Battery pack safety management method and system based on intelligent algorithm

Abstract

The invention relates to the field of battery safety management, in particular to a battery pack safety management method and system based on an intelligent algorithm. A battery pack safety management system based on an intelligent algorithm comprises a battery pack parameter monitoring module, a weight calculation module and a fault prediction module. According to the method, key parameters such as the single voltage, the total voltage, the charging and discharging current and the cell surface temperature are acquired in real time, the dynamic threshold values of the total voltage, the charging and discharging current and the charge state are calculated respectively, judgment is performed in combination with the preset fixed threshold value, abnormal conditions in the running process of the battery pack can be found in time, and corresponding abnormal response measures are executed; comprehensive real-time monitoring and dynamic threshold judgment of the parameters of the battery pack are achieved, the safety and reliability of the battery pack are effectively improved, and safety accidents caused by abnormal parameters are prevented.

Description

Technical Field

[0001] The present invention relates to the field of battery safety management, and particularly to a battery pack safety management method and system based on an intelligent algorithm. Background Art

[0002] With the wide application of lithium-ion battery packs in the fields of new energy vehicles, energy storage systems, consumer electronics, etc., their safety management has become an important topic in the industry development; during the operation of the battery pack, faults such as overvoltage, overcurrent, and local overheating are likely to be caused by factors such as monomer parameter differences, aging effects, heat accumulation, and complex load conditions. In severe cases, it may lead to thermal runaway or even explosion; traditional safety management methods mostly rely on fixed threshold detection mechanisms, for example, setting static safety intervals for parameters such as monomer voltage and total current, and triggering protection actions once parameter overlimits are detected; however, the adaptability of such methods under dynamic working conditions is insufficient: in actual use of the battery pack, the attenuation of the state of health (SoH), the change of ambient temperature, the fluctuation of charge and discharge rate, etc. will all significantly affect the reasonable threshold range of each parameter; if the threshold is set too conservatively, false alarms may be frequently triggered, interfering with normal use; if the threshold is too loose, potential risks cannot be effectively warned.

[0003] In the prior art, some improvement schemes attempt to introduce a dynamic threshold adjustment strategy, such as correcting the maximum charge and discharge current according to the battery SoH; however, such methods are still limited to single-parameter correction and fail to establish a global framework for coordinated adjustment of multiple parameters (such as total voltage, state of charge); in addition, traditional methods usually treat the contribution degrees of different parameters to faults with equal weights, ignoring the priority indication effect of the mutation of key parameters (such as cell temperature difference, maximum monomer voltage) on fault prediction; although some studies propose to use machine learning models for fault prediction, their input features usually directly splice the time series data of each parameter and do not perform weighted fusion based on parameter correlation degrees, resulting in insufficient sensitivity of the model to core risk signals.

[0004] Therefore, the present invention proposes a battery pack safety management method and system based on an intelligent algorithm, constructs an end-to-end safety warning and response system, and breaks through the performance bottleneck of the prior art. Summary of the Invention

[0005] By obtaining key parameters such as monomer voltage, total voltage, charge and discharge current, and cell surface temperature in real time, calculating the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively, and making judgments in combination with preset fixed thresholds, the present invention can timely detect abnormal situations during the operation of the battery pack and execute corresponding abnormal response measures, realizing comprehensive real-time monitoring and dynamic threshold judgment of battery pack parameters, effectively improving the safety and reliability of the battery pack, and preventing safety accidents caused by parameter abnormalities.

[0006] A battery pack safety management method based on intelligent algorithms, including:

[0007] Real-time monitoring of battery pack parameters, where the battery pack parameters include single-cell voltage, total voltage, charge and discharge current, surface temperature of the battery cell, temperature difference between adjacent battery cells, state of charge, and state of health;

[0008] Calculate and obtain the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively, and based on the preset fixed thresholds of the single-cell voltage, surface temperature of the battery cell, temperature difference between adjacent battery cells, and state of health, determine whether the currently obtained battery pack parameters are at a normal level. If so, no operation is performed; if not, an abnormal response measure is executed;

[0009] Obtain several battery pack data samples, calculate the correlation between each battery pack parameter and the battery pack fault state, and then calculate and obtain the weight coefficient of each battery pack parameter;

[0010] Sort the maximum single-cell voltage, total voltage, charge and discharge current, maximum surface temperature of the battery cell, maximum temperature difference between adjacent battery cells, average state of charge, and minimum state of health obtained in the most recent t time according to time to form a battery pack parameter time series set combination; use the battery pack parameter time series set combination and the weight coefficient of each battery pack parameter as the input of the fault prediction model, and output the fault prediction result; if the fault prediction result is that there is a fault, an early warning response measure is executed.

