain the abnormal classification result and the parameters for adaptive adjustment of the result weight;

[0077] S62. According to the multi-source data features extracted previously, for the feature vector of the currently detected abnormal data, calculate its Euclidean distance from each feature template. The calculation formula is:

[0078] ;

[0079] In the formula, is the number of features, is the feature vector is the th eigenvalue of the quantity related to the th identification object corresponding to the th quantity related to the

[0080] S63. Further determine whether it is abnormal data by calculating the deviation degree between the current data and the normal data distribution. The calculation formula is:

[0081] ;

[0082] In the formula, is the value of the deviation degree, is the battery current-related data after fusion, is the mean value, is the standard deviation;

[0083] S64. Assign weights to each influencing factor and evaluate the severity of the abnormality. The calculation formula is:

[0084] ;

[0085] In the formula, is the evaluation value of the severity of the abnormality, with a value range of [0 - 1], is the total number of factors affecting the severity of the abnormality, is the weight of the th influencing factor, is the quantization value of the

[0086] S65. Combine the type of abnormality and the severity of the abnormality to divide the risk level.

[0087] Preferably, in step S63, when the value is less than the set threshold, it indicates that the abnormal classification result is accurate, and then proceed to step S64. When is greater than the set threshold, it indicates that the abnormal classification is inaccurate, and then return to step S3.

[0088] Preferably, in step S65, after dividing the risk level, establish a risk matrix. The risk level is set with three levels: high risk, medium risk, and low risk. When the risk level is divided into high risk. When the risk level is divided into medium risk. the risk level is divided into low risk.

[0089] Preferably, in step S8, when sending the abnormal result, risk level, and emergency measures to generate a warning message to the blockchain, each block contains the hash value of the previous block, timestamp, and data content information, and through the distributed ledger and encryption algorithm of the blockchain, the data is made immutable and traceable. When verifying the authenticity and integrity of the data, compare the calculated hash value with the hash value stored on the blockchain to ensure the security of the data throughout its life cycle.

[0090] The present invention provides a new energy battery current monitoring and abnormality detection method based on adaptive control. It has the following beneficial effects:

[0091] 1. The present invention monitors the change value of battery current in real time and accurately, obtains multi-source data during vehicle driving, analyzes the cause of impending battery failure in combination with slight changes in battery current, generates risk levels and emergency measures based on the cause of impending failure, fuses multi-source data using dynamic weight adjustment, comprehensively considers battery management system data, vehicle driving status data, etc., can more comprehensively and accurately reflect the battery operating status, capture early signs of battery failure in advance, buy sufficient time for the driver to adjust the vehicle position, and effectively avoid emergencies caused by battery failure.

[0092] 2. The present invention establishes a complete verification and evaluation system through verification and evaluation of multiple prediction results, strictly analyzes and compares each prediction result, continuously optimizes the prediction model, improves the accuracy and reliability of the prediction, and continuously reversely optimizes the parameters and structure of the prediction model through multiple rounds and multi-dimensional verification and evaluation. When it is found that the prediction error is large in a certain specific case, the weight of the corresponding feature in the model is adjusted in a targeted manner, or new feature variables are added, so that the model can more accurately capture the complex relationship between battery current changes and faults. After multiple rounds of iterative optimization, the accuracy and reliability of the prediction model are greatly improved, and the false alarm rate and missed alarm rate are significantly reduced. This enables the driver to be convinced of the authenticity and reliability of the warning information when receiving it, so as to make timely and correct decisions to ensure driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0094] Figure 2 It is a schematic diagram of the model deduction prediction process of the present invention;

[0095] Figure 3 It is a schematic diagram of the weight adaptive adjustment process of the present invention;

[0096] Figure 4 It is a schematic diagram of the abnormal result evaluation process of the present invention; DETAILED DESCRIPTION

[0097] 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.

