Battery early-stage micro short circuit fault detection method and device, computer equipment and medium

Abstract

The invention relates to the technical field of battery energy storage, and discloses a battery early-stage micro short circuit fault detection method and device, computer equipment and a medium, and the method comprises the steps: respectively obtaining the charging process sample data of a normal battery and a fault battery; extracting fault characteristic features from the charging process sample data, and dividing the fault characteristic features into a training set and a test set; training a preset initial boosting tree model by adopting the training set to obtain a trained to-be-tested boosting tree model; testing the boosting tree model to be tested by adopting the test set to obtain a tested battery fault prediction model; and acquiring voltage and current time sequence data when the to-be-predicted battery is charged, and inputting the voltage and current time sequence data when the to-be-predicted battery is charged into the battery fault prediction model so as to detect the fault state of the battery. The method focuses on pre-extraction of fault features, model parameters are small and easy to deploy, and normal operation of the battery pack is ensured through early diagnosis of the micro short circuit fault of the battery.

Description

Technical Field

[0001] The present invention relates to the technical field of battery energy storage, and in particular to a method, device, computer equipment and medium for detecting early micro-short circuit faults of batteries. Background Art

[0002] In order to meet the large-scale power demand of industry and commerce, existing energy storage battery packs are often composed of a large number of high-capacity single cells connected in series or in parallel. When a single cell in the battery pack fails, the energy storage and discharge efficiency of the entire battery pack will be reduced, and it will not be able to meet the expected power demand, affecting the normal operation of industry and commerce. At the same time, the faulty battery may cause safety problems such as overheating, leakage, and even explosion, especially in large-scale energy storage systems, which may cause serious property losses and safety accidents.

[0003] At present, the methods for battery fault diagnosis include directly analyzing the faulty battery voltage to extract fault features. Fault features include sample entropy, etc., and judgment is made by setting thresholds. This method is simple to implement, but is easily affected by noise and misjudgment, and is only applicable to sudden serious faults, and cannot detect early micro-short circuit faults in battery packs. Alternatively, a deep neural network model is used to automatically extract features from the voltage, temperature and other information of normal batteries and faulty batteries to estimate the probability of battery failure. This method usually requires a large amount of data as support, which is costly. At the same time, the scale of the neural network model is too large and cannot achieve rapid diagnosis. In addition, in terms of series battery pack failure, the consistency of the battery is used to judge the faulty battery in the battery pack, which is usually reflected in the inconsistency of voltage. The full charge voltage of the faulty battery is low. The voltage inconsistency under full charge conditions is used to diagnose the faulty battery in the actual battery use environment. The applicability is limited. Due to the differences in battery production, the sensor measurement accuracy is greatly affected, and the early micro-short circuit fault of the battery cannot be accurately determined. Summary of the invention

[0004] In view of this, the present invention provides a method, device, computer equipment and medium for detecting early micro-short circuit faults of batteries, so as to solve the problem in the prior art that the amount of data required is large and difficult to obtain, resulting in the inability to accurately and quickly diagnose early micro-short circuit faults of batteries.

[0005] In a first aspect, the present invention provides a method for detecting an early micro short circuit fault of a battery, the method comprising:

[0006] Obtain charging process sample data of a normal battery and a faulty battery respectively;

[0007] Extract fault characteristic features from charging process sample data, and divide the fault characteristic features into a training set and a test set;

[0008] The training set is used to train the preset initial boosting tree model to obtain a trained boosting tree model to be tested;

[0009] The test set is used to test the boosted tree model to be tested, and a battery fault prediction model that has been tested is obtained;

[0010] The voltage and current time series data of the battery to be predicted when it is charged are obtained, and the voltage and current time series data of the battery to be predicted when it is charged are input into a battery fault prediction model to detect the fault state of the battery.

[0011] The battery early micro-short circuit fault detection method provided by the present invention obtains sample data of the charging process of normal batteries and faulty batteries respectively, extracts features with fault characteristics based on the charging process sample data, uses a boosted tree model for training, and uses the trained fault detection model to detect battery faults, which can effectively detect battery early micro-short circuit faults and is less affected by noise. At the same time, compared with deep learning neural networks, this method requires less training data, focuses on pre-extracting fault features, and has smaller model parameters and is easy to deploy. By diagnosing battery micro-short circuit faults at an early stage, the normal operation of the battery pack can be ensured, and the performance of the entire system can be avoided from being affected by individual faulty batteries. It also helps to maintain the consistency of the battery pack, reduce the loss of other batteries, extend the service life of the battery, and reduce the cost of maintenance and replacement, solving the problem in the prior art that the amount of data required is large and difficult to obtain, resulting in the inability to accurately and quickly diagnose battery early micro-short circuit faults.