[0011] Preferably, the dynamic thresholds of the total voltage, charge and discharge current, and state of charge are calculated and obtained respectively, and the specific operations are as follows:

[0012] Use the formula V total,max =N(V cell,max -ΔV cell ) to calculate and obtain the theoretical extreme value V total，max of the total voltage;

[0013] Use the formula I SOH =I rated -I rated (100 - SOH min )b to calculate and obtain the health state correction current value I SOH ;

[0014] Based on all the obtained surface temperatures T i of the single battery cells, select the highest surface temperature T max of them; if T max < - 20°C or T max ≥50°C, then the temperature coefficient α T =0; if - 20°C ≤ T max <50°C, then use the formula to calculate and obtain the temperature coefficient α T ;

[0015] Based on all the obtained single-cell SOCs i , if the battery pack is in a discharging state, select the lowest SOC from all the single-cell SOCs i , and use the formula min to calculate and obtain the state-of-charge coefficient β ; if the battery pack is in a charging state, select the highest SOC from all the single-cell SOCs SOC , and use the formula i to calculate and obtain the state-of-charge coefficient β max ; ; SOC ;

[0016] Finally, use the formula I max = min(I SOH , α T I rated , β SOC I rated ) to calculate and obtain the maximum charge and discharge current I max ;

[0017] Use the formula to calculate and obtain the dynamic thresholds of the state of charge of each single battery in the battery pack, including the minimum working state of charge SOC i，low and the maximum working state of charge SOC i，high .

[0018] Preferably, obtain several battery pack data samples, and the specific operations are as follows:

[0019] Obtain the historical time-series data set of the battery pack, including the historical battery pack parameters X (g) at several consecutive time points and the corresponding fault labels Y, g = 1, 2,..., 7. The historical battery pack parameters at each time point specifically include the maximum single-cell voltage X (1) , the total voltage X (2) , the charge and discharge current X (3) , the maximum cell surface temperature X (4) , the maximum adjacent cell temperature difference X (5) , the average state of charge X (6) and the lowest health state X (7) ; randomly select m fault time points with a fault label of 1 from the historical time-series data set of the battery pack, and intercept the data segments within t time before each fault time point as m fault samples; randomly select M normal time points with a fault label of 0 from the historical time-series data set of the battery pack, and intercept the data segments within t time before each normal time point as M normal samples, and finally obtain m + M battery pack data samples;

[0020] For any normal sample and faulty sample, based on the historical battery pack parameter set therein j = 1, 2, …, n; n represents the number of time points included in each sample; obtain the statistical feature set x(h) of the current battery pack data sample, h = 1, 2, …, 17.

[0021] Preferably, calculate the correlation degree between each battery pack parameter and the battery pack fault state, and the specific operation is as follows:

[0022] Perform normalization on the statistical feature set x(h) included in the m + M battery pack data samples k to obtain the m + M standardized statistical feature sets x(h) k,norm , k = 1, 2, …, m + M;

[0023] Calculate and obtain the maximum information coefficient MIC(x(h); Y) of the statistical feature set x(h) and the battery pack fault state, and further obtain the correlation degree R between each battery pack parameter and the battery pack fault state (g) .

[0024] Preferably, based on the correlation degree between each battery pack parameter and the battery pack fault state, calculate and obtain the weight coefficient of each battery pack parameter, and the specific operation is as follows:

[0025] Use the formula to calculate and obtain the weight coefficient ω of each battery pack parameter (g) .