[0098] likeFigures 1 - 4 As shown in Figures 1 - 4 , the embodiment of the present invention provides a new energy battery current monitoring and anomaly detection method based on adaptive control, including the following steps:

[0099] S1. Obtain multi-dimensional vehicle state data, determine the driving state of the vehicle. The multi-dimensional vehicle state data includes but is not limited to the battery discharge state, driving mode, and driving environment data;

[0100] S2. Collect the vehicle driving state and battery current output data at a fixed frequency;

[0101] S3. Perform filtering preprocessing on the collected original current data to remove noise and interference signals in the data;

[0102] S4. Predict the battery discharge current state of the vehicle through model deduction based on the preprocessed data;

[0103] S5. According to the anomaly determination result, adaptively adjust the result weight to improve the accuracy of anomaly detection and the ability of early warning, and timely discover potential battery fault hazards;

[0104] S6. After adaptively adjusting the result weight, evaluate the anomaly result, calculate the deviation degree, further evaluate the anomaly result after weight adjustment, and divide the risk level according to the evaluation result;

[0105] S7. Generate corresponding emergency measures based on the risk level and anomaly result, and control the discharge current of the vehicle battery;

[0106] S8. Generate a warning message including the anomaly result, risk level, and emergency measures and send it to the vehicle central control, user mobile phone APP, vehicle background management system, and blockchain, reminding the driver to operate or initiate intelligent driving to take over the driver's operation. After the vehicle stops at a safe location, cut off the battery discharge of the vehicle.

[0107] In this embodiment, in step S3, the calculation formula for filtering the original current data is:

[0108] ;

[0109] In the formula, is the filtered current value, is the filter window size, is the current data at a certain moment , is from to the respective original current values at the moment.

[0110] In this embodiment, in step S4, predicting the battery discharge current state of the vehicle includes the following steps:

[0111] S41. Construct a battery current prediction model based on the obtained multi-dimensional data. The calculation formula is:

[0112] ;

[0113] In the formula, is the output vector of the hidden layer, is the data matrix, is the weight matrix, is the bias vector, is the activation function;

[0114] S42. According to the prediction model of the battery current signal, adaptively and dynamically adjust the parameters to predict the battery current state, quantify the uncertainty of the prediction state, reflect the observation error, and correct the prediction result to extract the characteristics of the true current signal;

[0115] S43. Dynamically adjust the model parameters according to the real-time input data to perform high-precision prediction on the change of the vehicle battery current. The calculation formula is:

[0116] ;

[0117] In the formula, is the value of the model parameter at the th iteration, is the value of the model parameter at the th iteration, is the learning rate, is the number of iterations, is the loss function at ;

[0118] S44. Combine the obtained multi-dimensional data of the vehicle battery to evaluate the prediction result and analyze the accuracy of the model prediction in different situations;

[0119] S45. Classify the abnormal results according to the high-precision prediction and analysis results of the change of the vehicle battery current.

[0120] In this embodiment, in step S42, the prediction of the battery current state is based on the dynamic model of the system to make a preliminary estimate of the future state of the battery current, providing a basis for subsequent precise adjustment. The calculation formula is:

[0121] ;

[0122] In the formula, is the prior prediction value of the system state based on the information at time step at moment, is the transition matrix, at time step is the posterior prediction value of the system state based on its own information at the moment of ; is the control input matrix, at time step is the control input quantity applied to the system;

[0123] Quantifying the uncertainty of the predicted state takes into account the unpredictable noise factors in the system and provides important parameters for subsequent calculations. The calculation formula is:

[0124] ;

[0125] In the formula, at time step is the posterior covariance matrix based on the information at the moment of ; at time step is the posterior covariance matrix after fusing the observation information at the moment of ; is the matrix transpose, is the process noise covariance matrix;

[0126] reflects the error of the observation and reflects the noise level in the measurement process. The calculation formula is:

[0127] ;

[0128] In the formula, is the Kalman gain, is the observation noise covariance, is the observation matrix;

[0129] Correcting the prediction result is used to combine the actual measurement data with the prediction result to make the estimated value closer to the true battery current state. The calculation formula is:

[0130] ;

[0131] In the formula, is the observed value at time step ;

[0132] Extracting the characteristics of the true current signal adopts the method of adjusting the covariance, which provides a reliable data basis for subsequent battery current monitoring and anomaly detection tasks. The calculation formula is:

[0133] ;

[0134] In the formula, is the identity matrix.

