技术领域Technical Field
本发明属于动力电池健康状态监测技术领域，具体涉及一种融合健康衰退和一致性的动力电池风险评估方法。The present invention belongs to the technical field of power battery health status monitoring, and specifically relates to a power battery risk assessment method integrating health decline and consistency.
背景技术Background technique
现阶段，电动汽车上采用的动力电池系统呈现出高电压与高容量的发展趋势，电池组中单体成组数量和规模随之显著提高，但受目前电池制造工艺、材料差异等因素影响，导致电池组内各单体之间常存在电压、容量、内阻等参数的不一致性，而这些不一致问题在当前还不具备完全消除的条件。由于动力电池的不一致性问题对电池性能和寿命均容易产生不利影响，因此对动力电池的健康状态和一致性进行精确及时的评估具有非常重要的现实意义。现有技术中对于电池健康状态评估的手段主要可分为：基于实验分析的、基于模型的和基于数据驱动的SOH估计方法三大类，这其中基于实验分析的SOH估计方法，虽能够对锂离子电池衰退的机理进行较全面的分析，SOH估计精度高，但受限于实验设备、环境、研究周期及成本等问题，不能较好地在实车中应用；基于模型的方法对建模的依赖性较强，而模型参数的辨识过程则存在计算成本高、耗时长的缺点；基于数据驱动的方法虽相较于前两类方法在精确性和计算开销方面优势明显，但大多现有技术所采用的分析参数仍相对单一，对动力电池不一致性的评估还不够全面，同时常采用的如阻抗等的特征参数在实车上也难以准确测量，也限制了此类方法在实车应用时的真实效果。At present, the power battery system used in electric vehicles shows a development trend of high voltage and high capacity, and the number and scale of monomer groups in the battery pack have increased significantly. However, due to the current battery manufacturing process, material differences and other factors, there are often inconsistencies in parameters such as voltage, capacity, and internal resistance between the monomers in the battery pack, and these inconsistencies are not yet completely eliminated. Since the inconsistency of power batteries can easily have an adverse effect on battery performance and life, it is of great practical significance to accurately and timely evaluate the health status and consistency of power batteries. The existing methods for evaluating the health status of batteries can be mainly divided into three categories: SOH estimation methods based on experimental analysis, model-based and data-driven. Among them, the SOH estimation method based on experimental analysis can conduct a more comprehensive analysis of the mechanism of lithium-ion battery degradation and has high SOH estimation accuracy, but it is limited by experimental equipment, environment, research cycle and cost, and cannot be well applied in actual vehicles; the model-based method has a strong dependence on modeling, and the identification process of model parameters has the disadvantages of high computational cost and long time consumption; although the data-driven method has obvious advantages in accuracy and computational overhead compared with the first two methods, the analysis parameters used in most existing technologies are still relatively single, and the evaluation of power battery inconsistency is not comprehensive enough. At the same time, the commonly used characteristic parameters such as impedance are difficult to measure accurately on actual vehicles, which also limits the actual effect of such methods in actual vehicle applications.
发明内容Summary of the invention
鉴于此，针对本领域中存在的技术问题，本发明提供了一种融合健康衰退和一致性的动力电池风险评估方法，具体包括以下步骤：In view of this, in order to solve the technical problems existing in the art, the present invention provides a power battery risk assessment method integrating health decline and consistency, which specifically includes the following steps:
步骤一、从若干新能源汽车获取同类动力电池系统的实车运行数据，并利用实车运行数据中包括时间Time、总电流I、总电压U、SOC、车辆状态S、温度T、充电状态等的参数建立新能源汽车运行数据矩阵P：Step 1: Obtain the actual vehicle operation data of the same power battery system from several new energy vehicles, and use the parameters including time Time, total current I, total voltage U, SOC, vehicle state S, temperature T, charging state, etc. in the actual vehicle operation data to establish the new energy vehicle operation data matrix P:
P＝[Time,U,S,a1,a2,...,an]P＝[Time,U,S,a1 ,a2 ,...,an ]
其中，an表示第n个可用数据项对应的数据向量；Where an represents the data vector corresponding to the nth available data item;
步骤二、对所建立的新能源运行数据矩阵P中所包含的数据异常、数据缺失及错误数据执行包括线性插补和数据删除的数据预处理；Step 2: Perform data preprocessing including linear interpolation and data deletion on the data anomalies, data missing and erroneous data contained in the established new energy operation data matrix P;
步骤三、根据各新能源汽车实际使用时充电行为统计分析驾驶员的用车习惯，并在此基础上设定充电片段样本的选取规则；利用新能源汽车运行数据矩阵P获得切分后的各充电片段，并基于选取规则筛选出相应的充电片段样本；Step 3: Analyze the driver's driving habits according to the charging behaviors of each new energy vehicle during actual use, and set the selection rules of charging segment samples on this basis; use the new energy vehicle operation data matrix P to obtain the segmented charging segments, and select the corresponding charging segment samples based on the selection rules;
步骤四、利用充电片段样本计算固定SOC区间的容量增量和能量增量；提取不同时刻单体电压值与电池组SOC的乘积来建立反映电压一致性的电压表征参数序列，计算电池组在相应充电片段的基准电压表征序列，以及各单体电压表征参数向量相对该基准电压表征序列的曼哈顿距离矩阵；提取充电片段样本中电池探针温度数据，计算电池组探针温度在相应充电片段的基准温度表征序列，以及各探针温度向量相对该基准温度表征序列的曼哈顿距离矩阵；计算充电片段样本对应的充电功率表征参数序列以反映充电功率一致性，计算相应片段的基准充电功率序列，以及充电功率表征参数序列相对该基准充电功率序列的曼哈顿距离矩阵；Step 4: Calculate the capacity increment and energy increment of a fixed SOC interval using the charging segment samples; extract the product of the single cell voltage value and the battery pack SOC at different times to establish a voltage characterization parameter sequence reflecting voltage consistency, calculate the reference voltage characterization sequence of the battery pack in the corresponding charging segment, and the Manhattan distance matrix of each single cell voltage characterization parameter vector relative to the reference voltage characterization sequence; extract the battery probe temperature data in the charging segment samples, calculate the reference temperature characterization sequence of the battery pack probe temperature in the corresponding charging segment, and the Manhattan distance matrix of each probe temperature vector relative to the reference temperature characterization sequence; calculate the charging power characterization parameter sequence corresponding to the charging segment samples to reflect the charging power consistency, calculate the reference charging power sequence of the corresponding segment, and the Manhattan distance matrix of the charging power characterization parameter sequence relative to the reference charging power sequence;
将得到的容量增量、能量增量、各曼哈顿距离矩阵中离散程度最大单体所分别对应的表征参数序列Z分数进行相关性分析，从中筛选出最终用于评估电池组SOH与一致性的特征参数；The obtained capacity increment, energy increment, and Z-scores of the characterization parameter sequences corresponding to the monomers with the largest degree of dispersion in each Manhattan distance matrix are subjected to correlation analysis, from which the characteristic parameters ultimately used to evaluate the SOH and consistency of the battery pack are selected;
步骤五、基于CRITIC权重法来建立电池组SOH与一致性评估模型，并融合熵权法对CRITIC权重法进行改进，在计算充电片段风险评分时考虑其熵值，由此得到各充电片段的风险评估序列；Step 5: Establish a battery pack SOH and consistency assessment model based on the CRITIC weight method, and integrate the entropy weight method to improve the CRITIC weight method. Consider its entropy value when calculating the risk score of the charging segment, thereby obtaining the risk assessment sequence of each charging segment;
步骤六、基于风险评估序列确定正常车辆与故障车辆的风险评分分布情况，并基于正常车辆的所述分布确定相应的安全风险识别阈值；基于该阈值对实车风险进行预警决策。Step 6: Determine the risk score distribution of normal vehicles and faulty vehicles based on the risk assessment sequence, and determine the corresponding safety risk identification threshold based on the distribution of normal vehicles; and make early warning decisions on real vehicle risks based on the threshold.
