Sodium-ion battery fault multi-stage early warning method and device

Abstract

The invention provides a sodium ion battery fault multistage early warning method and device, and the method comprises the steps: obtaining multi-dimensional physical quantities such as voltage, temperature, stress and gas concentration, and the change rate of the multi-dimensional physical quantities, constructing a decision matrix through a 1-9 scale method, and calculating a weight vector. After an input data set is constructed, a classification decision tree model is generated by using a Gini index, and a fusion decision tree model is generated according to a weight vector distribution priority. By inputting the multi-dimensional physical quantity and the change rate thereof, the model can perform multi-stage early warning on faults of the sodium-ion battery in real time, and potential problems can be effectively found in advance. According to the method, multi-dimensional physical quantity monitoring, decision matrix construction and a fusion decision tree model are combined, and efficient early warning is achieved.

Description

Technical Field

[0001] The present application relates to the field of battery early warning, and in particular to a multi-level early warning method and device for sodium ion battery failure. Background Art

[0002] During the charging, discharging and storage process, batteries may encounter various adverse factors such as overcharging, over-discharging, excessive temperature, mechanical stress, etc. These factors may trigger complex chemical reactions inside the battery, leading to serious safety accidents such as thermal runaway, fire, and even explosion.

[0003] Real-time monitoring through a battery management system (BMS) has become an effective safety precaution. Traditional BMSs mostly use a single-threshold monitoring approach, setting multiple safety thresholds (such as voltage, temperature, and pressure). When battery operating parameters exceed these thresholds, the system triggers an alarm or takes automatic protective measures.

[0004] However, traditional single-threshold monitoring methods have limitations when faced with complex battery state changes. Battery overheating may be accompanied by voltage fluctuations, and monitoring only a single physical quantity, temperature or voltage, cannot capture this complex failure mode. Summary of the Invention

[0005] The purpose of this application is to overcome the defects in the above-mentioned prior art and provide a multi-level early warning method and device for sodium ion battery failure.

[0006] The present application provides a multi-level early warning method for sodium ion battery failure, comprising:

[0007] Obtaining multidimensional physical quantities of a sodium ion battery during operation, the multidimensional physical quantities including voltage, temperature, stress, and gas concentration;

[0008] Calculating a multidimensional physical quantity change rate of the multidimensional physical quantity, wherein the multidimensional physical quantity change rate includes a voltage change rate, a temperature change rate, a stress change rate, and a gas concentration change rate;

[0009] Constructing a decision matrix including pairwise comparison relationships of the multidimensional physical quantity change rates by a 1-9 scaling method;

[0010] Calculating a weight vector of each of the multidimensional physical quantity change rates according to the decision matrix;

[0011] constructing an input data set including a time series queue according to the multidimensional physical quantity and the rate of change of the multidimensional physical quantity;

[0012] Based on the input data set, recursively dividing feature thresholds by the Gini index to generate a classification decision tree model;

[0013] Prioritizing the multidimensional physical quantity change rates in the decision tree model according to the weight vector to generate a fusion decision tree model;

[0014] Based on the fusion decision tree model, a multi-level warning signal of a sodium ion battery failure is generated by inputting the multi-dimensional physical quantity and the rate of change of the multi-dimensional physical quantity.

[0015] Optionally, the weight vector is:

[0016] The voltage change rate weight w1 = 0.6088, the temperature change rate weight w2 = 0.1012, the stress change rate weight w3 = 0.2583, and the gas concentration change rate weight w4 = 0.0317.

[0017] Optionally, the method further includes: performing a consistency check on the weight vector using a random consistency indicator.

[0018] Optionally, the time series queue of the input data set is:

[0019] M=[V;T;P;Rho]

[0020] Wherein V, T, P, and Rho are the original time series data including voltage, temperature, stress, and gas concentration and the corresponding time series of the rate of change of the multi-dimensional physical quantity.

[0021] Optionally, the classification of the fault multi-level warning signal includes: normal, overheat, overcharge and fault.