[0012] In an optional implementation, extracting fault characteristic features from charging process sample data and dividing the fault characteristic features into a training set and a test set includes:

[0013] The voltage standard deviation, voltage skewness, average charging current, left and right endpoints of the curve of the ratio of battery capacity change to battery voltage change, peak value of the curve of the ratio of battery capacity change to battery voltage change, and slope of the voltage line at the end of charging for 6 minutes are extracted from the sample data of the charging process as fault characteristic features;

[0014] The fault characteristic features are divided into training set and test set according to the preset numerical ratio.

[0015] In an optional implementation, the charging process sample data includes voltage series data, current series data and time series data;

[0016] The voltage standard deviation, voltage skewness, average charging current, left and right endpoints of the curve of the ratio of battery capacity change to battery voltage change, peak value of the curve of the ratio of battery capacity change to battery voltage change, and slope of the voltage line at the end of charging for 6 minutes are extracted from the charging process sample data as fault characteristic features including:

[0017] Based on the voltage series data and time series data, the voltage standard deviation, voltage skewness, battery voltage change, battery capacity change and the slope of the voltage line at the end of charging for 6 minutes are calculated;

[0018] Calculate the average charging current based on the current series data and the time series data;

[0019] A curve of the ratio of the battery capacity change to the battery voltage change is constructed based on the battery voltage change and the battery capacity change, and the left and right endpoints and the peak value of the ratio curve are obtained.

[0020] The battery early micro-short circuit fault detection method provided by the present invention extracts the voltage standard deviation, voltage skewness, average charging current, the left and right endpoint values ​​of the curve of the ratio of battery capacity change to battery voltage change, the peak value of the curve of the ratio of battery capacity change to battery voltage change, and the slope of the voltage line at the end of charging for 6 minutes from the charging process sample data as fault characteristic features, adds more obvious fault characteristic features, further highlights the difference between normal batteries and faulty batteries through the battery capacity change and battery voltage change, and divides the fault characteristic features into a training set and a test set, providing a data basis for the training of the boosting tree model.

[0021] In an optional implementation, the preset initial boosting tree model is constructed in the following manner:

[0022] Set the hyperparameters of the boosted tree model, including the tree depth, learning rate, and number of training rounds;

[0023] The grid search algorithm is used to calculate the tree depth, learning rate and number of training rounds under different parameter range combinations, and corresponding different results are obtained;

[0024] The best result is determined from different results, and the depth, learning rate and number of training rounds of the tree under different parameter range combinations corresponding to the best result are taken as the optimal parameters. The initial boosting tree model is constructed based on the optimal parameters.

[0025] The battery early micro-short circuit fault detection method provided by the present invention sets the hyperparameters of the boosted tree model, and the hyperparameters include the depth of the tree, the learning rate and the number of training rounds; a grid search algorithm is used to calculate the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations to obtain corresponding different results; the best result is determined from the different results, and the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations corresponding to the best result are used as the optimal parameters, thereby realizing the construction of an initial boosted tree model based on the optimal parameters, and providing initial model conditions for the subsequent training of the boosted tree model.

[0026] In an optional implementation, the initial boosting tree model is trained using a training set, and the trained boosting tree model to be tested is obtained, including:

[0027] The training set is input into the initial boosting tree model for iterative calculation, and in each round of iteration, a normal battery classification prediction value or a faulty battery classification prediction value is output;

[0028] The residual between the normal battery classification prediction value or the faulty battery classification prediction value and its corresponding true value is calculated, and the residual is used as the input data for the next round of iteration. The iteration ends when the iteration result reaches the preset objective function, and the boosted tree model obtained in the last iteration is used as the boosted tree model to be tested.

[0029] The battery early micro-short circuit fault detection method provided by the present invention inputs a training set into an initial boosting tree model for iterative calculation, and in each round of iteration, outputs a normal battery classification prediction value or a faulty battery classification prediction value; calculates the residual between the normal battery classification prediction value or the faulty battery classification prediction value and its corresponding true value, and uses the residual as input data for the next round of iteration to iterate until the iteration result reaches a preset objective function and the iteration ends, thereby achieving the goal of using the boosting tree model obtained in the last iteration as the boosting tree model to be tested. The obtained boosting tree model to be tested is small in scale and has the advantages of small parameters and easy deployment. The residual is continuously fitted during the training process, thereby continuously correcting errors and gradually improving the prediction accuracy of the model.

[0030] In an optional implementation, the preset objective function is calculated using the following formula:

[0031]

[0032] Where: Obj(t) is the objective function of the tth round, is the loss function, l is the binary cross entropy loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample in the t-1 round of the boosted tree model, f t (x i ) is the current t-th boosted tree model for sample x i The predicted value of the sample is the fault characteristic feature, Ω(f t ) is a regularization term used to control the complexity of the boosted tree model, including the number of leaf nodes and weights. T is the number of leaf nodes in the boosted tree, w j is the weight of the jth leaf, and γ and λ are regularization coefficients.