进一步地，步骤一中针对时间Time、总电压U、车辆状态S和温度T数据具体采用以下向量展开形式：Furthermore, in step 1, the following vector expansion form is specifically used for the time Time, total voltage U, vehicle state S and temperature T data:
Time＝[time1,time2,time3,···,timej]TTime＝[time1 , time2 , time3 ,···,timej ]T
其中，timei表示在数据采集时间i(i＝1,2,3,...,j)；Wherein, timei represents the data collection time i (i = 1, 2, 3, ..., j);
其中，k是电池单体的数量，j是对应与不同采样时间的行数，Ut,h表示电池单体h(h＝1,2,3,...,k)在时刻t(t＝1,2,3,...,j)的电压值；Wherein, k is the number of battery cells, j is the number of rows corresponding to different sampling times, and Ut,h represents the voltage value of battery cell h (h = 1, 2, 3, ..., k) at time t (t = 1, 2, 3, ..., j);
其中，It、vt、soct、rt分别表示电池组在时刻t(t＝1,2,3,...,j)的总电流值、车辆行驶速度、电池组SOC、绝缘电阻值；Wherein, It , vt , soct , and rt represent the total current value of the battery pack at time t (t=1, 2, 3, ..., j), the vehicle speed, the battery pack SOC, and the insulation resistance value, respectively;
其中，m是温度探针的数量，j是对应于不同采样时间的行数，Tt,l表示电池组温度探针l(l＝1,2,3,...,m)在时刻t(t＝1,2,3,...,j)的温度值。Wherein, m is the number of temperature probes, j is the number of rows corresponding to different sampling times, and Tt,l represents the temperature value of battery pack temperature probe l (l=1,2,3,...,m) at time t (t=1,2,3,...,j).
进一步地，步骤三中通过对充电行为和驾驶员习惯的分析，选择包含60％～70％SOC区间的充电片段样本。Furthermore, in step three, by analyzing the charging behavior and driver habits, a charging segment sample including a SOC range of 60% to 70% is selected.
进一步地，步骤四中针对充电片段样本的60％～70％ SOC区间，利用以下安时积分计算固定SOC区间的容量增量：Furthermore, in step 4, for the 60% to 70% SOC interval of the charging segment sample, the capacity increment in the fixed SOC interval is calculated using the following ampere-hour integral:
其中，I表示动力电池实时充电电流，TSOCstart和TSOCend分别代表动力电池充电SOC区间开始和结束的时间；Where I represents the real-time charging current of the power battery, TSOCstart and TSOCend represent the start and end time of the power battery charging SOC interval respectively;
针对充电片段样本的相同SOC区间，利用以下安时积分计算固定SOC区间的能量增量：For the same SOC interval of the charging segment sample, the energy increment of the fixed SOC interval is calculated using the following ampere-hour integration:
利用充电片段样本的单体电压数据以及电池组SOC计算所述电压表征参数序列M：The voltage characterization parameter sequence M is calculated using the single cell voltage data of the charging segment sample and the battery pack SOC:
式中，Mh表示电池单体h的表征向量，ut,h表示电池单体h在时刻t(t＝1,2,3,...,j)的电压值，j是对应不同采样时间的行数，soctut,h表示电池单体h在t时刻电压值与此刻电池SOC的乘积，即电压表征参数值；Wherein, Mh represents the characterization vector of battery cell h,ut,h represents the voltage value of battery cell h at time t (t=1, 2, 3, ..., j), j is the number of rows corresponding to different sampling times, soct ut,h represents the product of the voltage value of battery cell h at time t and the battery SOC at this moment, that is, the voltage characterization parameter value;
计算基准电压表征序列Mdatum为：The reference voltage characterization sequence Mdatum is calculated as:
式中，表示时刻t(t＝1,2,3,...,j)的所有k个单体电池电压表征参数的平均值；In the formula, represents the average value of the voltage characteristic parameters of all k single cells at time t (t=1, 2, 3, ..., j);
计算充电片段内各电池单体电压表征向量相对于基准电压表征序列的曼哈顿距离矩阵DM：Calculate the Manhattan distance matrix DM of the voltage characterization vector of each battery cell in the charging segment relative to the reference voltage characterization sequence:
DM＝[DM,1 DM,2 … DM,h … DM,k]DM =[DM,1 DM,2 ... DM,h ... DM,k ]
式中，|soctut,h-socdatumut,datum|表示单体h在时刻t相对于基准表征值的曼哈顿距离；In the formula, |soct ut,h -socdatum ut,datum | represents the Manhattan distance of monomer h relative to the benchmark representation value at time t;
电压离散程度最大单体所对应的表征参数序列Z分数基于以下公式计算：The Z score of the characterization parameter sequence corresponding to the monomer with the largest voltage dispersion is calculated based on the following formula:
式中，表示各个电池单体在本充电片段内相对于基准电压表征序列的曼哈顿距离平均值，σDM表示曼哈顿距离的标准差，δDM,max表示相应的Z分数；In the formula, represents the average Manhattan distance of each battery cell relative to the reference voltage characterization sequence in this charging segment, σDM represents the standard deviation of the Manhattan distance, and δDM,max represents the corresponding Z score;
提取充电片段样本中的探针温度时间序列Th并计算基准温度表征序列Tdatum：Extract the probe temperature time seriesTh in the charging segment sample and calculate the reference temperature characterization seriesTdatum :
式中，表示时刻t(t＝1,2,3,...,j)的所有探针温度的平均值；In the formula, represents the average value of all probe temperatures at time t (t = 1, 2, 3, ..., j);
计算充电片段样本内各探针温度时间序列相对于温度基准表征序列的曼哈顿距离矩阵DT：Calculate the Manhattan distance matrix DT of each probe temperature time series in the charging segment sample relative to the temperature reference characterization sequence:
DT＝[DT,1 DT,2 … DT,h … DT,l]DT =[DT,1 DT,2 ... DT,h ... DT,l ]
其中，DT,h表示探针h相对于探针温度基准表征序列的曼哈顿距离，其计算方式如下：Where DT,h represents the Manhattan distance of probe h relative to the probe temperature benchmark characterization sequence, which is calculated as follows:
式中，tt,h表示探针h在时刻t(t＝1,2,3,...,j)的温度值，|tt,h-tt,datum|表示探针h在时刻t相对于温度基准序列的曼哈顿距离；Where tt,h represents the temperature value of probe h at time t (t = 1, 2, 3, ..., j), |tt,h -tt,datum | represents the Manhattan distance of probe h at time t relative to the temperature reference sequence;
温度离散程度最大单体所对应的表征参数序列Z分数基于以下公式计算：The Z score of the characterization parameter sequence corresponding to the monomer with the largest temperature dispersion is calculated based on the following formula:
式中，表示各探针温度在本充电片段内相对于基准温度序列的曼哈顿距离平均值，σDT表示此曼哈顿距离的标准差，δDT,max表示相应的Z分数；In the formula, represents the average Manhattan distance of each probe temperature relative to the reference temperature sequence in this charging segment, σDT represents the standard deviation of this Manhattan distance, and δDT,max represents the corresponding Z score;
基准充电功率序列基于以下公式计算：The base charging power sequence is calculated based on the following formula:
式中，Ph表示单体h的充电功率时间序列；Where,Ph represents the charging power time series of cell h;
通过与电压一致性和温度一致性相同的过程计算曼哈顿距离矩阵DP：The Manhattan distance matrix DP is calculated by the same process as for voltage consistency and temperature consistency:
DP＝[DP,1 DP,2 … DP,h … DP,k]DP =[DP,1 DP,2 ... DP,h ... DP,k ]
并利用以下公式计算得到充电功率离散程度最大单体所对应的表征参数序列Z分数：The Z score of the characterization parameter sequence corresponding to the monomer with the largest degree of charging power dispersion is calculated using the following formula:
表示上述曼哈顿距离的平均值，σDP表示曼哈顿距离的标准差，δDP,max表示相应的Z分数。 represents the mean of the above Manhattan distance, σDP represents the standard deviation of the Manhattan distance, and δDP,max represents the corresponding Z score.