[0135] In this embodiment, the evaluation of the battery current prediction result in step S44 includes the following steps:

[0136] S441. Obtain data on current amplitude, change rate, current spectrum, time-frequency domain, and correlations with voltage, temperature, driving, and operation of in-vehicle electrical equipment;

[0137] S442. Calculate the current change rate of the current battery, and the calculation formula is:

[0138] ;

[0139] In the formula, is the change rate of the current with time, is the current value at time is the current value at time is the sampling time interval;

[0140] S443. Calculate the current spectrum characteristics of the current, and the calculation formula is:

[0141] ;

[0142] In the formula, is the Fourier transform result of the signal i.e., a function of frequency , is the current signal varying with time , is the imaginary unit, is time the calculus of;

[0143] S444. Calculate the time-frequency domain characteristics of the current, and the calculation formula is:

[0144] ;

[0145] In the formula, is the short-time Fourier transform result of the signal at time and frequency , is the original signal varying with the time variable , is the window function used to intercept the local segment of the signal , is the time variable calculus;

[0146] S445. Calculate the correlation characteristics of the current and voltage, and the calculation formula is:

[0147] ;

[0148] Wherein, is the Pearson correlation coefficient between the current and the voltage , is the th current sampling value, is the mean value of the current sampling, is the th voltage sampling value, is the mean value of the voltage sampling, is the number of samplings;

[0149] S446. Adopt a feature selection method based on the quantum genetic algorithm, and use the superposition state of qubits and quantum gate operations to quickly search for the optimal feature combination.

[0150] Specifically, by monitoring the change value of the battery current in real time and accurately, obtaining multi-source data during the vehicle driving process, combining with the micro-change of the battery current, analyzing the reasons for the upcoming battery failure, generating a risk level and emergency measures according to the reasons for the upcoming failure, and using dynamic weight adjustment to fuse the multi-source data, comprehensively considering the battery management system data, vehicle driving state data, etc., can more comprehensively and accurately reflect the battery operation state, capture the early signs of battery failure in advance, and buy sufficient time for the driver to adjust the vehicle position, effectively avoiding emergencies caused by battery failure.

[0151] In this embodiment, in step S5, the adaptive adjustment of the result weight includes the following steps:

[0152] S51. Collect the parameters of the abnormal data of the judgment result and the normal data of the battery current;

[0153] S52. Judge the reliability of the abnormal data of the judgment result, and the calculation formula is:

[0154] ;

[0155] Wherein, is the abnormal data of the judgment result, is the abnormal data of the judgment result The reliability of is Under the condition of, the probability of observing the data , is the data set The th sample value of, and and are respectively the abnormal data of the judgment result When the reliability is the mean, variance, and standard deviation of the data follow a normal distribution;

[0156] S53. Measure the correlation between the normal data of the battery current and the abnormal data of the determination result. The calculation formula is:

[0157] ;

[0158] In the formula, is the normal data of the battery current, is the normal data of the battery current and the abnormal data of the determination result The correlation between them, is the normal data of the battery current and the abnormal data of the determination result The joint probability distribution of, and are the marginal probability distributions of the normal data of the battery current and the abnormal data of the determination result respectively;

[0159] S54. Train the neural network using the backpropagation algorithm. The loss function is defined as the predicted weight. The calculation formula is:

[0160] ;

[0161] In the formula, is the th monitoring data, is the correlation between the abnormal data of the determination result and the th monitoring data, is the reliability of the th data source, is the number of output nodes;

[0162] The mean square error between the theoretically calculated weights:

[0163] ;

[0164] In the formula, is the calculation result of the mean square error between the theoretically calculated weights, is the weight predicted by the neural network, is the weight calculated theoretically;

[0165] S55. Use the fused data for battery current anomaly detection, calculate the accuracy of the anomaly detection result, and verify the error. The calculation formula is:

[0166] ;

[0167] Wherein, is the actual number of abnormal samples, is the number of abnormal samples correctly detected, is the accuracy rate;

[0168] S56. When the error verification flag is in the error state, re-analyze the data features and reliability evaluation results, and adjust the parameters of the weight calculation model according to the re-analysis results.