[0033] The battery early micro-short circuit fault detection method provided by the present invention finds the optimal splitting point achieved by the fault characteristic features through the loss function and regularization coefficient in the preset objective function, minimizes the objective function, and makes the trained boosting tree model have high prediction accuracy, efficient calculation, good regularization ability, and prevents overfitting during the training process.

[0034] In an optional implementation, the test set is used to test the boosted tree model to be tested, and the battery fault prediction model after the test is obtained includes:

[0035] Input the test set into the boosted tree model to be tested to obtain the test output value;

[0036] The test output value is compared with the true value corresponding to the test set, and the prediction accuracy of the boosting tree model to be tested is calculated. If the prediction accuracy of the boosting tree model to be tested reaches a preset threshold, the boosting tree model to be tested is determined as a battery fault prediction model.

[0037] The battery early micro-short circuit fault detection method provided by the present invention obtains a test output value by inputting a test set into a boosted tree model to be tested; the test output value is compared with a true value corresponding to the test set, and the prediction accuracy of the boosted tree model to be tested is calculated. If the prediction accuracy of the boosted tree model to be tested reaches a preset threshold, the boosted tree model to be tested is determined as a battery fault prediction model, so that the battery fault prediction model has good detection capability and effectively improves the accuracy of model detection.

[0038] In a second aspect, the present invention provides a battery early micro short circuit fault detection device, the device comprising:

[0039] A sample data acquisition module, used to respectively acquire sample data of the charging process of a normal battery and a faulty battery;

[0040] A fault characteristic feature extraction module is used to extract fault characteristic features from charging process sample data and divide the fault characteristic features into a training set and a test set;

[0041] A training module is used to train a preset initial boosting tree model using a training set to obtain a trained boosting tree model to be tested;

[0042] A testing module, used to test the boosted tree model to be tested using a test set to obtain a battery fault prediction model that has been tested;

[0043] The prediction module is used to obtain the voltage and current time series data of the battery to be predicted when it is charged, and input the voltage and current time series data of the battery to be predicted when it is charged into the battery fault prediction model to detect the fault state of the battery.

[0044] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the battery early micro-short circuit fault detection method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the battery early micro-short circuit fault detection method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 is a flow chart of a method for detecting an early micro-short circuit fault of a battery according to an embodiment of the present invention;

[0048] Figure 2 is a flow chart of another method for detecting early micro-short circuit faults of a battery according to an embodiment of the present invention;

[0049] Figure 3 is a flow chart of another method for detecting early micro-short circuit faults of a battery according to an embodiment of the present invention;

[0050] Figure 4 is a flow chart of another method for detecting early micro-short circuit faults of a battery according to an embodiment of the present invention;

[0051] Figure 5 is a structural block diagram of a battery early micro-short circuit fault detection device according to an embodiment of the present invention;

[0052] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0054] According to an embodiment of the present invention, an embodiment of a method for detecting early micro-short circuit faults in a battery is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0055] In this embodiment, a method for detecting early micro short circuit faults of a battery is provided, which can be used in mobile terminals, such as mobile phones, central processing units, tablet computers, etc. Figure 1 is a flow chart of a method for detecting an early micro short circuit fault of a battery according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0056] Step S101, respectively obtaining charging process sample data of a normal battery and a faulty battery.

[0057] Specifically, when the battery is in an early micro-short circuit state, the voltage standard deviation and voltage skewness characteristics of normal batteries and faulty batteries are not obvious enough, and are also easily affected by noise. Therefore, more obvious dQ (battery capacity change) / dV (battery voltage change) related characteristics are added. The above-mentioned voltage standard deviation, voltage skewness and dQ (battery capacity change) / dV (battery voltage change) related characteristics are all related to the voltage data and current data during the charging process and the charging time. Therefore, the sample data of the charging process of normal batteries and faulty batteries may include voltage series data, current series data and time series data.

[0058] Step S102: extract fault characteristic features from the charging process sample data, and divide the fault characteristic features into a training set and a test set.

[0059] Specifically, due to the leakage current of the early micro-short circuit fault battery, the voltage performance of the normal battery and the faulty battery are different. During the same charging current, the faulty battery takes a longer time to reach the same voltage as the normal battery. Therefore, under the same current, the voltage standard deviation and voltage skewness of the normal battery and the faulty battery are different. When the battery is in an early micro-short circuit state, the above characteristics are not obvious enough and are also easily affected by noise. Therefore, more obvious dQ (battery capacity change) / dV (battery voltage change) related features are added. The voltage standard deviation, voltage skewness and dQ (battery capacity change) / dV (battery voltage change) extracted from the charging process sample data can be used as fault characteristic features. The preset numerical ratio is set according to the actual situation, and there is no specific restriction again. The preset numerical ratio of this embodiment divides the fault characteristic features into training set and test set according to the ratio of 7:3.

[0060] Step S103, using the training set to train the preset initial boosting tree model to obtain a trained boosting tree model to be tested.