[0022] The present application also provides a sodium ion battery fault multi-level early warning device, comprising:

[0023] An acquisition module is used to acquire multidimensional physical quantities during operation of the sodium ion battery, wherein the multidimensional physical quantities include voltage, temperature, stress, and gas concentration;

[0024] a calculation module for calculating a multidimensional physical quantity change rate of the multidimensional physical quantity, wherein the multidimensional physical quantity change rate includes a voltage change rate, a temperature change rate, a stress change rate, and a gas concentration change rate;

[0025] A matrix module constructs a decision matrix containing pairwise comparison relationships of the multi-dimensional physical quantity change rates through a 1-9 scaling method;

[0026] A vector module, calculating a weight vector of each of the multidimensional physical quantity change rates according to the decision matrix;

[0027] A data module constructs an input data set including a time series queue according to the multidimensional physical quantity and the rate of change of the multidimensional physical quantity;

[0028] A model module, based on the input data set, recursively divides the feature threshold by the Gini index to generate a classification decision tree model;

[0029] a fusion module, which assigns priorities to the multidimensional physical quantity change rates in the decision tree model according to the weight vector to generate a fusion decision tree model;

[0030] The early warning module generates a multi-level early warning signal of a sodium ion battery failure by inputting the multi-dimensional physical quantity and the rate of change of the multi-dimensional physical quantity based on the fusion decision tree model.

[0031] Optionally, the weight vector is:

[0032] The voltage change rate weight w1 = 0.6088, the temperature change rate weight w2 = 0.1012, the stress change rate weight w3 = 0.2583, and the gas concentration change rate weight w4 = 0.0317.

[0033] Optionally, the vector module further includes: performing a consistency check on the weight vector using a random consistency indicator.

[0034] Optionally, the time series queue of the input data set is:

[0035] M=[V;T;P;Rho]

[0036] Wherein V, T, P, and Rho are the original time series data including voltage, temperature, stress, and gas concentration and the corresponding time series of the rate of change of the multi-dimensional physical quantity.

[0037] Optionally, the classification of the fault multi-level warning signal includes: normal, overheat, overcharge and fault.

[0038] The beneficial effects of this application are:

[0039] The present application provides a multi-level early warning method for sodium-ion battery failure, comprising: obtaining multidimensional physical quantities during operation of the sodium-ion battery, the multidimensional physical quantities including voltage, temperature, stress, and gas concentration; calculating multidimensional physical quantity change rates of the multidimensional physical quantities, the multidimensional physical quantity change rates including voltage change rate, temperature change rate, stress change rate, and gas concentration change rate; constructing a decision matrix containing pairwise comparison relationships of the multidimensional physical quantity change rates using a 1-9 scaling method; calculating a weight vector for each multidimensional physical quantity change rate based on the decision matrix; constructing an input data set including a time series queue based on the multidimensional physical quantities and the multidimensional physical quantity change rates; generating a classification decision tree model based on the input data set by recursively partitioning feature thresholds using a Gini index; assigning priorities to the multidimensional physical quantity change rates in the decision tree model based on the weight vectors to generate a fused decision tree model; and generating a multi-level early warning signal for sodium-ion battery failure based on the fused decision tree model by inputting the multidimensional physical quantities and the multidimensional physical quantity change rates. This application obtains the multi-dimensional physical quantities and the rate of change of the multi-dimensional physical quantities during the operation of the sodium-ion battery, constructs a decision matrix and calculates the weight vector, and then constructs a fusion decision tree model to achieve real-time multi-level early warning of sodium-ion battery failures, which can effectively detect potential failures in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a schematic diagram of the multi-level warning process for sodium ion battery failure in this application;

[0041] Figure 2 It is a schematic diagram of the fluctuation of physical quantities over time at each stage of this application;

[0042] Figure 3 It is a schematic diagram of AHP warning values and multidimensional parameters in this application;

[0043] Figure 4 It is a partially enlarged schematic diagram of the AHP warning value and multidimensional parameters in this application. DETAILED DESCRIPTION

[0044] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that various forms of implementation of the present disclosure are not limited to the embodiments set forth herein. Rather, the embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0045] Please refer to Figure 1 As shown, the present application provides a multi-level early warning method for sodium ion battery failure, comprising:

[0046] S101, obtaining multidimensional physical quantities during operation of a sodium ion battery, wherein the multidimensional physical quantities include voltage, temperature, stress, and gas concentration;

[0047] S102, calculating the multidimensional physical quantity change rate of the multidimensional physical quantity, where the multidimensional physical quantity change rate includes the voltage change rate, the temperature change rate, the stress change rate, and the gas concentration change rate;

[0048] like Figure 2 As shown in the figure, each physical quantity in the multidimensional physical quantity shows a significant trend in the comparison between normal operating conditions and overcharge conditions, accompanied by a certain degree of noise fluctuation, which can be regarded as high-frequency noise. Therefore, it is necessary to preprocess the data signals of the multidimensional physical quantities and then perform differentiation processing on the preprocessed signals to obtain the rate of change of each physical quantity.