[0169] In this embodiment, in step S6, evaluating the abnormal result and dividing the risk level includes the following steps:

[0170] S61. Obtain the abnormal classification result and the parameter for adaptively adjusting the result weight;

[0171] S62. According to the multi-source data features extracted previously, for the feature vector of the currently detected abnormal data, calculate its Euclidean distance from each feature template. The calculation formula is:

[0172] ;

[0173] Wherein, is the number of features, is the feature vector is the eigenvalue of the th identification-related quantity, is the th identification object corresponding to the th identification-related quantity;

[0174] S63. Further determine whether it is abnormal data by calculating the deviation degree of the current data from the normal data distribution. The calculation formula is:

[0175] ;

[0176] Wherein, is the value of the deviation degree, is the battery current-related data after fusion, is the mean value, is the standard deviation;

[0177] S64. Assign weights to each influencing factor and evaluate the severity of the abnormality. The calculation formula is:

[0178] ;

[0179] Wherein, is the evaluation value of the severity of the abnormality, with a value range of [0 - 1], is the total number of factors that affect the severity of the abnormality, For the The weight of the impact factor, For the The quantitative value of the impact factor;

[0180] S65. Risk levels are determined based on the type and severity of the anomaly.

[0181] In this embodiment, in step S63, when If the value is less than the set threshold, it means that the abnormal classification result is accurate, then continue to step S64. If it is greater than the set threshold, it means that the abnormal classification is inaccurate, and then return to step S3.

[0182] In this embodiment, in step S65, after dividing the risk levels, a risk matrix is ​​established, and the risk levels are set to three levels: high risk, medium risk, and low risk. When the risk level is classified as high risk, When the risk level is classified as medium risk, The risk level is classified as low risk.

[0183] In this embodiment, in step S8, when sending abnormal results, risk levels, and emergency measures to generate warning information to the blockchain, each block contains the hash value, timestamp, and data content information of the previous block, and through the distributed ledger and encryption algorithm of the blockchain, the data is tamper-proof and traceable. When it is necessary to verify the authenticity and integrity of the data, the hash value is calculated and compared with the hash value stored on the blockchain to ensure the security of the data throughout its life cycle.

[0184] Specifically, through verification and evaluation of multiple prediction results, a complete verification and evaluation system is established, each prediction result is strictly analyzed and compared, the prediction model is continuously optimized, and the accuracy and reliability of the prediction are improved. Through multiple rounds and multi-dimensional verification and evaluation, the parameters and structure of the prediction model are continuously reversely optimized. When it is found that the prediction error is large in a certain situation, the weights of the corresponding features in the model are adjusted in a targeted manner, or new feature variables are added, so that the model can more accurately capture the complex relationship between battery current changes and faults. After multiple rounds of iterative optimization, the accuracy and reliability of the prediction model are greatly improved, and the false alarm rate and missed alarm rate are significantly reduced. This enables the driver to be convinced of the authenticity and reliability of the warning information when receiving it, so as to make timely and correct decisions to ensure driving safety.