[0061] Specifically, the boosted tree model is also called the XGBoost (eXtreme Gradient Boosting) model. The XGBoost model is an efficient gradient boosting framework that builds a powerful prediction model by integrating multiple weak learners (decision trees). The basic idea of ​​XGBoost is that gradient boosting gradually builds the model in an iterative manner, and each step attempts to reduce the prediction error of the previous model. Specifically, in each round of iteration, the newly added model will fit the residual of the previous model (that is, the difference between the true value and the predicted value), thereby continuously correcting errors and gradually improving the prediction accuracy. The boosted tree model of the last iteration is used as the boosted tree model to be tested.

[0062] Step S104, using the test set to test the boosted tree model to be tested, to obtain a battery fault prediction model that has completed the test.

[0063] Specifically, in order to evaluate the prediction accuracy of the boosted tree model to be tested, the test set is input into the boosted tree model to be tested, and the output is the sum of the prediction values ​​of all trees. When the output accuracy is greater than the preset accuracy threshold, it is determined that the model has good detection ability, and the final battery fault prediction model is obtained.

[0064] Step S105, obtaining the voltage and current time series data of the battery to be predicted when it is charged, and inputting the voltage and current time series data of the battery to be predicted when it is charged into a battery fault prediction model to detect the fault state of the battery.

[0065] Specifically, after the training is completed and passed the evaluation, the XGBoost-based battery fault prediction model is applied to the actual fault scenario for prediction, and the battery fault state is judged based on the voltage and current time series data of the battery to be predicted during charging, where the true value of a normal battery is 0 and the true value of a faulty battery is 1.

[0066] The battery early micro-short circuit fault detection method provided in this embodiment obtains sample data of the charging process of normal batteries and faulty batteries respectively, extracts features with fault characteristics based on the charging process sample data, uses the boosting tree model for training, and uses the trained fault detection model to detect battery faults, which can effectively detect battery early micro-short circuit faults and is less affected by noise. At the same time, compared with deep learning neural networks, this method requires less training data, focuses on pre-extracting fault features, and has small model parameters and is easy to deploy. By diagnosing battery micro-short circuit faults at an early stage, the normal operation of the battery pack can be ensured to avoid affecting the performance of the entire system due to individual faulty batteries. It also helps to maintain the consistency of the battery pack, reduce the loss of other batteries, extend the service life of the battery, and reduce maintenance and replacement costs, solving the problem in the prior art that the amount of data required is large and difficult to obtain, resulting in the inability to accurately and quickly diagnose battery early micro-short circuit faults.

[0067] In this embodiment, a method for detecting early micro short circuit faults of a battery is provided, which can be used in mobile terminals, such as mobile phones, central processing units, tablet computers, etc. Figure 2 is a flow chart of a method for detecting an early micro short circuit fault of a battery according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0068] Step S201, respectively obtain the charging process sample data of the normal battery and the faulty battery. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0069] Step S202: extract fault characteristic features from the charging process sample data, and divide the fault characteristic features into a training set and a test set.

[0070] Specifically, the above step S202 includes:

[0071] Step S2021, extracting the voltage standard deviation, voltage skewness, average charging current, left and right endpoints of the curve of the ratio of battery capacity change to battery voltage change, the peak value of the curve of the ratio of battery capacity change to battery voltage change, and the slope of the voltage line at the end of charging for 6 minutes from the charging process sample data as fault characteristic features.

[0072] In some optional implementations, the sample data includes voltage sequence data, current sequence data, and time series data. Taking a charge and discharge test of a certain model of commercial lithium-ion battery as an example, the charging includes a fault simulation charge and discharge cycle and a normal charge and discharge cycle, and the battery charge and discharge data are recorded. In the battery charging process data, and the state of charge SOC (State of Charge, SOC) is in the range of 20% to 90%, the voltage sequence data V, current sequence data I, and time series data T of the normal battery and the faulty battery are extracted respectively.

[0073] Among them, Figure 4 As shown, the voltage sequence data V=[V1,V2,V 3,...... ,V t ]; current sequence data I = [I1, I2, I 3,...... ,I t ], time series data T = [T1, T2, T 3,...... ,T t ], t is the length of time series data T.

[0074] The above step S2021 includes:

[0075] Step a1, based on the voltage series data and the time series data, calculate the voltage standard deviation, voltage skewness, battery voltage change, battery capacity change and the slope of the voltage line at the end of charging for 6 minutes.

[0076] The calculation formulas for voltage standard deviation, voltage skewness and the slope of the voltage line at the end of charging for 6 minutes are as follows:

[0077]

[0078]

[0079]

[0080] Among them, V std is the voltage standard deviation, V skew is the voltage deflection, k is the slope of the voltage line at the end of charging for 6 minutes, V i is each voltage value in the voltage sequence data, i=1,2,3,......,t, where Battery capacity change dQ = Q t2 -Q t1 , battery voltage change dV = V t2 -V t1 , Q t is the battery capacity.

