CN-121978568-A - Dynamic adjustment method for safety threshold of reliable battery management system

CN121978568ACN 121978568 ACN121978568 ACN 121978568ACN-121978568-A

Abstract

The invention discloses a dynamic adjustment method of a safety threshold of a reliable battery management system, which relates to the technical field of battery management, and comprises the steps of obtaining real-time operation data of a battery and the aging degree of a battery core, calculating a normal operation threshold interval of the battery voltage through a preset Gaussian process regression model to obtain a voltage threshold, taking the threshold as a safety boundary, respectively calculating a unified battery threshold and a single battery cell differentiation threshold by combining the real-time operation data of the aging degree of the battery core and the voltage-dividing data to obtain an initial threshold set, predicting temperature and voltage through an LSTM model after the initial threshold set is calibrated, performing risk assessment according to a calibration threshold set and a predicted value to obtain a safety risk value, and adjusting the calibration threshold set according to the safety risk value to obtain the safety threshold set. The method integrates real-time, aging and prediction data to dynamically adjust the threshold value, ensures the safety of the whole period of the battery, adapts to different working conditions, improves the performance utilization rate, and gives consideration to unified management of the battery pack and differentiated protection of the battery cell monomers, thereby improving the safety judgment accuracy.

Inventors

JI XIANG
ZENG GUOJIAN
YANG YANHUI
ZHUANSUN MINGMING

Assignees

安徽锐能科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251224

Claims (8)