进一步地，步骤四中对得到的容量增量、能量增量、各曼哈顿距离矩阵中离散程度最大单体所分别对应的表征参数序列Z分数，具体利用最大信息系数法(MaximalInformation Coefficient,MIC)进行相关性分析，具体公式如下：Furthermore, in step 4, the capacity increment, energy increment, and Z score of the characterization parameter sequence corresponding to the monomer with the largest degree of dispersion in each Manhattan distance matrix are respectively analyzed by using the maximum information coefficient method (Maximal Information Coefficient, MIC). The specific formula is as follows:
式中，X和Y是两个随机变量，B是MIC算法的变量，I[X；Y]是随机变量X和Y的互信息，定义为联合分布p(x,y)与边缘分布p(x)p(y)相对熵，其计算公式如下：Where X and Y are two random variables, B is the variable of the MIC algorithm, and I[X;Y] is the mutual information of random variables X and Y, which is defined as the relative entropy of the joint distribution p(x,y) and the marginal distribution p(x)p(y). The calculation formula is as follows:
式中，p(x,y)是两个随机变量(X,Y)的联合分布，p(x)、p(y)为随机变量(X,Y)边缘分布；Where p(x,y) is the joint distribution of two random variables (X,Y), p(x) and p(y) are the marginal distributions of the random variables (X,Y);
通过相关性分析最终确定固定SOC区间容量增量、电压与温度一致性分别对应的Z分数作为特征参数。Through correlation analysis, the Z scores corresponding to the capacity increment, voltage and temperature consistency in a fixed SOC interval are finally determined as characteristic parameters.
进一步地，步骤五中电池组SOH与一致性评估模型建立与改进的具体包括：Furthermore, the establishment and improvement of the battery pack SOH and consistency assessment model in step 5 specifically include:
首先提取新能源汽车每个充电片段对应的特征参数，得到动力电池风险评价矩阵A：First, the characteristic parameters corresponding to each charging segment of new energy vehicles are extracted to obtain the power battery risk assessment matrix A:
式中，A1,A2,A3分别表示容量增量、电压一致性特征参数、温度一致性特征参数序列,aij代表充电片段i(i＝1,2,…,n)的第j(j＝1,2,3)项表征参数值；Wherein, A1 , A2 , A3 represent the capacity increment, voltage consistency characteristic parameter, and temperature consistency characteristic parameter sequence, respectively, and aij represents the jth (j＝1,2,3) characterization parameter value of charging segment i (i＝1,2,…,n);
对各特征参数执行归一化处理，将容量增量数据处理为负向指标，电压和温度一致性特征参数处理为正向指标：Normalize each characteristic parameter, treat the capacity increment data as a negative indicator, and treat the voltage and temperature consistency characteristic parameters as positive indicators:
正向指标处理公式：Positive indicator processing formula:
负向指标处理公式：Negative indicator processing formula:
式中，aij为正向/负向指标处理后所得到的充电片段i的第j项特征参数值；Where aij is the jth characteristic parameter value of charging segment i obtained after positive/negative index processing;
计算数据各特征参数的信息承载量：Calculate the information carrying capacity of each characteristic parameter of the data:
式中，表示第j项特征参数的均值，Sj是第j项特征参数的标准差，用以承载该特征参数的波动性；In the formula, represents the mean of the j-th characteristic parameter,Sj is the standard deviation of the j-th characteristic parameter, which is used to carry the volatility of the characteristic parameter;
式中，rij表示第i项与第j项特征参数的相关系数，此处采用由MIC法执行特征筛选时得到的值，Rj用以承载该特征参数的冲突性；In the formula, Rij represents the correlation coefficient between the i-th and j-th feature parameters. The value obtained when the MIC method is used to perform feature screening is used here, and Rj is used to carry the conflict of the feature parameter;
式中，pij表示第j项特征参数在第i个充电片段时的比重，Ej为第j项特征参数的熵值，承载了该特征参数的离散性；In the formula, pij represents the weight of the jth characteristic parameter in the i-th charging segment, and Ej is the entropy value of the jth characteristic parameter, which carries the discreteness of the characteristic parameter;
由此得到第j项表征参数的信息量Cj：Thus, the information content Cj of the j-th characterization parameter is obtained:
Cj＝(Ej+Sj)RjCj =(Ej +Sj )Rj
计算权重：Calculate weights:
计算各充电片段综合得分：Calculate the comprehensive score of each charging segment:
得到动力电池充电片段风险评估序列F：Get the power battery charging segment risk assessment sequence F:
F＝[F1 F2 … Fi … Fn]TF＝[F1 F2 … Fi … Fn ]T
风险评估序列F即为电动车辆各充电片段所得的综合风险评分。The risk assessment sequence F is the comprehensive risk score obtained from each charging segment of the electric vehicle.