[0026] Preferably, the fault prediction model is established based on LSTM and CNN, and includes an input layer, a local feature extraction layer, an LSTM layer, a feature weighted fusion layer, a fully connected layer, and an output layer;

[0027] The input layer is used to receive the battery pack parameter time series set combination and the weight coefficient of each battery pack parameter;

[0028] The local feature extraction layer performs local feature extraction on the battery pack parameter time series set combination based on a one-dimensional convolutional neural network to obtain local feature vectors;

[0029] The LSTM layer is used to extract the time series feature vectors of each battery pack parameter in the battery pack parameter time series set combination respectively;

[0030] The feature weighted fusion layer is used to apply the weight coefficient of each battery pack parameter to perform weighted fusion on the time series feature vectors corresponding to each battery pack parameter to obtain weighted time series feature vectors; subsequently, the weighted time series feature vectors are concatenated with the static feature vectors to obtain comprehensive feature vectors;

[0031] The fully connected layer is used to perform non-linear transformation and high-order abstraction on the comprehensive feature vectors, learn the internal discrimination boundary of the fault pattern, and output high-dimensional feature representations;

[0032] The output layer is used to output the fault prediction result.

[0033] Preferably, for the training of the fault prediction model, the specific operations are as follows:

[0034] Divide all the obtained battery pack data samples into a training set and a validation set. Use the training set to train the fault prediction model with initialized parameters, and then use the validation set to verify the fault prediction model to obtain the verification result. Set the training conditions, and judge whether the obtained verification result meets the training conditions. If so, output the trained fault prediction model.

[0035] A battery pack safety management system based on an intelligent algorithm, comprising:

[0036] The battery pack parameter monitoring module includes a data acquisition unit, a threshold adjustment unit, and a first judgment unit; the data acquisition unit is used to monitor the battery pack parameters in real time; the threshold adjustment unit is used to calculate the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively; the first judgment unit is used to judge whether the currently obtained battery pack parameters are at a normal level. If so, no operation is performed; if not, an abnormal response measure is executed;

[0037] The weight calculation module includes a sample acquisition unit and a weight calculation unit; the sample acquisition unit is used to acquire a number of battery pack data samples; the weight calculation unit is used to calculate the weight coefficients of each battery pack parameter;

[0038] The fault prediction module includes a fault prediction unit and a second judgment unit. The fault prediction unit is used to use the battery pack parameter time series set combination and the weight coefficients of each battery pack parameter as the input of the fault prediction model and output the fault prediction result; the second judgment unit is used to execute a warning response measure when the fault prediction result indicates that a fault exists.

[0039] The present invention has the following advantages:

[0040] 1. By obtaining key parameters such as the individual voltage, total voltage, charge and discharge current, and cell surface temperature in real time, calculating the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively, and making judgments in combination with preset fixed thresholds, the present invention can timely detect abnormal situations during the operation of the battery pack and execute corresponding abnormal response measures, realizing comprehensive real-time monitoring of battery pack parameters and dynamic threshold judgment, effectively improving the safety and reliability of the battery pack, and preventing safety accidents caused by abnormal parameters.

[0041] 2. The present invention obtains the historical time series data set of the battery pack and intercepts data samples, calculates the correlation degree and weight coefficient between each battery pack parameter and the fault state, uses these data as inputs, and utilizes a fault prediction model that integrates LSTM and CNN for fault prediction. At the same time, the model focuses more on high-risk signals such as sudden temperature difference changes and single-cell voltage differences, can accurately predict the faults of the battery pack in advance, provides a basis for taking early warning response measures, further improves the safety management level of the battery pack, and ensures the stable operation of the battery pack. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic structural diagram of a battery pack safety management system based on an intelligent algorithm adopted in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention.

[0044] Embodiment 1, a battery pack safety management method based on an intelligent algorithm, includes:

[0045] Real-time monitoring of battery pack parameters, where the battery pack parameters include single-cell voltage, total voltage, charge and discharge current, surface temperature of the battery cell, temperature difference between adjacent battery cells, state of charge, and state of health; by real-time monitoring of the battery pack parameters, abnormal situations in the battery pack, such as overvoltage, overcurrent, overheating, etc., can be discovered in a timely manner, so as to take corresponding protection measures to prevent safety accidents such as thermal runaway, short circuit, and fire of the battery pack;