1. A method for dynamically adjusting a safety threshold of a reliable battery management system, the method comprising: Acquiring real-time operation data of a battery and the aging degree of a battery core, wherein the real-time operation data comprises a voltage value, a current value, a charge state, a temperature value and power; calculating a normal operation threshold interval of the voltage of the battery through a preset Gaussian process regression model to obtain a voltage confidence threshold; taking the voltage confidence threshold as a safety boundary, and respectively calculating a unified safety threshold and a battery cell monomer differentiation threshold of the battery according to the temperature value, the current value, the charge state, the power and the battery cell aging degree to obtain an initial threshold set; calibrating the initial threshold set to obtain a calibrated threshold set; Predicting the temperature value and the voltage value through an LSTM model to obtain a predicted temperature value and a predicted voltage value; Performing predictive risk assessment on the battery according to the calibration threshold set, the predicted temperature value and the predicted voltage value to obtain a safety risk value; And adjusting the calibration threshold set according to the security risk value to obtain a security threshold set.
2. The method for dynamically adjusting a safety threshold of a reliable battery management system according to claim 1, wherein the training process of the preset gaussian process regression model comprises: Acquiring historical operation data sets of a plurality of independent battery samples in a healthy state, and constructing a training data point set according to the data sets, wherein each training data point in the training data point set comprises a voltage value, a current value, a charge state and a temperature value; assigning a trainable memory cell parameter vector to each battery sample; each data point in the training data point set is spliced with a memory unit parameter vector of a battery sample corresponding to the data point, wherein the memory unit parameter vector comprises a current value, a charge state and a temperature value, so that a memory enhanced input characteristic set is obtained; The memory enhancement feature set is taken as input, and the voltage value is taken as output to construct a joint optimization objective function, wherein the joint optimization objective function is a fusion function of negative logarithmic marginal likelihood terms of all training data points and distribution consistency regularization terms of all memory unit parameter vectors; carrying out minimum solution on the joint optimization objective function through a preset gradient coupling optimization algorithm to obtain an optimal super-parameter and an optimal memory unit parameter vector; And inputting the optimal super parameters and the optimal memory unit parameter vector into a Gaussian process regression model to obtain a preset Gaussian process regression model.
3. The method of claim 2, wherein constructing a joint optimization objective function with the memory enhancement feature set as input and the voltage value as output comprises: The joint optimization objective function is l=0.5y T ×K -1 ×y+0.5ln|K|+(n/2)×ln(2π)+β×∑||m i -μ|| 2 , wherein L is a joint optimization objective function value, y is an actual voltage value vector of all training data points, K is a kernel matrix constructed according to a memory-enhanced input feature set, K -1 is an inverse matrix of the kernel matrix, k|is a determinant of the kernel matrix, n is the number of samples of the training data points, β is a regularization coefficient, m i is a memory unit parameter vector of an ith battery sample, and μ is an average value of all memory unit parameter vectors.
4. The method for dynamically adjusting the safety threshold of a reliable battery management system according to claim 2, wherein the working principle of the preset gradient coupling optimization algorithm comprises: respectively calculating a first gradient vector of the super parameter and a second gradient vector of all memory unit parameter vectors according to the joint optimization objective function; splicing the first gradient vector and the second gradient vector to obtain a joint gradient vector; performing linear transformation on the joint gradient vector to obtain an attention weight vector; respectively carrying out weighted combination on the first gradient vector and the second gradient vector according to the attention weight vector to obtain a second coupling vector and a first coupling vector; Weighting the first gradient vector and the first coupling vector to obtain a first updated gradient; Carrying out weighted summation on the second gradient vector and the second coupling vector to obtain a second updated gradient; And respectively carrying out gradient descent operation on the super-parameters and the memory unit parameter vectors according to the first updating gradient and the second updating gradient based on a minimum target of the joint optimization objective function to obtain optimal super-parameters and optimal memory unit parameter vectors.
5. The method of claim 1, wherein calculating the unified safety threshold and the cell differentiation threshold of the battery to obtain the initial threshold set according to the temperature value, the current value, the state of charge, the power, and the cell aging degree, respectively, comprises: Using a battery as a global node and using each battery cell as a local node to construct a double-layer time sequence chart network model; Calculating dynamic coupling weights between the global node and each local node and between the local nodes according to the temperature value, the current value, the charge state and the power to obtain a dynamic coupling weight set; Inputting the voltage confidence threshold value into a first risk quantization function in the time sequence diagram network model to calculate an initial risk value of the global node, and inputting the cell aging degree of each cell unit into a second risk quantization function to calculate an initial risk value of each local node; Performing multi-round coupling iterative computation on initial risk values of all nodes in the time sequence graph network model according to the dynamic coupling weight set until absolute values of variation amounts of the risk values of all nodes in two adjacent iterative computation are smaller than a preset convergence threshold; And extracting risk values of global nodes and local nodes in the time sequence graph network model and converting the risk values to obtain an initial threshold set.
6. The method for dynamic adjustment of a reliable battery management system safety threshold as recited in claim 4, wherein the first risk quantization function and the second risk quantization function comprise: The first risk quantization function is Wherein, R 1 is the initial risk potential value of the global node, V 1 is the rated voltage value of the battery, V 2 is the voltage confidence threshold, V 3 is the real-time voltage value, and k 1 is the first normalization coefficient; The second risk quantization function is Wherein S is the health state of the battery cell monomer and the value range (0, 1), The S attenuation rate of the battery cell is, alpha is an aging rate sensitivity coefficient, and k 2 is a first normalization coefficient.
7. The method of claim 1, wherein calibrating the initial set of thresholds to obtain a calibrated set of thresholds comprises: Acquiring an actual safety boundary offset under a historical working condition; Calculating a unified calibration factor and a differential calibration factor according to the actual safety boundary offset; Calibrating the unified safety thresholds in the initial threshold set according to the unified calibration factors to obtain calibrated unified safety thresholds; Calibrating each cell monomer differential threshold in the initial threshold set according to the differential calibration factor to obtain a calibrated differential threshold; and counting the calibration unified safety threshold and all calibration differentiation thresholds to obtain a calibration threshold set.
8. The method of claim 1, wherein performing predictive risk assessment on the battery based on the set of calibration thresholds, the predicted temperature value, and the predicted voltage value to obtain a safety risk value comprises: if the predicted voltage value and the predicted temperature value are both larger than the calibrated unified safety threshold value in the calibrated threshold set, respectively calculating the deviation degree of the voltage and the temperature to obtain a voltage deviation risk value and a temperature deviation risk value; Calculating parameter dispersion among the cells according to the numerical distribution characteristics of the cell monomer differentiation threshold value in the calibration threshold set to obtain a monomer consistency risk coefficient; and carrying out weighted fusion on the voltage deviation risk value, the temperature deviation risk value and the monomer consistency risk coefficient to obtain a safety risk value.