进一步地，步骤六中具体基于高斯分布“3σ”原则，由正常车风险评分的概率分布的均值μ和标准差σ，来确定安全风险识别阈值RiskT；若动力电池系统风险评估综合评分大于该阈值，则发出风险预警信号，提醒车辆驾乘人员对动力电池系统进行及时地检查和维修，避免动力电池系统故障恶化，甚至热失控的发生。Furthermore, in step six, based on the "3σ" principle of Gaussian distribution, the safety risk identification threshold RiskT is determined by the mean μ and standard deviation σ of the probability distribution of the risk score of a normal vehicle; if the comprehensive score of the power battery system risk assessment is greater than the threshold, a risk warning signal is issued to remind the vehicle driver and passengers to inspect and repair the power battery system in a timely manner to avoid the deterioration of power battery system faults or even thermal runaway.
上述本发明所提供的融合健康衰退和一致性的动力电池风险评估方法，通过分析筛选用于表征动力电池系统监控衰退和一致性的特征参数，较准确地分析电池SOH了随累计行驶里程增加的演变规律，基于CRITIC权重法融合熵权法对其改进来建立SOH与一致性评估模型，并基于正常车辆与故障车辆的风险评分分布确定风险评估阈值，从而能够及时有效地发现动力电池系统潜在风险，有利于做出风险预警决策。The power battery risk assessment method that integrates health decline and consistency provided by the present invention analyzes and screens the characteristic parameters used to characterize the power battery system monitoring decline and consistency, and more accurately analyzes the evolution of the battery SOH with the increase of cumulative mileage. The SOH and consistency assessment model is established based on the CRITIC weight method and the entropy weight method. The risk assessment threshold is determined based on the risk score distribution of normal vehicles and faulty vehicles, so that potential risks of the power battery system can be discovered in a timely and effective manner, which is conducive to making risk warning decisions.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
图1为本发明所提供方法的流程框图；FIG1 is a flowchart of the method provided by the present invention;
图2为动力电池各充电片段综合评分结果图；FIG2 is a diagram showing the comprehensive scoring results of each charging segment of the power battery;
图3为正常车与故障车风险评分概率分布图。Figure 3 is the probability distribution diagram of risk scores for normal vehicles and faulty vehicles.
具体实施方式Detailed ways
下面将结合附图对本发明的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
本发明提供的融合健康衰退和一致性的动力电池风险评估方法，如图1所示，具体包括以下步骤：The power battery risk assessment method integrating health decline and consistency provided by the present invention, as shown in FIG1 , specifically comprises the following steps:
步骤一、从若干新能源汽车获取同类动力电池系统的实车运行数据，并利用实车运行数据中包括时间Time、总电流I、总电压U、SOC、车辆状态S、温度T、充电状态等的参数建立新能源汽车运行数据矩阵P：Step 1: Obtain the actual vehicle operation data of the same power battery system from several new energy vehicles, and use the parameters including time Time, total current I, total voltage U, SOC, vehicle state S, temperature T, charging state, etc. in the actual vehicle operation data to establish the new energy vehicle operation data matrix P:
P＝[Time,U,S,a1,a2,...,an]P＝[Time,U,S,a1 ,a2 ,...,an ]
其中，an表示第n个可用数据项对应的数据向量；Where an represents the data vector corresponding to the nth available data item;
步骤二、对所建立的新能源运行数据矩阵P中所包含的数据异常、数据缺失及错误数据执行包括线性插补和数据删除的数据预处理；Step 2: Perform data preprocessing including linear interpolation and data deletion on the data anomalies, data missing and erroneous data contained in the established new energy operation data matrix P;
步骤三、根据各新能源汽车实际使用时充电行为统计分析驾驶员的用车习惯，并在此基础上设定充电片段样本的选取规则；利用新能源汽车运行数据矩阵P获得切分后的各充电片段，并基于选取规则筛选出相应的充电片段样本；Step 3: Analyze the driver's driving habits according to the charging behaviors of each new energy vehicle during actual use, and set the selection rules of charging segment samples on this basis; use the new energy vehicle operation data matrix P to obtain the segmented charging segments, and select the corresponding charging segment samples based on the selection rules;
步骤四、利用充电片段样本计算固定SOC区间的容量增量和能量增量；提取不同时刻单体电压值与电池组SOC的乘积来建立反映电压一致性的电压表征参数序列，计算电池组在相应充电片段的基准电压表征序列，以及各单体电压表征参数向量现对该基准电压表征序列的曼哈顿距离矩阵；提取充电片段样本中电池探针温度数据，计算电池组探针温度在相应充电片段的基准温度表征序列，以及各探针温度向量相对该基准温度表征序列的曼哈顿距离矩阵；计算充电片段样本对应的充电功率表征参数序列以反映充电功率一致性，计算相应片段的基准充电功率序列，以及充电功率表征参数序列相对该基准充电功率序列的曼哈顿距离矩阵；Step 4: Calculate the capacity increment and energy increment of the fixed SOC interval using the charging segment samples; extract the product of the single cell voltage value and the battery pack SOC at different times to establish a voltage characterization parameter sequence reflecting voltage consistency, calculate the reference voltage characterization sequence of the battery pack in the corresponding charging segment, and the Manhattan distance matrix of each single cell voltage characterization parameter vector to the reference voltage characterization sequence; extract the battery probe temperature data in the charging segment samples, calculate the reference temperature characterization sequence of the battery pack probe temperature in the corresponding charging segment, and the Manhattan distance matrix of each probe temperature vector relative to the reference temperature characterization sequence; calculate the charging power characterization parameter sequence corresponding to the charging segment samples to reflect the charging power consistency, calculate the reference charging power sequence of the corresponding segment, and the Manhattan distance matrix of the charging power characterization parameter sequence relative to the reference charging power sequence;
将得到的容量增量、能量增量、各曼哈顿距离矩阵中离散程度最大单体所分别对应的表征参数序列Z分数进行相关性分析，从中筛选出最终用于评估电池组SOH与一致性的特征参数；The obtained capacity increment, energy increment, and Z-scores of the characterization parameter sequences corresponding to the monomers with the largest degree of dispersion in each Manhattan distance matrix are subjected to correlation analysis, from which the characteristic parameters ultimately used to evaluate the SOH and consistency of the battery pack are selected;
步骤五、基于CRITIC权重法来建立电池组SOH与一致性评估模型，并融合熵权法对CRITIC权重法进行改进，在计算充电片段风险评分时考虑其熵值，由此得到各充电片段的风险评估序列；Step 5: Establish a battery pack SOH and consistency assessment model based on the CRITIC weight method, and integrate the entropy weight method to improve the CRITIC weight method. Consider its entropy value when calculating the risk score of the charging segment, thereby obtaining the risk assessment sequence of each charging segment;
步骤六、基于风险评估序列确定正常车辆与故障车辆的风险评分分布情况，并基于正常车辆的所述分布确定相应的安全风险识别阈值；基于该阈值对实车风险进行预警决策。Step 6: Determine the risk score distribution of normal vehicles and faulty vehicles based on the risk assessment sequence, and determine the corresponding safety risk identification threshold based on the distribution of normal vehicles; and make early warning decisions on real vehicle risks based on the threshold.