[0081] Step a2, calculating the average charging current based on the current series data and the time series data.

[0082] Specifically, the calculation formula for the average charging current is as follows:

[0083]

[0084] Among them, I mean is the average charging current, I i are the current values ​​in the current series data, i=1,2,3,......,t.

[0085] Step a3, constructing a curve of the ratio of the battery capacity change to the battery voltage change based on the battery voltage change and the battery capacity change, and obtaining the left and right endpoints and the peak value of the ratio curve.

[0086] Specifically, the ratio between the battery capacity change and the battery voltage change at each charging time t is calculated, and the ratio obtained at each time is connected into a ratio curve, and the left and right endpoints and the peak value of the ratio curve are obtained.

[0087] The formulas for the left and right endpoints and the peak value of the curve of the ratio of battery capacity change to battery voltage change are as follows:

[0088]

[0089]

[0090]

[0091] in, is the left endpoint of the curve of the ratio of battery capacity change to battery voltage change, is the right endpoint of the curve of the ratio of battery capacity change to battery voltage change, is the peak value of the curve of the ratio of battery capacity change to battery voltage change. dQ1 and dV1 are the capacitance change and voltage change at the first charging moment, respectively. dQ t , dV t are the capacitance change and voltage change at the tth charging moment respectively.

[0092] Step S2022: divide the fault characteristic features into a training set and a test set according to a preset numerical ratio.

[0093] Specifically, the fault characteristic feature data set D = {X1, X2, X3, ..., X n}, where each sample X i (i=1,2,…,n) is an array containing the above 7 fault characteristic features, which can be further expressed as: n is the total number of samples, each sample X i There is a corresponding label true value. For the fault characteristic feature of a normal battery, the label value is 0, and for the fault characteristic feature of a faulty battery, the label value is 1.

[0094] According to the total number of samples, the fault characteristic feature data set is randomly divided into training set D in a ratio of 7:3. train ={X1,X2,X3,…,X n1} and test set D test ={X1,X2,X3,…,X n2},in

[0095] Step S203: Use the training set to train the preset initial boosting tree model to obtain a trained boosting tree model to be tested. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0096] Step S204: Use the test set to test the boosted tree model to obtain a battery fault prediction model that has been tested. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0097] Step S205, obtaining the voltage and current time series data of the battery to be predicted when it is charged, and inputting the voltage and current time series data of the battery to be predicted when it is charged into the battery fault prediction model to detect the fault state of the battery. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0098] The battery early micro-short circuit fault detection method provided in this embodiment extracts the voltage standard deviation, voltage skewness, average charging current, the left and right endpoint values ​​of the curve of the ratio of battery capacity change to battery voltage change, the peak value of the curve of the ratio of battery capacity change to battery voltage change, and the slope of the voltage line at the end of charging for 6 minutes as fault characteristic features from sample data, adds more obvious fault characteristic features, further highlights the difference between normal batteries and faulty batteries through the battery capacity change and battery voltage change, and divides the fault characteristic features into training sets and test sets, providing a data basis for the training of the boosting tree model.

[0099] In this embodiment, a method for detecting early micro short circuit faults of a battery is provided, which can be used in mobile terminals, such as mobile phones, central processing units, tablet computers, etc. Figure 3 is a flow chart of a method for detecting an early micro short circuit fault of a battery according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0100] Step S301, respectively obtain the charging process sample data of the normal battery and the faulty battery. Figure 1 Step S201 of the illustrated embodiment will not be described in detail here.

[0101] Step S302: extract fault characteristic features from the charging process sample data, and divide the fault characteristic features into a training set and a test set. Figure 1 Step S202 of the illustrated embodiment will not be described in detail here.

[0102] Step S303: Use the training set to train the preset initial boosting tree model to obtain a trained boosting tree model to be tested.

[0103] Specifically, the above step S303 includes:

[0104] Step S3031, set the hyperparameters of the boosted tree model (hereinafter referred to as the XGBoost model), the hyperparameters include the depth of the tree, the learning rate and the number of training rounds, use the grid search algorithm to calculate the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations, and obtain corresponding different results, determine the best result from the different results, and use the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations corresponding to the best result as the optimal parameters, and build an initial boosted tree model based on the optimal parameters.

[0105] Specifically, build an XGBoost model and set the hyperparameters of the XGBoost model. The hyperparameters include max_depth: the depth of the tree, which affects the model complexity and overfitting risk. learning_rate: learning rate, which controls the step size of each iteration. n_estimators: the number of boosted trees, that is, the number of training rounds. Automatically find the optimal parameters through the gridSearchCV (grid search) method, set the parameter range, where the tree depth max_depth range is 3-10 layers, the learning rate learning_rate range is 0.05-0.15, and the number of boosted trees range is 100-1000, calculate the results under different parameter combinations, and finally save the parameter settings that give the best results on the data. The tree depth, learning rate and number of training rounds under different parameter range combinations corresponding to the best results are used as the optimal parameters, and the initial boosted tree model is constructed based on the optimal parameters.