[0046] Calculate and obtain the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively, and based on the preset fixed thresholds of the single-cell voltage, surface temperature of the battery cell, temperature difference between adjacent battery cells, and state of health, determine whether the currently obtained battery pack parameters meet the dynamic thresholds of the total voltage, charge and discharge current, and state of charge and the fixed thresholds of the single-cell voltage, surface temperature of the battery cell, temperature difference between adjacent battery cells, and state of health. If they are met, no operation is performed; if not, an abnormal response measure is executed; through the calculation and judgment of the dynamic thresholds and the fixed thresholds, the system can monitor the operating state of the battery pack in real time, discover abnormal situations in a timely manner and issue early warnings, and avoid safety accidents caused by parameter abnormalities; the calculation of the dynamic thresholds takes into account the real-time state of the battery pack (such as SOH and temperature), can more flexibly adapt to the changes of the battery pack, and avoid misjudgment or missed judgment caused by the inability of the fixed thresholds to adapt to battery aging or environmental changes. Thus, when the battery pack parameters exceed the thresholds, the system can take protection measures in a timely manner, such as current limiting or power off, to prevent overcharging, over-discharging, overheating, etc. of the battery pack and extend the service life of the battery pack;

[0047] Obtain the historical time-series dataset of the battery pack, extract several battery pack data samples from the historical time-series dataset of the battery pack, and then calculate the correlation degree between each battery pack parameter and the battery pack fault status; by calculating the correlation degree between each battery pack parameter and the fault status, it is possible to identify which parameters have significant change characteristics when a fault occurs; these characteristics can be used as an important basis for fault diagnosis to help improve the accuracy of fault detection; the calculation results of the correlation degree can be used to determine the weights of each battery pack parameter in the fault prediction model; the higher the correlation degree between the parameter and the fault status, the greater its weight in the model, thus playing a more important role in fault prediction; by calculating the correlation degree, parameters highly correlated with the fault status can be screened out, reducing the interference of irrelevant parameters, thereby optimizing the input of the fault prediction model and improving the prediction performance and generalization ability of the model;

[0048] Based on the correlation degree between each battery pack parameter and the battery pack fault status, calculate and obtain the weight coefficients of each battery pack parameter;

[0049] Sort the maximum single-cell voltage, total voltage, charge and discharge current, maximum cell surface temperature, maximum adjacent cell temperature difference, average state of charge, and minimum health status obtained within the most recent t time in chronological order to form a battery pack parameter time-series set combination, including a single-cell voltage time-series set, a total voltage time-series set, a charge and discharge current time-series set, a cell temperature time-series set, a cell temperature difference time-series set, a state of charge time-series set, and a health status time-series set. This kind of time-series feature is very important for fault prediction because many faults will have obvious parameter change trends before they occur; use the battery pack parameter time-series set combination and the weight coefficients of each battery pack parameter as the input of the fault prediction model, and output the fault prediction result; if the fault prediction result is that there is a fault, then execute warning response measures, such as alarm, current limiting, power off, etc., to ensure the safety of the battery pack; this kind of timely response measure can effectively prevent the further development of the fault and protect the battery pack and related equipment.

[0050] Calculate and obtain the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively. The specific operations are as follows:

[0051] For the calculation of the total voltage dynamic threshold, set the fixed upper limit V of the single-cell voltage cell,max , and use the formula V total,max = N(V cell,max - ΔV cell ) to calculate and obtain the theoretical extreme value V of the total voltage total，max , where i = 1, 2,..., N; N represents the number of series-connected single cells in the battery pack; V cell，max is the fixed upper limit of the single-cell voltage;

[0052] For the calculation of the charge and discharge current dynamic threshold, based on all the obtained single-cell SOHi , select the smallest SOH among them min , and use formula I SOH = I rated - I rated (100 - SOH min )b to calculate and obtain the healthy state correction current value I SOH , where I rated is the rated current value, b is the reduction coefficient, and for every 1% decrease in SOH min , the healthy state correction current value I SOH decreases by b%;

[0053] Based on all the surface temperatures T of the individual battery cells obtained i , select the highest surface temperature T of the individual battery cells max ; if T max < -20°C or T max ≥ 50°C, then the temperature coefficient α T = 0; if -20°C ≤ T max < 50°C, then use the formula to calculate and obtain the temperature coefficient α T ;

[0054] Based on all the individual SOCs obtained i , if the battery pack is in the discharge state, then select the lowest SOC i from all the individual SOCs min , and use the formula to calculate and obtain the state of charge coefficient β SOC ; if the battery pack is in the charging state, then select the highest SOC i from all the individual SOCs max , and use the formula to calculate and obtain the state of charge coefficient β SOC ;