[0049] The data signal preprocessing includes low-pass filtering the data signal with a cutoff frequency of 2000 Hz, such as a Butterworth low-pass filter or a Chebyshev low-pass filter. After low-pass filtering, the voltage of the timing queue is represented as V, the temperature is represented as T, the battery wall stress is represented as P (i.e., pressure), and the gas concentration at the battery tab is represented as Rho.

[0050] According to the differential formula, the rate of change of the corresponding physical quantity is calculated as follows:

[0051]

[0052] S103, constructing a decision matrix including pairwise comparison relationships of the multi-dimensional physical quantity change rates by a 1-9 scaling method;

[0053] The above multi-dimensional physical quantities are constructed as a time sequence queue matrix and expressed as:

[0054] M=[V,T,P,Rho]

[0055] According to the experience of battery system and actual application data (the rate of change of multi-dimensional physical quantities), a hierarchical structure model of the analytic hierarchy process (AHP) is constructed.

[0056] The change rates of each physical quantity (voltage, temperature, stress, gas concentration) are compared in pairs to obtain the relative importance weights of each physical quantity. The expression of the decision matrix A is as follows:

[0057]

[0058] Among them, each element a in the decision matrix A ij The subscripts of represent the rate of change of each physical quantity; and a ij Indicates the importance of physical quantity i to physical quantity j.

[0059] Each factor is compared with each other, and the weight of each criterion layer to the target layer is determined, and the weight value uses Santy's 1-9 scaling method.

[0060] S104, calculating a weight vector of each multi-dimensional physical quantity change rate according to the decision matrix;

[0061] Specifically, the evaluation index of the AHP model weight uses the consistency index CI and the consistency ratio CR, which are calculated as follows:

[0062]

[0063] Wherein, RI represents a constant predetermined according to the order A of the decision matrix.

[0064] The RI corresponding to the 4th-order decision matrix mentioned above is 0.9; λ max It represents the maximum value of the eigenvalue of the decision matrix and is calculated by the following formula

[0065] Aw=λ max w

[0066] Where w is the optimal weight vector of the AHP model. In order to ensure that the sum of the weights is 1, the eigenvector needs to be normalized:

[0067]

[0068] Example: According to the corresponding ratio of the rate of change of physical quantities, the decision matrix is determined as:

[0069]

[0070] An example of the decision matrix shown is as follows:

[0071] <![CDATA[w1]]> <![CDATA[w2]]> <![CDATA[w3]]> <![CDATA[w4]]> CI CR Numerical 0.6088 0.1012 0.2583 0.0317 0.1 0.11

[0072] The AHP method is used to calculate the weight of the change rate of each physical quantity, obtain the weight ratio of each physical quantity of the sodium-ion battery relative to other physical quantities, and perform a consistency test.

[0073] S105, constructing an input data set including a time series queue according to the multidimensional physical quantity and the rate of change of the multidimensional physical quantity;

[0074] The physical quantity data and change rate data of sodium ion batteries are used as inputs to the decision tree model.

[0075] The input of the decision tree model constructs the time series queue matrix M, M = [V, T, P, Rho], and constructs the data set: D = {(x i ,y i )}, where x i∈M is the feature of the i-th sample, y i ∈{1, 2, …, C} is the label, and the number of categories is C.

[0076] S106: Based on the input data set, recursively divide the feature threshold by the Gini index to generate a classification decision tree model;

[0077] At each node, a feature x and its threshold t are selected to divide the dataset into two subsets DL and DR, including samples less than t and greater than or equal to t, respectively.

[0078] Use the Gini index Gini(D) to measure the quality of the division:

[0079]

[0080] Among them, p k is the proportion of samples in dataset D that belong to category k.

[0081] For a feature x j And threshold t, the Gini index after division is:

[0082]

[0083] Recursive partitioning is achieved by minimizing the Gini index until the depth of the tree reaches a preset value or the number of subsets is reduced to a certain level, thereby completing the construction of the CART decision tree.

[0084] S107: assigning priorities to the multidimensional physical quantity change rates in the decision tree model according to the weight vector to generate a fusion decision tree model;

[0085] The weights obtained by the AHP method are combined with the decision tree model to form a weighted fusion decision tree model.

[0086] The weighted fusion decision tree model can not only classify the battery status, such as classification based on the original data of physical quantities, but also accurately identify the fault type based on the weights of the change rates of different physical quantities, thereby facilitating the linkage of fire and explosion prevention under different strategies.

[0087] like Figure 3 As shown in the figure, by training the weighted fusion decision tree, early warning signals of battery status are output, including normal, overheating, overcharge, fault, etc.