[0106] Step S3032, input the training set into the initial boosting tree model for iterative calculation, and in each round of iteration, output the normal battery classification prediction value or the faulty battery classification prediction value.

[0107] Specifically, the training set D train ={X1,X2,X3,…,Xn1 Input the XGBoost model for training and perform binary classification tasks. The XGBoost model inputs the fault characteristic features in the training set and outputs the classification prediction value results of the battery normal state 0 or fault state 1.

[0108] Step S3033, calculate the residual between the normal battery classification prediction value or the faulty battery classification prediction value and its corresponding true value, and use the residual as the input data for the next round of iteration until the iteration result reaches the preset objective function, and the iteration ends. The boosted tree model obtained in the last iteration is used as the boosted tree model to be tested.

[0109] Specifically, the preset objective function is calculated using the following formula:

[0110]

[0111] Where: Obj(t) is the objective function of the tth round, is the loss function, l is the binary cross entropy loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample in the t-1 round of the boosted tree model, f t (x i ) is the current t-th boosted tree model for sample x i The predicted value of the sample is the fault characteristic feature, Ω(f t ) is a regularization term used to control the complexity of the boosted tree model, including the number of leaf nodes and weights. T is the number of leaf nodes in the boosted tree, w j is the weight of the jth leaf, and γ and λ are regularization coefficients.

[0112] In some optional implementations, the above step S3033 includes:

[0113] Step b1, initialize the XGBoost model: build a constant model, usually the mean of the target variable.

[0114] Step b2, calculate the residual between the current XGBoost model prediction value and the true value.

[0115] Step b3, using the residual as a new target variable, constructing a new tree, that is, using a greedy algorithm or an approximate histogram algorithm to find the best feature splitting point to minimize the above objective function.

[0116] Step b4: for each fault characteristic feature in the training set, calculate its contribution to the residual, and select the best fault characteristic feature as the basis for splitting.

[0117] Step b5, finding the best splitting point on the selected fault characteristic feature to minimize the objective function.

[0118] Step b6, divide the data into left and right subsets according to the optimal split point, and continue to recursively build the next layer of the tree until the stopping condition is met (for example, the stopping condition is reaching the maximum tree depth).

[0119] Step b7: After the tree is built, backward traversal pruning is performed to remove leaf nodes that contribute less to the overall model to simplify the XGBoost model structure and prevent overfitting.

[0120] Step b8, add the newly constructed tree to the existing model after weighting it at a certain learning rate to form a new comprehensive model. When the preset maximum number of iterations is reached or the model performance is no longer significantly improved, stop the iteration and obtain the final XGBoost model, that is, the boosted tree model to be tested.

[0121] Step S304: Use the test set to test the boosted tree model to be tested, and obtain a battery fault prediction model that has completed the test.

[0122] Specifically, Figure 4 As shown, the above step S304 includes:

[0123] Step S3041, input the test set into the boosting tree model to be tested to obtain the test output value.

[0124] Specifically, the test set D test ={X1,X2,X3,…,X n2}Input into the boosting tree model to be tested, evaluate the performance of the boosting tree model to be tested, and obtain the test output value.

[0125] Step S3042, compare the test output value with the true value corresponding to the test set, calculate the prediction accuracy of the boosting tree model to be tested, and if the prediction accuracy of the boosting tree model to be tested reaches a preset threshold, the boosting tree model to be tested is determined as a battery fault prediction model.

[0126] Specifically, all test output values ​​are compared with the true values ​​corresponding to the test set, and the prediction accuracy of the boosted tree model to be tested is calculated. The preset threshold of this embodiment is set to 90%. When the output accuracy is greater than 90%, it is determined that the boosted tree model to be tested has a good detection ability, and the boosted tree model to be tested is determined as a battery fault prediction model.

[0127] Step S305, obtaining the voltage and current time series data of the battery to be predicted when it is charged, and inputting the voltage and current time series data of the battery to be predicted when it is charged into the battery fault prediction model to detect the fault state of the battery. Figure 1Step S205 of the illustrated embodiment will not be described in detail here.

[0128] The battery early micro-short circuit fault detection method provided in this embodiment is to input the training set into the initial boosting tree model for iterative calculation, and in each round of iteration, output the normal battery classification prediction value or the faulty battery classification prediction value; calculate the residual between the normal battery classification prediction value or the faulty battery classification prediction value and its corresponding true value, and use the residual as the input data of the next round of iteration for iteration, until the iteration ends when the iteration result reaches the preset objective function, so as to realize that the boosting tree model obtained by the last iteration is used as the boosting tree model to be tested, and the obtained boosting tree model to be tested is small in scale, and the model has the advantages of small parameters and easy deployment. The residual is continuously fitted during the training process, so as to continuously correct errors and gradually improve the prediction accuracy of the model. The optimal splitting point reached by the fault characteristic feature is found through the loss function and regularization coefficient in the preset objective function, and the objective function is minimized, so that the trained boosting tree model has high prediction accuracy, high computational efficiency, and good regularization ability, and prevents overfitting during the training process.

[0129] In this embodiment, a battery early micro-short circuit fault detection device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0130] This embodiment provides a battery early micro short circuit fault detection device, such as Figure 5 As shown, including:

[0131] The sample data acquisition module 501 is used to respectively acquire sample data of the charging process of a normal battery and a faulty battery.

[0132] The fault characteristic feature extraction module 502 is used to extract fault characteristic features from the charging process sample data and divide the fault characteristic features into a training set and a test set.

[0133] The training module 503 is used to train the preset initial boosting tree model using the training set to obtain a trained boosting tree model to be tested.

[0134] The testing module 504 is used to test the boosted tree model to be tested using the test set to obtain a battery fault prediction model that has completed the test.

[0135] The prediction module 505 is used to obtain the voltage and current time series data of the battery to be predicted when it is charged, and input the voltage and current time series data of the battery to be predicted when it is charged into the battery fault prediction model to detect the fault state of the battery.

[0136] In some optional implementations, the fault characteristic feature extraction module 502 includes:

[0137] The extraction unit is used to extract the voltage standard deviation, voltage skewness, average charging current, left and right endpoint values ​​of the curve of the ratio of battery capacity change to battery voltage change, the peak value of the curve of the ratio of battery capacity change to battery voltage change, and the slope of the voltage connection line at the end of charging for 6 minutes from the sample data of the charging process as fault characteristic features.

[0138] The data set division unit is used to divide the fault characteristic features into a training set and a test set according to a preset numerical ratio.

[0139] In some optional implementations, the charging process sample data includes voltage series data, current series data and time series data; and the extraction unit includes:

[0140] The first calculation subunit is used to calculate the voltage standard deviation, voltage skewness, battery voltage change, battery capacity change and the slope of the voltage line at the charging end for 6 minutes based on the voltage series data and the time series data.

[0141] The second calculation subunit is used to calculate the average charging current based on the current series data and the time series data.

[0142] The third calculation subunit is used to construct a curve of the ratio of the battery capacity change to the battery voltage change based on the battery voltage change and the battery capacity change, and obtain the left and right endpoints and peak value of the ratio curve.

[0143] In some optional implementations, the training module 503 includes:

[0144] The initial boosting tree model construction unit is used to set the hyperparameters of the boosting tree model, including the depth of the tree, the learning rate and the number of training rounds; the grid search algorithm is used to calculate the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations to obtain corresponding different results; the best result is determined from the different results, and the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations corresponding to the best result are used as the optimal parameters, and the initial boosting tree model is constructed based on the optimal parameters.

[0145] The input-output unit is used to input the training set into the initial boosting tree model for iterative calculation, and in each round of iteration, output the normal battery classification prediction value or the faulty battery classification prediction value.

[0146] The training iteration unit is used to calculate the residual between the normal battery classification prediction value or the faulty battery classification prediction value and its corresponding true value, and use the residual as the input data for the next round of iteration until the iteration result reaches the preset objective function. The iteration ends and the boosted tree model obtained in the last iteration is used as the boosted tree model to be tested.

[0147] The preset objective function is calculated using the following formula:

[0148]

[0149] Where: Obj(t) is the objective function of the tth round, is the loss function, l is the binary cross entropy loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample in the t-1 round of the boosted tree model, f t (x i ) is the current t-th boosted tree model for sample x i The predicted value of the sample is the fault characteristic feature, Ω(f t ) is a regularization term used to control the complexity of the boosted tree model, including the number of leaf nodes and weights. T is the number of leaf nodes in the boosted tree, w j is the weight of the jth leaf, and γ and λ are regularization coefficients.

[0150] In some optional implementations, the testing module 504 includes:

[0151] The test unit is used to input the test set into the boosted tree model to be tested to obtain the test output value.

[0152] A comparison unit is used to compare the test output value with the true value corresponding to the test set, and calculate the prediction accuracy of the boosting tree model to be tested. If the prediction accuracy of the boosting tree model to be tested reaches a preset threshold, the boosting tree model to be tested is determined as a battery fault prediction model.

[0153] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0154] The battery early micro-short circuit fault detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0155] The embodiment of the present invention also provides a computer device having the above Figure 6 The battery early micro short circuit fault detection device shown.

[0156] See also Figure 6 , Figure 6 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.

[0157] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0158] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0159] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0160] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0161] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0162] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0163] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0164] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for detecting early micro short circuit faults of a battery, characterized in that: The method comprises: Obtain charging process sample data of a normal battery and a faulty battery respectively; Extracting fault characteristic features from the charging process sample data, and dividing the fault characteristic features into a training set and a test set; The training set is used to train the preset initial boosting tree model to obtain a trained boosting tree model to be tested; Using a test set to test the boosted tree model to be tested, to obtain a battery fault prediction model that has been tested; The voltage and current time series data of the battery to be predicted when it is charged are obtained, and the voltage and current time series data of the battery to be predicted when it is charged are input into the battery fault prediction model to detect the fault state of the battery.

2. The method according to claim 1, characterized in that: Extracting fault characteristic features from the charging process sample data and dividing the fault characteristic features into a training set and a test set includes: Extracting the voltage standard deviation, voltage skewness, average charging current, left and right endpoints of the curve of the ratio of battery capacity change to battery voltage change, the peak value of the curve of the ratio of battery capacity change to battery voltage change, and the slope of the voltage line at the end of charging for 6 minutes from the charging process sample data as fault characteristic features; The fault characteristic features are divided into a training set and a test set according to a preset numerical ratio.

3. The method according to claim 1 or 2, characterized in that: The charging process sample data includes voltage series data, current series data and time series data; The voltage standard deviation, voltage skewness, average charging current, left and right endpoints of the curve of the ratio of battery capacity change to battery voltage change, peak value of the curve of the ratio of battery capacity change to battery voltage change, and slope of the voltage line at the end of charging for 6 minutes are extracted from the charging process sample data as fault characteristic features, including: Calculate the voltage standard deviation, voltage skewness, battery voltage change, battery capacity change and the slope of the voltage line at the end of charging for 6 minutes based on the voltage series data and time series data; Calculate an average charging current based on the current series data and the time series data; A curve of the ratio of the battery capacity change to the battery voltage change is constructed based on the battery voltage change and the battery capacity change, and the left and right endpoints and the peak value of the ratio curve are obtained.

4. The method according to claim 1, characterized in that The preset initial boosting tree model is constructed in the following way: Setting hyperparameters of the boosted tree model, including tree depth, learning rate, and number of training rounds; The grid search algorithm is used to calculate the tree depth, learning rate and number of training rounds under different parameter range combinations, and corresponding different results are obtained; The best result is determined from different results, the depth of the tree, the learning rate and the number of training rounds under different parameter range combinations corresponding to the best result are taken as the optimal parameters, and an initial boosted tree model is constructed based on the optimal parameters.

5. The method according to claim 4, characterized in that The initial boosting tree model is trained using the training set, and the trained boosting tree model to be tested includes: Inputting the training set into the initial boosting tree model for iterative calculation, and outputting a normal battery classification prediction value or a faulty battery classification prediction value in each round of iteration; The residual between the normal battery classification prediction value or the faulty battery classification prediction value and its corresponding true value is calculated, and the residual is used as the input data for the next round of iteration. The iteration ends when the iteration result reaches the preset objective function, and the boosted tree model obtained in the last iteration is used as the boosted tree model to be tested.

6. The method according to claim 5, characterized in that The preset objective function is calculated using the following formula: Where: Obj(t) is the objective function of the tth round, is the loss function, l is the binary cross entropy loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample in the t-1 round of the boosted tree model, f t (x i ) is the current t-th boosted tree model for sample x i The predicted value of the sample is the fault characteristic feature, Ω(f t ) is a regularization term used to control the complexity of the boosted tree model, including the number of leaf nodes and weights. T is the number of leaf nodes in the boosted tree, w j is the weight of the jth leaf, and γ and λ are regularization coefficients.

7. The method according to claim 1, characterized in that The step of testing the boosted tree model to be tested by using a test set to obtain a battery fault prediction model after the test is completed includes: Inputting the test set into the boosted tree model to be tested to obtain a test output value; The test output value is compared with the true value corresponding to the test set, and the prediction accuracy of the boosting tree model to be tested is calculated. If the prediction accuracy of the boosting tree model to be tested reaches a preset threshold, the boosting tree model to be tested is determined as a battery fault prediction model.

8. A battery early micro short circuit fault detection device, characterized in that: The device comprises: A sample data acquisition module, used to respectively acquire sample data of the charging process of a normal battery and a faulty battery; A fault characteristic feature extraction module is used to extract fault characteristic features from the charging process sample data and divide the fault characteristic features into a training set and a test set; A training module, used to train a preset initial boosting tree model using the training set to obtain a trained boosting tree model to be tested; A testing module, used to test the boosted tree model to be tested using a test set to obtain a battery fault prediction model that has been tested; The prediction module is used to obtain the voltage and current time series data of the battery to be predicted when it is charged, and input the voltage and current time series data of the battery to be predicted when it is charged into the battery fault prediction model to detect the fault state of the battery.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the battery early micro-short circuit fault detection method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the battery early micro-short circuit fault detection method according to any one of claims 1 to 7.

Images