[0055] Finally, use the formula I max = min(I SOH , α T I rated , β SOC I rated ) to calculate and obtain the maximum charge and discharge current I max ;

[0056] For the calculation of the state of charge dynamic threshold, use the formula to calculate and obtain the state of charge dynamic thresholds of each individual battery in the battery pack, including the minimum operating state of charge SOC i，low and the maximum operating state of charge SOC i，high , where SOC base，low and SOC base，highThey are the minimum rated state of charge and the maximum rated state of charge respectively;

[0057] In the calculation of the dynamic threshold, the specific numerical settings of some parameters (such as in the calculation formulas of the temperature coefficient and the state of charge coefficient) are optimized based on typical application scenarios and experimental data, and their values can be adjusted according to actual needs; for example, for different types of battery packs (such as lithium iron phosphate and ternary lithium batteries), different working conditions (such as high-rate charge and discharge or low-temperature environment) or specific safety level requirements, the relevant parameters can be adaptively corrected through theoretical analysis, simulation verification or actual testing; the physical meanings and dynamic correlation logics of the parameters in the formulas are the keys to the technical solutions of the present invention, and the selection of specific values does not affect the universality and innovativeness of the method.

[0058] Extract several battery pack data samples from the battery pack historical time series dataset. The specific operations are as follows:

[0059] Obtain the battery pack historical time series dataset, including the historical battery pack parameters X at several consecutive time points (g) and the corresponding fault labels Y, where Y is equal to 0 or 1. Y = 1 indicates a fault, and Y = 0 indicates normal. g = 1, 2,..., 7. The historical battery pack parameters at each time point specifically include the maximum single-cell voltage X (1) , the total voltage X (2) , the charge and discharge current X (3) , the maximum cell surface temperature X (4) , the maximum adjacent cell temperature difference X (5) , the average state of charge X (6) and the lowest health state X (7) ; randomly select m fault time points with a fault label of 1 from the battery pack historical time series dataset, and intercept the data segments within t time before each fault time point as fault data windows; take the intercepted m fault data windows as m fault samples; randomly select M normal time points with a fault label of 0 from the battery pack historical time series dataset, and intercept the data segments within t time before each normal time point as normal data windows; take the intercepted M normal data windows as M normal samples, and finally obtain m + M battery pack data samples;

[0060] For any normal sample and fault sample, based on the historical battery pack parameter set therein j = 1, 2,..., n; n represents the number of time points included in each sample; for the maximum single-cell voltage dataset the total voltage dataset the charge and discharge current dataset the maximum cell surface temperature dataset and the maximum adjacent cell temperature difference dataset Calculate the mean, standard deviation, and peak-to-peak value of each data set respectively, and obtain x(1), x(2), …, x(15). x(1) to x(3) are the mean, standard deviation, and peak-to-peak value of the maximum single-cell voltage data set in sequence. The mean, standard deviation, and peak-to-peak value of the total voltage data set are x(4) to x(6) in sequence, and so on. For the average state of charge data set and the lowest state of health data set calculate the mean of each data set respectively, and obtain x(16) and x(17). Take x(h) as the statistical feature set of the current sample, where h = 1, 2, …, 17.

[0061] Calculate the correlation degree between each battery pack parameter and the battery pack fault state. The specific operations are as follows:

[0062] Based on the statistical feature set x(h) k contained in m + M battery pack data samples k and the fault label Y calculate and obtain the standardized statistical feature set x(j) of m + M battery pack data samples k，norm , where max(x(h) k ) and min(x(h) k represent the maximum and minimum values in the m + M statistical feature sets x(h) k respectively;

[0063] Set multiple grid combinations v×w ∈ {2×2, 3×2, …, 10×2}, traverse any grid combination v×w, and for (x(h) k，norm , Y k ) of each battery pack data sample, determine its position in the current grid combination v×w. Use the formula to calculate and obtain the joint probability distribution p v，w ; d v，w represents the number of samples falling into the grid (v, w) among the M + m battery pack data samples. Subsequently, use the formulas and to calculate and obtain the marginal probability distributions p(x(h) v ) and p(Y w ), where v now represents the v value of the current network combination. Finally, use the formula to calculate and obtain the mutual information coefficient MI(x(h); Y) between the statistical feature set x(h) and the battery pack fault state;