在本发明的优选实施方式中，步骤一中针对时间Time、总电压U、车辆状态S和温度T数据具体采用以下向量展开形式：In a preferred embodiment of the present invention, in step 1, the following vector expansion form is specifically used for the time Time, total voltage U, vehicle state S and temperature T data:
Time＝[time1,time2,time3,···,timej]TTime＝[time1 , time2 , time3 ,···,timej ]T
其中，timei表示在数据采集时间i(i＝1,2,3,...,j)；Wherein, timei represents the data collection time i (i = 1, 2, 3, ..., j);
其中，k是电池单体的数量，j是对应与不同采样时间的行数，Ut,h表示电池单体h(h＝1,2,3,...,k)在时刻t(t＝1,2,3,...,j)的电压值；Wherein, k is the number of battery cells, j is the number of rows corresponding to different sampling times, and Ut,h represents the voltage value of battery cell h (h = 1, 2, 3, ..., k) at time t (t = 1, 2, 3, ..., j);
其中，It、vt、soct、rt分别表示电池组在时刻t(t＝1,2,3,...,j)的总电流值、车辆行驶速度、电池组SOC、绝缘电阻值；Wherein, It , vt , soct , and rt represent the total current value of the battery pack at time t (t=1, 2, 3, ..., j), the vehicle speed, the battery pack SOC, and the insulation resistance value, respectively;
其中，m是温度探针的数量，j是对应于不同采样时间的行数，Tt,l表示电池组温度探针l(l＝1,2,3,...,m)在时刻t(t＝1,2,3,...,j)的温度值。Wherein, m is the number of temperature probes, j is the number of rows corresponding to different sampling times, and Tt,l represents the temperature value of battery pack temperature probe l (l=1,2,3,...,m) at time t (t=1,2,3,...,j).
由于在电动汽车实际使用过程中，充电起始/终止SOC与车辆的用途和驾驶员习惯有关，并不局限于某一范围内，故可将充电起始SOC分布近似视为正态分布，通过观察，受使用者的里程焦虑等原因影响，充电起始SOC主要集中在40～60之间，充电终止SOC占比则随着SOC的增大而提高，有70％以上的充电过程都充至100％而结束。对被试车辆的统计，充电SOC变化值的分布图可以看出，充电深度分布集中于20％～70％之间，分布范围较大，且占比较为均匀，充电持续时间分布呈现峰值趋势，主要集中在200～500分钟。为了满足后续研究的数据可靠性和充足性，需要筛选合适的充电片段，通过上节对动力电池充电片段的分析，根据充电片段起始/终止SOC分布的统计结果，综合考虑充电SOC区间，若选择的区间太小，则会导致所含信息太少，且充电行为会对环境条件更加敏感，从而导致产生较大的数据波动，将影响结论的鲁棒性；若选择的区间太大，则会导致满足条件的充电片段样本数量太少，不足以支撑结论的可靠性。为此，在本发明的优选实施方式中，步骤三中通过对充电行为和驾驶员习惯的分析，选择包含60％～70％SOC区间的充电片段样本。Since in the actual use of electric vehicles, the charging start/end SOC is related to the purpose of the vehicle and the driver's habits, and is not limited to a certain range, the charging start SOC distribution can be approximately regarded as a normal distribution. Through observation, affected by the user's mileage anxiety and other reasons, the charging start SOC is mainly concentrated between 40 and 60, and the charging end SOC proportion increases with the increase of SOC. More than 70% of the charging process ends with charging to 100%. From the statistics of the tested vehicles, the distribution diagram of the charging SOC change value shows that the charging depth distribution is concentrated between 20% and 70%, with a large distribution range and a relatively uniform proportion. The charging duration distribution shows a peak trend, mainly concentrated in 200 to 500 minutes. In order to meet the data reliability and adequacy of subsequent research, it is necessary to screen appropriate charging segments. Through the analysis of the power battery charging segment in the previous section, according to the statistical results of the start/end SOC distribution of the charging segment, the charging SOC interval is comprehensively considered. If the selected interval is too small, it will lead to too little information, and the charging behavior will be more sensitive to environmental conditions, resulting in large data fluctuations, which will affect the robustness of the conclusion; if the selected interval is too large, the number of charging segment samples that meet the conditions will be too small, which is not enough to support the reliability of the conclusion. For this reason, in the preferred embodiment of the present invention, in step three, by analyzing the charging behavior and driver habits, a charging segment sample containing a 60% to 70% SOC interval is selected.
在本发明的优选实施方式中，步骤四中针对充电片段样本的60％～70％ SOC区间，利用以下安时积分计算固定SOC区间的容量增量：In a preferred embodiment of the present invention, in step 4, for the 60% to 70% SOC interval of the charging segment sample, the capacity increment in the fixed SOC interval is calculated using the following ampere-hour integral:
其中，I表示动力电池实时充电电流，TSOCstart和TSOCend分别代表动力电池充电SOC区间开始和结束的时间；Where I represents the real-time charging current of the power battery, TSOCstart and TSOCend represent the start and end time of the power battery charging SOC interval respectively;
针对充电片段样本的相同SOC区间，利用以下安时积分计算固定SOC区间的能量增量：For the same SOC interval of the charging segment sample, the energy increment of the fixed SOC interval is calculated using the following ampere-hour integration:
在实车使用过程中，动力电池的单体电压是其状态情况最直观的体现，因为其实时准确的特点，目前大部分研究均基于单体电压进行动力电池的故障诊断。研究表明，存在不一致问题的动力电池在充放电末期的强极化非线性会引起动力电池组“扫帚”效应，即充放电末期动力电池组内单体间的不一致性会显著增大，在“帚劲”之前电池单体的一致性良好，在“帚劲”之后，电池的不一致性会显现出来。基于此特点，本发明提出一种基于动力电池组“扫帚”效应的电压一致性分析的方案，针对实车充电历史数据，结合充电SOC，开展电压一致性分析。具体过程包括利用充电片段样本的单体电压数据以及电池组SOC计算所述电压表征参数序列M：During the use of actual vehicles, the single cell voltage of the power battery is the most intuitive reflection of its status. Due to its real-time and accurate characteristics, most current studies are based on single cell voltage for power battery fault diagnosis. Studies have shown that the strong polarization nonlinearity of power batteries with inconsistency problems at the end of charge and discharge will cause the "broom" effect of the power battery pack, that is, the inconsistency between the single cells in the power battery pack at the end of charge and discharge will increase significantly. Before the "broom effect", the consistency of the battery cells is good, and after the "broom effect", the inconsistency of the battery will be revealed. Based on this feature, the present invention proposes a voltage consistency analysis scheme based on the "broom" effect of the power battery pack, and conducts voltage consistency analysis based on the actual vehicle charging history data and the charging SOC. The specific process includes calculating the voltage characterization parameter sequence M using the single cell voltage data of the charging segment sample and the battery pack SOC:
式中，Mh表示电池单体h的表征向量，ut,h表示电池单体h在时刻t(t＝1,2,3,...,j)的电压值，j是对应不同采样时间的行数，soctut,h表示电池单体h在t时刻电压值与此刻电池SOC的乘积，即电压表征参数值；Wherein, Mh represents the characterization vector of battery cell h,ut,h represents the voltage value of battery cell h at time t (t=1, 2, 3, ..., j), j is the number of rows corresponding to different sampling times, soct ut,h represents the product of the voltage value of battery cell h at time t and the battery SOC at this moment, that is, the voltage characterization parameter value;
计算基准电压表征序列Mdatum为：The reference voltage characterization sequence Mdatum is calculated as:
式中，表示时刻t(t＝1,2,3,...,j)的所有k个单体电池电压表征参数的平均值；In the formula, represents the average value of the voltage characteristic parameters of all k single cells at time t (t=1, 2, 3, ..., j);
计算充电片段内各电池单体电压表征向量相对于基准电压表征序列的曼哈顿距离矩阵DM：Calculate the Manhattan distance matrix DM of each battery cell voltage characterization vector in the charging segment relative to the reference voltage characterization sequence:
DM＝[DM,1 DM,2 … DM,h … DM,k]DM =[DM,1 DM,2 ... DM,h ... DM,k ]
式中，|soctut,h-socdatumut,datum|表示单体h在时刻t相对于基准表征值的曼哈顿距离；In the formula, |soct ut,h -socdatum ut,datum | represents the Manhattan distance of monomer h relative to the benchmark representation value at time t;
电压离散程度最大单体所对应的表征参数序列Z分数基于以下公式计算：The Z score of the characterization parameter sequence corresponding to the monomer with the largest voltage dispersion is calculated based on the following formula:
式中，表示各个电池单体在本充电片段内相对于基准电压表征序列的曼哈顿距离平均值，σDM表示曼哈顿距离的标准差，δDM,max表示相应的Z分数；In the formula, represents the average Manhattan distance of each battery cell relative to the reference voltage characterization sequence in this charging segment, σDM represents the standard deviation of the Manhattan distance, and δDM,max represents the corresponding Z score;
在实车使用过程中，由于外界环境温度变化、电池单体内阻不一致以及电池模组设计的差异，致使电池内部不同位置的温度产生差异，这会进一步影响电池单体的内阻，从而发生恶性循环，引发各电池单体的衰退速度不一致，影响动力电池性能。因此针对电池组探针温度开展一致性分析，对风险预警和评估具有重要意义。温度一致性分析过程包括：提取充电片段样本中的探针温度时间序列Th并计算基准温度表征序列Tdatum：During the actual use of the vehicle, due to changes in the external environment temperature, inconsistent internal resistance of battery cells and differences in battery module design, the temperature of different locations inside the battery will be different, which will further affect the internal resistance of the battery cells, thus causing a vicious cycle, causing inconsistent decay rates of each battery cell, and affecting the performance of the power battery. Therefore, conducting consistency analysis on the battery pack probe temperature is of great significance for risk warning and assessment. The temperature consistency analysis process includes: extracting the probe temperature time series Th in the charging segment sample and calculating the reference temperature characterization sequence Tdatum :
式中，表示时刻t(t＝1,2,3,...,j)的所有探针温度的平均值；In the formula, represents the average value of all probe temperatures at time t (t = 1, 2, 3, ..., j);
计算充电片段样本内各探针温度时间序列相对于温度基准表征序列的曼哈顿距离矩阵DT：Calculate the Manhattan distance matrix DT of each probe temperature time series in the charging segment sample relative to the temperature reference characterization sequence:
DT＝[DT,1 DT,2 … DT,h … DT,l]DT =[DT,1 DT,2 ... DT,h ... DT,l ]
其中，DT,h表示探针h相对于探针温度基准表征序列的曼哈顿距离，其计算方式如下：Where DT,h represents the Manhattan distance of probe h relative to the probe temperature benchmark characterization sequence, which is calculated as follows:
式中，tt,h表示探针h在时刻t(t＝1,2,3,...,j)的温度值，|tt,h-tt,datum|表示探针h在时刻t相对于温度基准序列的曼哈顿距离；Where tt,h represents the temperature value of probe h at time t (t = 1, 2, 3, ..., j), |tt,h -tt,datum | represents the Manhattan distance of probe h at time t relative to the temperature reference sequence;
温度离散程度最大单体所对应的表征参数序列Z分数基于以下公式计算：The Z score of the characterization parameter sequence corresponding to the monomer with the largest temperature dispersion is calculated based on the following formula:
式中，表示各探针温度在本充电片段内相对于基准温度序列的曼哈顿距离平均值，σDT表示此曼哈顿距离的标准差，δDT,max表示相应的Z分数；In the formula, represents the average Manhattan distance of each probe temperature relative to the reference temperature sequence in this charging segment, σDT represents the standard deviation of this Manhattan distance, and δDT,max represents the corresponding Z score;
动力电池实际使用过程中各种因素相互耦合，除了导致电压、温度的不一致，也会进一步导致各单体的功率衰退速率不一致，最终形成不一致扩大正反馈效应，为此本发明执行以下充电功率一致性分析过程：Various factors are coupled with each other during the actual use of power batteries. In addition to causing inconsistencies in voltage and temperature, they will also further lead to inconsistent power decay rates of each monomer, ultimately forming an inconsistent and amplified positive feedback effect. For this reason, the present invention performs the following charging power consistency analysis process:
基准充电功率序列基于以下公式计算：The base charging power sequence is calculated based on the following formula:
式中，Ph表示单体h的充电功率时间序列；Where,Ph represents the charging power time series of cell h;
通过与电压一致性和温度一致性相同的过程计算曼哈顿距离矩阵DP：The Manhattan distance matrix DP is calculated by the same process as for voltage consistency and temperature consistency:
DP＝[DP,1 DP,2 … DP,h … DP,k]DP =[DP,1 DP,2 ... DP,h ... DP,k ]
并利用以下公式计算得到充电功率离散程度最大单体所对应的表征参数序列Z分数：The Z score of the characterization parameter sequence corresponding to the monomer with the largest degree of charging power dispersion is calculated using the following formula:
表示上述曼哈顿距离的平均值，σDP表示曼哈顿距离的标准差，δDP,max表示相应的Z分数。 represents the mean of the above Manhattan distance, σDP represents the standard deviation of the Manhattan distance, and δDP,max represents the corresponding Z score.