[0088] The optimal weight vector calculated by AHP is introduced, and a comprehensive early warning indicator is constructed by weighted fusion. It is then added to the above decision tree model. The AHP early warning is expressed as:

[0089] η=w1·V′+w2·T′+w3·P′+w4·Rho′

[0090] The weights obtained from the AHP model are used as the feature weights of the decision tree, and these weights are incorporated into the decision tree construction process to enhance the decision tree's recognition of the importance of each factor.

[0091] The decision tree model is trained using the training set to establish a classification model for battery failure warning.

[0092] S108. Based on the fusion decision tree model, generate a multi-level warning signal for sodium ion battery failure by inputting the multi-dimensional physical quantity and the rate of change of the multi-dimensional physical quantity.

[0093] During the use of sodium-ion batteries, the system monitors multidimensional data of various physical quantities in real time and uses a weighted fusion decision tree model to predict and warn of faults. A multi-level warning mechanism corresponding to the type classification instance is also established.

[0094] According to the operating status of the sodium-ion battery and the early warning signals output by the weighted fusion decision tree model, the fault warning is divided into multiple levels, thereby achieving multi-level response in system monitoring.

[0095] Specifically, based on the changes in various physical quantities and the classification results of the weighted fusion decision tree, the following warning levels are defined:

[0096] Level 1 Warning (Normal): The battery status is normal and no action is required.

[0097] Level 2 warning (minor warning): The battery operating status is abnormal, but it does not immediately affect safety.

[0098] For example, temperature, pressure, concentration, etc. are at the lower end of the warning threshold, such as the battery temperature is slightly elevated but still within the safe range.

[0099] Level 3 Warning (Serious Warning): The battery has obvious safety hazards and can easily turn into a very dangerous thermal runaway state. Close attention must be paid and preliminary safety measures must be taken.

[0100] For example, if the battery voltage approaches the overcharge / overdischarge threshold, the temperature approaches the overheat warning value, or the pressure and gas concentration are too high, serious consequences such as loss of control and fire may occur.

[0101] Level 4 warning (out-of-control prediction): The system predicts that the sodium-ion battery will experience a thermal runaway event, and combined with the fact that various physical quantities are approaching the threshold, the battery needs to be disconnected immediately and emergency treatment must be carried out.

[0102] like Figure 4As shown in the figure, when the model identifies a potential failure risk in the sodium-ion battery, it automatically issues an early warning signal based on the weighted fusion decision tree to prompt the user or initiate corresponding safety protection measures, effectively supporting the implementation of fire prevention, explosion prevention and threshold linkage design solutions.

[0103] The present application also provides a sodium ion battery fault multi-level early warning device, comprising:

[0104] An acquisition module is used to acquire multidimensional physical quantities during operation of the sodium ion battery, wherein the multidimensional physical quantities include voltage, temperature, stress, and gas concentration;

[0105] a calculation module for calculating a multidimensional physical quantity change rate of the multidimensional physical quantity, wherein the multidimensional physical quantity change rate includes a voltage change rate, a temperature change rate, a stress change rate, and a gas concentration change rate;

[0106] A matrix module constructs a decision matrix containing pairwise comparison relationships of the multi-dimensional physical quantity change rates through a 1-9 scaling method;

[0107] A vector module, calculating a weight vector of each of the multidimensional physical quantity change rates according to the decision matrix;

[0108] A data module constructs an input data set including a time series queue according to the multidimensional physical quantity and the rate of change of the multidimensional physical quantity;

[0109] A model module, based on the input data set, recursively divides the feature threshold by the Gini index to generate a classification decision tree model;

[0110] a fusion module, which assigns priorities to the multidimensional physical quantity change rates in the decision tree model according to the weight vector to generate a fusion decision tree model;

[0111] The early warning module generates a multi-level early warning signal of a sodium ion battery failure by inputting the multi-dimensional physical quantity and the rate of change of the multi-dimensional physical quantity based on the fusion decision tree model.

[0112] Furthermore, the weight vector is:

[0113] The voltage change rate weight w1 = 0.6088, the temperature change rate weight w2 = 0.1012, the stress change rate weight w3 = 0.2583, and the gas concentration change rate weight w4 = 0.0317.

[0114] Furthermore, the vector module further includes: performing consistency check on the weight vector using a random consistency indicator.

[0115] Furthermore, the time series queue of the input data set is:

[0116] M=[V;T;P;Rho]

[0117] Wherein V, T, P, and Rho are the original time series data including voltage, temperature, stress, and gas concentration and the corresponding time series of the rate of change of the multi-dimensional physical quantity.

[0118] Furthermore, the classification of the fault multi-level warning signal includes: normal, overheat, overcharge and fault.

[0119] The above description of the embodiments is intended to facilitate understanding and application of the present invention by those skilled in the art. It will be readily apparent to those skilled in the art that various modifications to the above embodiments can be made, and the general principles described herein can be applied to other embodiments without requiring inventive effort. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the present disclosure are intended to fall within the scope of protection of the present invention.

Claims

1. A multi-level early warning method for sodium ion battery failure, characterized in that: include: Obtaining multidimensional physical quantities of a sodium ion battery during operation, the multidimensional physical quantities including voltage, temperature, stress, and gas concentration; Calculating a multidimensional physical quantity change rate of the multidimensional physical quantity, wherein the multidimensional physical quantity change rate includes a voltage change rate, a temperature change rate, a stress change rate, and a gas concentration change rate; Constructing a decision matrix including pairwise comparison relationships of the multidimensional physical quantity change rates by a 1-9 scaling method; Calculating a weight vector of each of the multidimensional physical quantity change rates according to the decision matrix; constructing an input data set including a time series queue according to the multidimensional physical quantity and the rate of change of the multidimensional physical quantity; Based on the input data set, recursively dividing feature thresholds by the Gini index to generate a classification decision tree model; Prioritizing the multidimensional physical quantity change rates in the decision tree model according to the weight vector to generate a fusion decision tree model; Based on the fusion decision tree model, a multi-level warning signal of a sodium ion battery failure is generated by inputting the multi-dimensional physical quantity and the rate of change of the multi-dimensional physical quantity.

2. A sodium ion battery fault multi-level early warning method according to claim 1, characterized in that: The weight vector is: The voltage change rate weight w1 = 0.6088, the temperature change rate weight w2 = 0.1012, the stress change rate weight w3 = 0.2583, and the gas concentration change rate weight w4 = 0.0317.

3. A sodium ion battery fault multi-level early warning method according to claim 2, characterized in that: Also includes: A random consistency indicator is used to perform consistency check on the weight vector.

4. A sodium ion battery fault multi-level early warning method according to claim 1, characterized in that: The time series queue of the input dataset is: M=[V;T;P;Rho] Wherein V, T, P, and Rho are the original time series data including voltage, temperature, stress, and gas concentration and the corresponding time series of the rate of change of the multi-dimensional physical quantity.

5. A sodium ion battery fault multi-level early warning method according to claim 1, characterized in that: The classification of the fault multi-level warning signal includes: normal, overheat, overcharge and fault.

6. A multi-level early warning device for sodium ion battery failure, characterized in that: include: An acquisition module is used to acquire multidimensional physical quantities during operation of the sodium ion battery, wherein the multidimensional physical quantities include voltage, temperature, stress, and gas concentration; a calculation module for calculating a multidimensional physical quantity change rate of the multidimensional physical quantity, wherein the multidimensional physical quantity change rate includes a voltage change rate, a temperature change rate, a stress change rate, and a gas concentration change rate; A matrix module constructs a decision matrix containing pairwise comparison relationships of the multi-dimensional physical quantity change rates through a 1-9 scaling method; A vector module, calculating a weight vector of each of the multidimensional physical quantity change rates according to the decision matrix; A data module constructs an input data set including a time series queue according to the multidimensional physical quantity and the rate of change of the multidimensional physical quantity; A model module, based on the input data set, recursively divides the feature threshold by the Gini index to generate a classification decision tree model; a fusion module, which assigns priorities to the multidimensional physical quantity change rates in the decision tree model according to the weight vector to generate a fusion decision tree model; The early warning module generates a multi-level early warning signal of a sodium ion battery failure by inputting the multi-dimensional physical quantity and the rate of change of the multi-dimensional physical quantity based on the fusion decision tree model.

7. A sodium ion battery fault multi-level early warning device according to claim 6, characterized in that: The weight vector is: The voltage change rate weight w1 = 0.6088, the temperature change rate weight w2 = 0.1012, the stress change rate weight w3 = 0.2583, and the gas concentration change rate weight w4 = 0.0317.

8. A sodium ion battery fault multi-level early warning device according to claim 7, characterized in that: The vector module further includes: performing consistency check on the weight vector using a random consistency indicator.

9. A sodium ion battery fault multi-level early warning device according to claim 6, characterized in that: The time series queue of the input dataset is: M=[V;T;P;Rho] Wherein V, T, P, and Rho are the original time series data including voltage, temperature, stress, and gas concentration and the corresponding time series of the rate of change of the multi-dimensional physical quantity.

10. A sodium ion battery fault multi-level early warning device according to claim 6, characterized in that: The classification of the fault multi-level warning signal includes: normal, overheat, overcharge and fault.

Images