[0064] Based on the mutual information coefficients corresponding to all the obtained grid combinations, select the maximum mutual information coefficient as the maximum information coefficient MIC(x(h); Y) between the statistical feature set x(h) and the battery pack fault state;

[0065] Take the maximum value among MIC(x(1); Y) to MIC(x(3); Y) as the maximum single cell voltage X (1) The correlation degree R with the battery pack fault state (1) ;

[0066] Take the maximum value among MIC(x(4); Y) to MIC(x(6); Y) as the total voltage X (2) The correlation degree R with the battery pack fault state (2) ;

[0067] Take the maximum value among MIC(x(7)Y) to MIC(x(9)Y) as the charge and discharge current X (3) The correlation degree R with the battery pack fault state (3) ;

[0068] Take the maximum value among MIC(x(10); Y) to MIC(x(12); Y) as the maximum cell surface temperature X (4) The correlation degree R with the battery pack fault state (4) ;

[0069] Take the maximum value among MIC(x(13); Y) to MIC(x(15); Y) as the maximum adjacent cell temperature difference X (5) The correlation degree R with the battery pack fault state (5) ;

[0070] Take MIC(X(16); Y) as the average state of charge X (6) The correlation degree R with the battery pack fault state (6) ;

[0071] Take MIC(x(17); Y) as the lowest health state X (7) The correlation degree R with the battery pack fault state (7) 。

[0072] Based on the correlation degrees between each battery pack parameter and the battery pack fault state, calculate the weight coefficients of each battery pack parameter. The specific operations are as follows:

[0073] Use the formula Calculate the weight coefficient ω of each battery pack parameter (g) 。

[0074] The fault prediction model is established based on LSTM and CNN, and includes an input layer, a local feature extraction layer, an LSTM layer, a feature weighted fusion layer, a fully connected layer, and an output layer;

[0075] The input layer is used to receive the time - series set combination of battery pack parameters and the weight coefficients of each battery pack parameter;

[0076] The local feature extraction layer performs local feature extraction on the time - series set combination of battery pack parameters based on a one - dimensional convolutional neural network to obtain local feature vectors;

[0077] The LSTM layer is used to extract the time - series feature vectors of each battery pack parameter in the time - series set combination of battery pack parameters respectively;

[0078] The feature weighted fusion layer is used to apply the weight coefficients of each battery pack parameter to perform weighted fusion on the time - series feature vectors corresponding to each battery pack parameter to obtain weighted time - series feature vectors; subsequently, the weighted time - series feature vectors are concatenated with static feature vectors to obtain comprehensive feature vectors;

[0079] The fully - connected layer is used to perform non - linear transformation and high - order abstraction on the comprehensive feature vectors, learn the internal discrimination boundary of the fault mode, and output high - dimensional feature representations;

[0080] The output layer is used to output the fault prediction result.

[0081] For the training of the fault prediction model, the specific operations are as follows:

[0082] All the obtained battery pack data samples are divided into a training set and a validation set. The fault prediction model with initialized parameters is trained using the training set, and then the fault prediction model is verified through the validation set to obtain a verification result; set training conditions, and judge whether the obtained verification result meets the training conditions. If so, output the trained fault prediction model.

[0083] Embodiment 2, a battery pack safety management system based on an intelligent algorithm, as Figure 1 shown, includes:

[0084] The battery pack parameter monitoring module includes a data acquisition unit, a threshold adjustment unit, and a first judgment unit; the data acquisition unit is used to monitor the battery pack parameters in real - time. The battery pack parameters include single - cell voltage, total voltage, charge - discharge current, surface temperature of the battery cell, temperature difference between adjacent battery cells, state of charge, and state of health; the threshold adjustment unit is used to calculate and obtain the dynamic thresholds of the total voltage, charge - discharge current, and state of charge respectively; the first judgment unit judges whether the currently obtained battery pack parameters meet the dynamic thresholds of the total voltage, charge - discharge current, and state of charge and the fixed thresholds of the single - cell voltage, surface temperature of the battery cell, temperature difference between adjacent battery cells, and state of health. If they are met, no operation is performed; if not, an abnormal response measure is executed;

[0085] The weight calculation module includes a sample acquisition unit and a weight calculation unit; the sample acquisition unit is used to acquire the historical time series dataset of the battery pack and intercept several battery pack data samples from the historical time series dataset of the battery pack; the weight calculation unit is used to calculate the correlation degree between each battery pack parameter and the fault state of the battery pack, and calculate and obtain the weight coefficient of each battery pack parameter based on the correlation degree between each battery pack parameter and the fault state of the battery pack.

[0086] The fault prediction module includes a fault prediction unit and a second judgment unit. The fault prediction unit is used to sort the maximum single cell voltage, total voltage, charge and discharge current, maximum cell surface temperature, maximum adjacent cell temperature difference, average state of charge, and minimum health state obtained within the most recent t time in time sequence to form a battery pack parameter time series set combination, including a single cell voltage time series set, a total voltage time series set, a charge and discharge current time series set, a cell temperature time series set, a cell temperature difference time series set, a state of charge time series set, and a health state time series set; use the battery pack parameter time series set combination and the weight coefficient of each battery pack parameter as the input of the fault prediction model, and output the fault prediction result; the second judgment unit is used to execute the warning response measure when the fault prediction result is that a fault exists.

[0087] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention. The parts not described in detail in this specification belong to the prior art well known to those of ordinary skill in the art.

Claims

1. A battery pack safety management method based on intelligent algorithms, characterized in that, Including: Real-time monitoring of battery pack parameters, which include single-cell voltage, total voltage, charge and discharge current, cell surface temperature, adjacent cell temperature difference, state of charge, and state of health; Respectively calculate and obtain the dynamic thresholds of the total voltage, charge and discharge current, and state of charge, and based on the preset fixed thresholds of the single-cell voltage, cell surface temperature, adjacent cell temperature difference, and state of health, determine whether the currently obtained battery pack parameters are at a normal level. If so, no operation is performed; if not, an abnormal response measure is executed; Obtain several battery pack data samples, calculate the correlation between each battery pack parameter and the battery pack fault state, and then calculate and obtain the weight coefficient of each battery pack parameter; Sort the maximum single-cell voltage, total voltage, charge and discharge current, maximum cell surface temperature, maximum adjacent cell temperature difference, average state of charge, and minimum state of health obtained within the most recent t time in chronological order to form a battery pack parameter time series set combination; Use the battery pack parameter time series set combination and the weight coefficient of each battery pack parameter as the input of the fault prediction model, and output the fault prediction result; If the fault prediction result indicates a fault, execute a warning response measure.

2. The battery pack safety management method based on an intelligent algorithm according to claim 1, wherein Respectively calculate and obtain the dynamic thresholds of the total voltage, charge and discharge current, and state of charge. The specific operations are as follows: Use the formula V total，max = N(V cell，max - ΔV cell ) to calculate and obtain the theoretical extreme value V total，max ; Using formula I SOH = I rated - I rated (100 - SOH min )b to calculate and obtain the healthy state correction current value I SOH ; Based on the obtained surface temperature T of all single cells i , select the highest surface temperature T of the single cells max ; if T max < -20°C or T max ≥ 50°C, then the temperature coefficient α T = 0; if -20°C ≤ T max < 50°C, then use the formula to calculate and obtain the temperature coefficient α T ; Based on all the obtained individual SOCs i , if the battery pack is in a discharging state, select the lowest SOC from all the individual SOCs i , and use the formula min to calculate and obtain the state-of-charge coefficient β ; if the battery pack is in a charging state, select the highest SOC from all the individual SOCs SOC , and use the formula i to calculate and obtain the state-of-charge coefficient β max ; SOC ​​ Finally, use formula I max = min(I SOH , α T I rated , β SOC I rated ) to calculate and obtain the maximum charge and discharge current I max ; Using the formula calculate and obtain the dynamic threshold of the state of charge of each single battery in the battery pack, including the minimum working state of charge SOC i，low and the maximum working state of charge SOC i，high .

3. The battery pack safety management method based on an intelligent algorithm according to claim 2, characterized in that, Obtain several battery pack data samples. The specific operations are as follows: Obtain the historical time series dataset of the battery pack, including the historical battery pack parameters X at a number of consecutive time points (g) and the corresponding fault labels Y, where g = 1, 2, …, 7. The historical battery pack parameters at each time point specifically include the maximum single cell voltage X (1) , the total voltage X (2) , the charge and discharge current X (3) , the maximum cell surface temperature X (4) , the maximum adjacent cell temperature difference X (5) , the average state of charge X (6) and the lowest state of health X (7) ; Randomly select m fault time points with a fault label of 1 from the historical time series dataset of the battery pack, and intercept the data segments within t time before each fault time point as m fault samples; Randomly select M normal time points with a fault label of 0 from the historical time series dataset of the battery pack, and intercept the data segments within t time before each normal time point as M normal samples, and finally obtain m + M battery pack data samples For any normal sample and fault sample, based on the historical battery pack parameter set therein n represents the number of time points included in each sample; obtain the statistical feature set x(h) of the current battery pack data sample, where h = 1, 2,..., 17.

4. The battery pack safety management method based on intelligent algorithm according to claim 3, wherein Calculate the correlation degree between each battery pack parameter and the battery pack fault state. The specific operations are as follows: Normalize the statistical feature set x(h) contained in the m+M battery pack data samples k to obtain the m+M standardized statistical feature sets x(h) k，norm , where k = 1, 2, …, m+M; Calculate and obtain the maximum information coefficient MIC(x(h); Y) between the statistical feature set x(h) and the battery pack fault status, and then obtain the correlation degree R between each battery pack parameter and the battery pack fault status (g) .

5. The method for battery pack safety management based on intelligent algorithm according to claim 4, characterized in that, Based on the correlation degree between each battery pack parameter and the battery pack fault state, calculate and obtain the weight coefficient of each battery pack parameter. The specific operations are as follows: Using the formula calculate and obtain the weight coefficient ω of each battery pack parameter (g) .

6. The battery pack safety management method based on an intelligent algorithm according to claim 5, characterized in that The fault prediction model is established based on LSTM and CNN, and includes an input layer, a local feature extraction layer, an LSTM layer, a feature weighted fusion layer, a fully connected layer, and an output layer; The input layer is used to receive the battery pack parameter time series set combination and the weight coefficient of each battery pack parameter; The local feature extraction layer performs local feature extraction on the battery pack parameter time series set combination based on a one-dimensional convolutional neural network to obtain local feature vectors; The LSTM layer is used to respectively extract the time series feature vectors of each battery pack parameter in the battery pack parameter time series set combination; The feature weighted fusion layer is used to apply the weight coefficient of each battery pack parameter to weight and fuse the time series feature vectors corresponding to each battery pack parameter to obtain weighted time series feature vectors; subsequently, the weighted time series feature vectors are concatenated with static feature vectors to obtain comprehensive feature vectors; The fully connected layer is used to perform non-linear transformation and high-order abstraction on the comprehensive feature vectors, learn the internal discrimination boundary of the fault mode, and output high-dimensional feature representations; The output layer is used to output the fault prediction result.

7. A battery pack safety management method based on an intelligent algorithm according to claim 6, characterized in that, Regarding the training of the fault prediction model, the specific operations are as follows: Divide all the obtained battery pack data samples into a training set and a validation set, use the training set to train the fault prediction model with initialized parameters, and then verify the fault prediction model through the validation set to obtain a verification result; Set training conditions, and judge whether the obtained verification result meets the training conditions. If so, output the trained fault prediction model.

8. A battery pack safety management system based on intelligent algorithms, characterized in that, The system is applied to a battery pack safety management method based on an intelligent algorithm described in any one of claims 1-7 above, including: The battery pack parameter monitoring module includes a data acquisition unit, a threshold adjustment unit, and a first judgment unit; the data acquisition unit is used to monitor the battery pack parameters in real time; the threshold adjustment unit is used to calculate the dynamic thresholds of the total voltage, charge and discharge current, and state of charge respectively; the first judgment unit is used to judge whether the currently acquired battery pack parameters are at a normal level. If so, no operation is performed; if not, an abnormal response measure is executed. The weight calculation module includes a sample acquisition unit and a weight calculation unit; the sample acquisition unit is used to acquire a number of battery pack data samples; the weight calculation unit is used to calculate the weight coefficients of each battery pack parameter. The fault prediction module includes a fault prediction unit and a second judgment unit. The fault prediction unit is used to take the battery pack parameter time series set combination and the weight coefficients of each battery pack parameter as the input of the fault prediction model and output the fault prediction result; the second judgment unit is used to execute a warning response measure when the fault prediction result indicates that a fault exists.

Images