在传统的相关性分析算法，如Pearson相关系数或者Spearman相关系数能够有效地评估数据的线性相关性，然而很多变量由于其复杂性，其相关关系是非线性的，并很难用简单的公式进行数学表达，因此，基于阈值相关、相位同步相关、距离相关、互信息等的相关性计算方法被国内外学者提出，最大信息系数(Maximal Information Coefficient,MIC)即在互信息的基础上提出的，能够快速地给不同类型的相关关系进行评估，而动力电池系统是一个典型的非线性系统。为此在本发明的优选实施方式中，步骤四中对得到的容量增量、能量增量、各曼哈顿距离矩阵中离散程度最大单体所分别对应的表征参数序列Z分数，具体利用MIC法进行相关性分析，具体公式如下：In traditional correlation analysis algorithms, such as Pearson correlation coefficient or Spearman correlation coefficient, the linear correlation of data can be effectively evaluated. However, due to its complexity, the correlation of many variables is nonlinear and difficult to express mathematically with simple formulas. Therefore, correlation calculation methods based on threshold correlation, phase synchronization correlation, distance correlation, mutual information, etc. have been proposed by domestic and foreign scholars. The maximum information coefficient (MIC) is proposed on the basis of mutual information, which can quickly evaluate different types of correlations, and the power battery system is a typical nonlinear system. For this reason, in a preferred embodiment of the present invention, in step 4, the capacity increment, energy increment, and the characterization parameter sequence Z score corresponding to the maximum discrete monomer in each Manhattan distance matrix are obtained, and the MIC method is used for correlation analysis. The specific formula is as follows:
式中，X和Y是两个随机变量，B是MIC算法的变量，I[X；Y]是随机变量X和Y的互信息，定义为联合分布p(x,y)与边缘分布p(x)p(y)相对熵，其计算公式如下：Where X and Y are two random variables, B is the variable of the MIC algorithm, and I[X;Y] is the mutual information of random variables X and Y, which is defined as the relative entropy of the joint distribution p(x,y) and the marginal distribution p(x)p(y). The calculation formula is as follows:
式中，p(x,y)是两个随机变量(X,Y)的联合分布，p(x)、p(y)为随机变量(X,Y)边缘分布；Where p(x,y) is the joint distribution of two random variables (X,Y), p(x) and p(y) are the marginal distributions of the random variables (X,Y);
通过相关性分析最终确定固定SOC区间容量增量、电压与温度一致性分别对应的Z分数作为特征参数。Through correlation analysis, the Z scores corresponding to the capacity increment, voltage and temperature consistency in a fixed SOC interval are finally determined as characteristic parameters.
MIC值在区间[0,1]内，值越高，表示变量X和Y的相关性越强，通过对被试车辆的相关表征参数分析，在本实施例中最终舍弃固定SOC区间能量增量和功率一致性表征参数这两个变量，即选择固定SOC区间容量增量作为动力电池衰退的表征参数，选择电压一致性表征参数和温度一致性表征参数作为动力电池一致性的表征参数。The MIC value is in the interval [0,1]. The higher the value, the stronger the correlation between variables X and Y. By analyzing the relevant characterization parameters of the test vehicle, in this embodiment, the two variables of fixed SOC interval energy increment and power consistency characterization parameter are finally discarded, that is, the fixed SOC interval capacity increment is selected as the characterization parameter of power battery degradation, and the voltage consistency characterization parameter and temperature consistency characterization parameter are selected as the characterization parameters of power battery consistency.
在本发明的优选实施方式中，步骤五中电池组SOH与一致性评估模型建立与改进的具体包括：In a preferred embodiment of the present invention, the establishment and improvement of the battery pack SOH and consistency evaluation model in step 5 specifically includes:
首先提取新能源汽车每个充电片段对应的特征参数，得到动力电池风险评价矩阵A：First, the characteristic parameters corresponding to each charging segment of new energy vehicles are extracted to obtain the power battery risk assessment matrix A:
式中，A1,A2,A3分别表示容量增量、电压一致性特征参数、温度一致性特征参数序列,aij代表充电片段i(i＝1,2,…,n)的第j(j＝1,2,3)项表征参数值；Wherein, A1 , A2 , A3 represent the capacity increment, voltage consistency characteristic parameter, and temperature consistency characteristic parameter sequence, respectively, and aij represents the jth (j＝1,2,3) characterization parameter value of charging segment i (i＝1,2,…,n);
对各特征参数执行归一化处理，将容量增量数据处理为负向指标，电压和温度一致性特征参数处理为正向指标：Normalize each characteristic parameter, treat the capacity increment data as a negative indicator, and treat the voltage and temperature consistency characteristic parameters as positive indicators:
正向指标处理公式：Positive indicator processing formula:
负向指标处理公式：Negative indicator processing formula:
式中，aij为正向/负向指标处理后所得到的充电片段i的第j项特征参数值；Where aij is the jth characteristic parameter value of charging segment i obtained after positive/negative index processing;
计算数据各特征参数的信息承载量：Calculate the information carrying capacity of each characteristic parameter of the data:
式中，表示第j项特征参数的均值，Sj是第j项特征参数的标准差，用以承载该特征参数的波动性；In the formula, represents the mean of the j-th characteristic parameter,Sj is the standard deviation of the j-th characteristic parameter, which is used to carry the volatility of the characteristic parameter;
式中，rij表示第i项与第j项特征参数的相关系数，此处采用由MIC法执行特征筛选时得到的值，Rj用以承载该特征参数的冲突性；In the formula, Rij represents the correlation coefficient between the i-th and j-th feature parameters. The value obtained when the MIC method is used to perform feature screening is used here, and Rj is used to carry the conflict of the feature parameter;
式中，pij表示第j项特征参数在第i个充电片段时的比重，Ej为第j项特征参数的熵值，承载了该特征参数的离散性；In the formula, pij represents the weight of the jth characteristic parameter in the i-th charging segment, and Ej is the entropy value of the jth characteristic parameter, which carries the discreteness of the characteristic parameter;
由此得到第j项表征参数的信息量Cj：Thus, the information content Cj of the j-th characterization parameter is obtained:
Cj＝(Ej+Sj)RjCj =(Ej +Sj )Rj
计算权重：Calculate weights:
计算各充电片段综合得分：Calculate the comprehensive score of each charging segment:
得到动力电池充电片段风险评估序列F：Get the power battery charging segment risk assessment sequence F:
F＝[F1 F2 … Fi … Fn]TF＝[F1 F2 … Fi … Fn ]T
风险评估序列F即为电动车辆各充电片段所得的综合风险评分。The risk assessment sequence F is the comprehensive risk score obtained from each charging segment of the electric vehicle.
在本发明的优选实施方式中，对同车型的8辆实车历史运行数据进行参数提取与综合评分计算，共1641个充电片段，综合评分如图2所示，其中包括正常车3辆，共700个充电片段，故障车5辆，共941个充电片段，两类车辆数之比与充电片段之比大致相同。In a preferred embodiment of the present invention, parameter extraction and comprehensive score calculation are performed on the historical operation data of 8 real vehicles of the same model, with a total of 1641 charging segments. The comprehensive score is shown in Figure 2, including 3 normal vehicles with a total of 700 charging segments and 5 faulty vehicles with a total of 941 charging segments. The ratio of the two types of vehicles is roughly the same as the ratio of the charging segments.