Description

Dynamic adjustment method for safety threshold of reliable battery management system Technical Field The invention belongs to the technical field of battery management, and particularly relates to a dynamic adjustment method of a safety threshold of a reliable battery management system. Background In the new energy field, the battery is used as a core energy storage unit, the safety performance and the service life of the battery directly determine the operation reliability of the end products such as an energy storage system, an electric automobile and the like, the battery management system is a core component for guaranteeing the safety and stable operation of the battery, the safety threshold is a key judgment basis for realizing overcharge, overdischarge, overcurrent and overtemperature protection of the battery management system, the safety threshold of the traditional battery management system mostly adopts a fixed value, and the value is usually determined based on nominal parameters of the battery when leaving a factory or test data under the standard working condition of a laboratory and is kept constant in the whole life cycle of the battery. The patent application CN120327265A discloses a dynamic regulation system for thermal runaway and safety threshold of a battery, which comprises a data acquisition module, a data preprocessing module, an initial safety threshold setting module, a dynamic threshold regulation module and a thermal runaway prediction model module, wherein the data acquisition module is used for acquiring battery state parameters, environment parameters and vehicle use working conditions, the data preprocessing module is used for preprocessing acquired data, the initial safety threshold setting module is used for calculating an initial risk threshold of the thermal runaway of the battery by utilizing a statistical rule in combination with a large amount of acquired historical sample data and using the initial risk threshold as a judgment standard of an initial deployment stage, the dynamic threshold regulation module is used for carrying out dynamic personalized regulation on the initial risk threshold according to a Bayesian optimization algorithm, and the thermal runaway prediction model module is used for predicting the future 24-hour battery risk level and the future 1-hour battery temperature by combining a traditional machine learning and deep learning model. However, the initial threshold value is set according to a statistical rule, the physical and chemical properties of the battery core and the thermal runaway failure mechanism are not deeply combined, so that the initial reference lacks a reasonable support of a physical layer, and further, the inherent defects of the initial reference cannot be subjected to targeted correction in the subsequent dynamic adjustment process, so that the threshold value is always unfolded around the initial value lacking the physical reference, the actual safety boundary of the battery is difficult to accurately match finally, and the deviation exists in the threshold value after the dynamic adjustment, so that the accuracy of safety judgment is influenced. Disclosure of Invention The invention aims to solve the problem that the threshold is difficult to match with the actual safety boundary of a battery due to the fact that the initial threshold set by a statistical rule lacks of physical and chemical characteristics of the battery and is supported by a fault mechanism and the initial reference defect is not corrected pertinently by dynamic adjustment, and provides a dynamic adjustment method of the safety threshold of a reliable battery management system. The invention provides a dynamic adjustment method of a safety threshold of a reliable battery management system, which comprises the following steps: Acquiring real-time operation data of a battery and the aging degree of a battery core, wherein the real-time operation data comprises a voltage value, a current value, a charge state, a temperature value and power; calculating a normal operation threshold interval of the voltage of the battery through a preset Gaussian process regression model to obtain a voltage confidence threshold; taking the voltage confidence threshold as a safety boundary, and respectively calculating a unified safety threshold and a battery cell monomer differentiation threshold of the battery according to the temperature value, the current value, the charge state, the power and the battery cell aging degree to obtain an initial threshold set; calibrating the initial threshold set to obtain a calibrated threshold set; Predicting the temperature value and the voltage value through an LSTM model to obtain a predicted temperature value and a predicted voltage value; Performing predictive risk assessment on the battery according to the calibration threshold set, the predicted temperature value and the predicted voltage value to obtain a safety risk value; And adjust