[0185] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A new energy battery current monitoring and abnormality detection method based on adaptive control, characterized in that: The following steps are involved: S1. Acquire multi-dimensional data of vehicle status to determine the driving status of the vehicle, wherein the multi-dimensional data of vehicle status includes but is not limited to battery discharge status, driving mode, and driving environment data; S2, collecting vehicle driving status and battery current output data at a fixed frequency; S3, performing filtering preprocessing on the collected raw current data to remove noise and interference signals in the data; S4. Based on the preprocessed data, the vehicle battery discharge current state is predicted through model deduction to obtain an abnormal result; S5. Adaptively adjust the result weights based on the abnormal results to improve the accuracy of abnormal detection and early warning capabilities, and timely discover potential battery failure hazards; In step S5, the adaptive adjustment of the result weight includes the following steps: S51, collecting parameters of abnormal determination result data and normal battery current data; S52, determine the reliability of the abnormal data of the determination result, and the calculation formula is: ； In the formula, To determine the data with abnormal results, To determine abnormal data The reliability of Under the condition of The probability of For the dataset No. Sample values, and as well as The abnormal data of the judgment results are Reliability When, data The mean, variance and standard deviation follow a normal distribution; S53, measuring the correlation between the normal battery current data and the abnormal determination result data, the calculation formula is: ； In the formula, is the normal data of battery current, Normal data for battery current and abnormal data of the judgment results The correlation between Normal data for battery current and abnormal data of the judgment results The joint probability distribution of and They are respectively the normal data of battery current and abnormal data of the judgment results The marginal probability distribution of S54. The back propagation algorithm is used to train the neural network. The loss function is defined as the prediction weight, and the calculation formula is: ； In the formula, For the Monitoring data, To determine abnormal data With The correlation between the monitoring data For the The reliability of the data source, is the number of output nodes; The mean square error between the calculated theoretical weights is: ； In the formula, is the mean square error calculation result between theoretical weights, is the weight predicted by the neural network, is the weight calculated theoretically; S55. Use the fused data for battery current anomaly detection, calculate the accuracy of the anomaly detection result, and verify the error. The calculation formula is: ； In the formula, is the actual number of abnormal samples, The number of abnormal samples correctly detected, is the accuracy rate; S56. When the error verification mark is in an error state, reanalyze the data characteristics and reliability assessment results, and adjust the parameters of the weight calculation model according to the reanalysis results; S6. After the result weights are adaptively adjusted, the abnormal results are evaluated, the deviation is calculated, the abnormal results after weight adjustment are further evaluated, and the risk level is divided according to the evaluation results; S7. Generate corresponding emergency measures according to the risk level and abnormal results to control the discharge current of the vehicle battery; S8. Generate warning information about abnormal results, risk levels, and emergency measures and send it to the vehicle's central control, user mobile phone APP, vehicle backend management system, and blockchain to remind the driver to operate or start intelligent driving to take over the driver's operation until the vehicle reaches a safe location and stops, cutting off the vehicle battery discharge.

2. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 1 is characterized in that: In step S3, the calculation formula for filtering the original current data is: ； In the formula, is the current value after filtering, is the filter window size, for a certain moment Current data, For arrive The original current values ​​at each moment.

3. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 1 is characterized in that: In step S4, predicting the vehicle battery discharge current state includes the following steps: S41. Based on the acquired multi-dimensional data, a battery current prediction model is constructed, and the calculation formula is: ； In the formula, is the output vector of the hidden layer, is the data matrix, is the weight matrix, is the bias vector, is the activation function; S42, according to the prediction model of the battery current signal, adaptively and dynamically adjust the parameters, predict the battery current state, quantify the uncertainty of the predicted state, reflect the observation error, and correct the prediction result to extract the real current signal characteristics; S43, dynamically adjust the model parameters according to the real-time input data, and make a high-precision prediction of the vehicle battery current change. The calculation formula is: ； In the formula, For the The model parameter values ​​at the iteration, For the The model parameter values ​​at the iteration, is the learning rate, is the number of iterations, is the loss function exist The gradient at S44, evaluating the prediction results based on the obtained multi-dimensional data of the vehicle battery, and analyzing the accuracy of the model prediction under different circumstances; S45. Perform high-precision prediction and analysis results based on vehicle battery current changes, and classify abnormal results.

4. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 3 is characterized in that: In step S42, the battery current state prediction makes a preliminary estimate of the future state of the battery current based on the dynamic model of the system, providing a basis for subsequent precise adjustment. The calculation formula is: ； In the formula, For the time step Time based on The prior prediction value of the system state made by the time information, is the transfer matrix, For the time step Time based on The a posteriori estimate of the system state obtained from its own information at each moment, is the control input matrix, For the time step The control input applied to the system at The uncertainty of the quantitative prediction state takes into account the unpredictable noise factors in the system and provides important parameters for subsequent calculations. The calculation formula is: ； In the formula, For the time step Time based on The posterior covariance matrix obtained from the moment information is For the time step Time Fusion The posterior covariance matrix after observing the information at each moment, is the matrix transpose, is the process noise covariance matrix; The error in the observation reflects the noise in the measurement process, and the calculation formula is: ； In the formula, is the Kalman gain, is the observation noise covariance, is the observation matrix; The prediction results are corrected to combine the actual measured data with the prediction results to make the estimated value closer to the actual battery current state. The calculation formula is: ； In the formula, For the time step Observed value at time ; The covariance adjustment method is used to extract the real current signal features, providing a reliable data basis for subsequent battery current monitoring and anomaly detection tasks. The calculation formula is: ； In the formula, is the identity matrix.

5. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 3 is characterized in that: The evaluation of the battery current prediction result in step S44 includes the following steps: S441, obtaining data on current amplitude, rate of change, current spectrum, time-frequency domain, and correlation with voltage, temperature, driving, and operation of in-vehicle electrical equipment; S442, calculate the current battery current change rate, the calculation formula is: ； In the formula, is the rate of change of current with time, for The current value at the moment, for The current value at the moment, is the sampling time interval; S443, calculate the current current spectrum characteristics, the calculation formula is: ； In the formula, For signal The Fourier transform result, that is, the frequency The function of For over time The changing current signal, is an imaginary unit, For time Calculus of S444. Calculate the time-frequency domain characteristics of the current, and the calculation formula is: ； In the formula, For signal In time and frequency The short-time Fourier transform result at is a time-varying variable The original signal of the change, is the window function used to intercept the signal A partial fragment of Time variable Calculus; S445. Calculate the relevant characteristics of the current current and voltage. The calculation formula is: ； In the formula, For current With voltage The Pearson correlation coefficient between For the Current sampling value, is the mean value of current sampling, For the A voltage sampling value, is the mean value of the voltage sampling, is the number of samples; S446. A feature selection method based on quantum genetic algorithm is used to quickly search for the optimal feature combination by utilizing the superposition state of quantum bits and quantum gate operations.

6. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 1 is characterized in that: In step S6, evaluating the abnormal results and classifying the risk levels includes the following steps: S61, obtaining abnormal classification results and adaptively adjusting parameters of result weights; S62. Based on the previously extracted multi-source data features, for the feature vector of the currently detected abnormal data, calculate the Euclidean distance between it and each feature template. The calculation formula is: ； In the formula, is the number of features, is the feature vector No. eigenvalues ​​that identify the relevant quantity, For the The identification object and The numerical value corresponding to the relevant quantity of the identifier; S63. Further determine whether it is abnormal data by calculating the degree of deviation between the current data and the normal data distribution. The calculation formula is: ； In the formula, is the value of the deviation degree, is the battery current related data after fusion, is the mean, is the standard deviation; S64. Assign a weight to each influencing factor and evaluate the severity of the anomaly. The calculation formula is: ； In the formula, is the abnormal severity assessment value, ranging from [0-1], is the total number of factors that affect the severity of the abnormality, For the The weight of the impact factor, For the The quantitative value of the impact factor; S65. Risk levels are determined based on the type and severity of the anomaly.

7. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 6 is characterized in that: In step S63, when If the value is less than the set threshold, it means that the abnormal classification result is accurate, then continue to step S64. If it is greater than the set threshold, it means that the abnormal classification is inaccurate, and then return to step S3.

8. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 6 is characterized in that: In step S65, a risk matrix is ​​established after the risk levels are divided. The risk levels are set to three levels: high risk, medium risk and low risk. When the risk level is classified as high risk, When the risk level is classified as medium risk, The risk level is classified as low risk.

9. The method for current monitoring and abnormality detection of new energy batteries based on adaptive control according to claim 1, characterized in that: In step S8, when sending abnormal results, risk levels, and emergency measures to generate warning information to the blockchain, each block contains the hash value, timestamp, and data content information of the previous block, and through the distributed ledger and encryption algorithm of the blockchain, the data is tamper-proof and traceable. When it is necessary to verify the authenticity and integrity of the data, the hash value is calculated and compared with the hash value stored on the blockchain to ensure the security of the data throughout its life cycle.

Images