从图2中可以看出，3辆正常车各充电片段的综合评分都较为稳定，均处于0.3-1.0左右，5辆故障车在发生事故前，其充电片段的综合评分随着循环次数的增加都不同程度的向上偏移，并超过正常车充电片段的综合评分范围，因此分析正常车与故障车的充电片段综合评分的概率分布，如图3所示，根据其分布差异确定安全风险评估报警阈值。As can be seen from Figure 2, the comprehensive scores of each charging segment of the three normal vehicles are relatively stable, all around 0.3-1.0. Before the accident, the comprehensive scores of the charging segments of the five faulty vehicles shifted upward to varying degrees with the increase in the number of cycles, and exceeded the comprehensive score range of the charging segments of normal vehicles. Therefore, the probability distribution of the comprehensive scores of the charging segments of normal vehicles and faulty vehicles is analyzed, as shown in Figure 3, and the safety risk assessment alarm threshold is determined based on their distribution differences.
如图所示，正常车的综合评分分布曲线近似于高斯分布，而故障车分布曲线由于其综合评分的偏移与异常，并不完全符合高斯分布，且分布中心相对于正常车分布曲线向右偏移。下表为该8辆车的分布参数对比。As shown in the figure, the comprehensive score distribution curve of normal vehicles is close to Gaussian distribution, while the distribution curve of faulty vehicles does not completely conform to Gaussian distribution due to the deviation and abnormality of their comprehensive scores, and the distribution center is shifted to the right relative to the normal vehicle distribution curve. The following table compares the distribution parameters of the eight vehicles.
表1正常车与故障车风险评分分布参数表Table 1. Risk score distribution parameters of normal vehicles and faulty vehicles
由“小概率事件”和假设检验的基本思想，“小概率事件”的发生概率小于5％，可以认为在一次试验中，该事件几乎不可能发生。高斯分布中，事件落在(μ-3σ，μ+3σ)之内的概率为99.7％，即在此区间之外的概率仅为0.03％，远小于5％，因此在本节中利用“3σ”原则确定风险阈值，由表可知正常车1,2,3的μ+3σ分别为1.13,0.97,1.24，为避免误诊断，将此风险阈值设定为1.25。；若动力电池系统风险评估综合评分大于该阈值，则发出风险预警信号，提醒车辆驾乘人员对动力电池系统进行及时地检查和维修，避免动力电池系统故障恶化，甚至热失控的发生。According to the basic idea of "low probability events" and hypothesis testing, the probability of occurrence of "low probability events" is less than 5%. It can be considered that in a test, the event is almost impossible to occur. In Gaussian distribution, the probability of an event falling within (μ-3σ, μ+3σ) is 99.7%, that is, the probability outside this interval is only 0.03%, which is much less than 5%. Therefore, in this section, the "3σ" principle is used to determine the risk threshold. It can be seen from the table that the μ+3σ of normal vehicles 1, 2, and 3 are 1.13, 0.97, and 1.24 respectively. In order to avoid misdiagnosis, this risk threshold is set to 1.25. If the comprehensive score of the power battery system risk assessment is greater than the threshold, a risk warning signal is issued to remind the vehicle driver and passengers to check and repair the power battery system in a timely manner to avoid the deterioration of power battery system failures and even thermal runaway.
应理解，本发明实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本发明实施例的实施过程构成任何限定。It should be understood that the size of the serial numbers of the steps in the embodiment of the present invention does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
尽管已经示出和描述了本发明的实施例，对于本领域的普通技术人员而言，可以理解在不脱离本发明的原理和精神的情况下可以对这些实施例进行多种变化、修改、替换和变型，本发明的范围由所附权利要求及其等同物限定。Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410570035.2ACN118425818A (en) | 2024-05-09 | 2024-05-09 | A power battery risk assessment method integrating health decline and consistency |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410570035.2ACN118425818A (en) | 2024-05-09 | 2024-05-09 | A power battery risk assessment method integrating health decline and consistency |
| Publication Number | Publication Date |
|---|---|
| CN118425818Atrue CN118425818A (en) | 2024-08-02 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410570035.2APendingCN118425818A (en) | 2024-05-09 | 2024-05-09 | A power battery risk assessment method integrating health decline and consistency |
| Country | Link |
|---|---|
| CN (1) | CN118425818A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119471402A (en)* | 2025-01-15 | 2025-02-18 | 佛山大学 | A method and device for detecting internal short circuit of lithium-ion battery based on improved GAF |
| CN120314798A (en)* | 2025-06-19 | 2025-07-15 | 国家市场监督管理总局缺陷产品召回技术中心 | Adaptive evaluation method, device, equipment and medium for operating safety status of electrochemical energy storage battery |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119471402A (en)* | 2025-01-15 | 2025-02-18 | 佛山大学 | A method and device for detecting internal short circuit of lithium-ion battery based on improved GAF |
| CN120314798A (en)* | 2025-06-19 | 2025-07-15 | 国家市场监督管理总局缺陷产品召回技术中心 | Adaptive evaluation method, device, equipment and medium for operating safety status of electrochemical energy storage battery |
| Publication | Publication Date | Title |
|---|---|---|
| CN112731159B (en) | A method for predicting and locating battery faults in battery compartments of energy storage power stations | |
| CN113933732B (en) | New energy automobile power battery health state analysis method, system and storage medium | |
| CN112092675B (en) | Battery thermal runaway early warning method, system and server | |
| CN114578251B (en) | Battery module safety status assessment method and device based on convolutional neural network | |
| CN102749589B (en) | Recession-mode predicting method of power battery of electric automobile | |
| Xu et al. | A vehicle-cloud collaborative method for multi-type fault diagnosis of lithium-ion batteries | |
| CN114280479B (en) | Retired battery rapid sorting method based on electrochemical impedance spectroscopy | |
| CN103954913B (en) | Electric automobile power battery life-span prediction method | |
| CN119024189B (en) | Low-speed electric vehicle battery life assessment prediction method and system based on big data | |
| CN118425818A (en) | A power battery risk assessment method integrating health decline and consistency | |
| CN115366683B (en) | A fault diagnosis strategy based on multi-dimensional model fusion for new energy vehicle power battery | |
| CN114036647A (en) | Power battery safety risk assessment method based on real vehicle data | |
| WO2023124220A1 (en) | Vehicle detection method and device | |
| CN114460481B (en) | Bi-LSTM and attention mechanism-based energy storage battery thermal runaway early warning method | |
| CN115128468A (en) | Chemical energy storage battery PHM undervoltage fault prediction method | |
| CN116901707A (en) | A power battery pack failure early warning method, system and vehicle | |
| CN118938016A (en) | A method for judging the aging trend of power batteries | |
| CN118132959A (en) | A method for quantifying inconsistency of power batteries based on battery cell voltage values | |
| CN119619858A (en) | A safety early warning method and device for power battery | |
| CN118759370B (en) | Accident battery damage assessment method and system based on Internet of vehicles big data | |
| CN119224592A (en) | New energy battery health status monitoring method and system based on big data analysis | |
| CN119471443A (en) | Battery failure detection method, device and non-volatile storage medium | |
| CN112782603B (en) | A method for fitting the cycle life distribution of lithium-ion batteries based on interval-censored data | |
| Zhang et al. | Battery defect detection for real world vehicles based on Gaussian distribution parameterization developed LCSS | |
| CN119511125A (en) | An intelligent monitoring method based on dynamic modeling of charging screen and battery pack |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |