New energy vehicle battery health status assessment method based on multidimensional data analysis

Abstract

The present invention discloses a method for evaluating the health status of a new energy vehicle battery based on multidimensional data analysis, and relates to the technical field of new energy vehicle batteries. The method comprises the following steps: S1, initializing the structure and parameters of a linear model; S2, obtaining the index data of a vehicle single battery, processing the index data, and dividing the training set and the test set, wherein the index data includes the monthly voltage consistency, temperature consistency, internal resistance consistency, and capacity retention rate of the battery; S3, pre-training and cross-training the linear model; S4, using the data of the test set to evaluate the linear model and obtain the evaluation result; S5, comparing the evaluation result with the size of a preset threshold, and if the evaluation result is less than or equal to the preset threshold, repeating S3 and S4; if the evaluation result is greater than the preset threshold, the linear model outputs the evaluation result. By calculating the four-dimensional index data of the vehicle battery, the overall health status of the new energy vehicle battery during operation can be more accurately reflected.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy vehicle batteries, and in particular to a new energy vehicle battery health status assessment method based on multidimensional data analysis. Background Art

[0002] As the new energy vehicle market continues to grow, the assessment of the health status of new energy vehicle batteries is a topic that urgently needs to be improved.

[0003] Since the battery system generates a variety of data types, including voltage, current, temperature, charging times, etc., these data come from different sensors and control systems with different formats and sampling frequencies. Effectively integrating and synchronously analyzing these multi-dimensional data requires complex data preprocessing and conversion processes, which is very challenging. Therefore, most of the current assessments of the health status of new energy vehicle batteries are based on analysis of single indicators.

[0004] For example, Chinese invention patent CN117741446B discloses a method and system for evaluating the health status of new energy vehicles batteries, which specifically include: calculating the current capacity corresponding to each charging segment data, thereby calculating SOH as a label; using the local outlier factor algorithm, box plot principle or 3σ principle to eliminate outliers for the label capacity; analyzing and selecting features that are highly correlated with the battery health status SOH; inputting the selected feature data into a linear model for training; extracting the corresponding features of the vehicle to be predicted, and inputting them into the trained model for capacity prediction. It can be seen that this technical solution also only analyzes from the indicator of battery capacity. Obviously, analyzing a single battery indicator cannot fully reflect the health status of the battery. Summary of the invention

[0005] The technical problem to be solved by the present invention is how to analyze the health status of new energy vehicle batteries based on multiple indicators.

[0006] In order to solve the above technical problems, the present invention provides a new energy vehicle battery health status assessment method based on multidimensional data analysis, the method comprising the following steps:

[0007] S1. Initialize the structure and parameters of the linear model constructed by the linear regression method; wherein the linear model satisfies the formula:

[0008]

[0009] in, is the dependent variable; is the independent variable; is the corresponding weight coefficient; is the intercept;

[0010] S2. Calculate and obtain the index data of the vehicle single battery, normalize and standardize the index data, and divide the index data into a training set and a test set; wherein the index data includes the monthly voltage consistency, temperature consistency, internal resistance consistency and capacity retention rate of the vehicle single battery;

[0011] S3, pre-training and cross-training the linear model using the indicator data to update the linear model;

[0012] S4, using the indicator data of the test set to evaluate the linear model and obtain an evaluation result;

[0013] S5. Compare the evaluation result with a preset threshold value. If the evaluation result is less than or equal to the preset threshold value, repeat S3 and S4. If the evaluation result is greater than the preset threshold value, the linear model outputs the evaluation result, and then outputs an evaluation report on the battery health status.

[0014] Furthermore, the calculation method of the monthly voltage consistency and temperature consistency of the vehicle single battery includes:

[0015] Obtaining a voltage data set and a temperature data set of the single battery of the vehicle at all times on the day;

[0016] The extreme values ​​of the voltage and temperature of the single battery at each moment are obtained respectively, and the root mean square error values ​​of the extreme values ​​of the voltage and temperature are calculated respectively. The root mean square error value satisfies the calculation formula:

[0017]

[0018] in, The total number of voltage and temperature data of the single battery at all times of the vehicle on that day; They are Maximum voltage and temperature at a given moment; They are Minimum voltage and temperature at a given moment.

[0019] Furthermore, the method for calculating the monthly voltage consistency of the vehicle single battery also includes:

[0020] Calculate the average value of the vehicle's voltage data at all times of the day;

[0021] The voltage variation coefficient is calculated according to the average value of the voltage data to evaluate the discrete situation of the voltage data. The voltage variation coefficient satisfies the calculation formula:

[0022]

[0023] in, The root mean square error of the voltage extreme value, It is the average value of the vehicle's voltage data at all times of the day.

[0024] Furthermore, the calculation method of the monthly internal resistance consistency of the vehicle single battery includes:

[0025] Get the first and last data of the vehicle's complete charging period;

[0026] Compare the voltage data difference and current data difference at the time corresponding to the first and last data to obtain the internal resistance change set of the single battery The average value of internal resistance change , calculate the standard deviation of the internal resistance change of the single battery, and the standard deviation of the internal resistance change satisfies the calculation formula:

[0027]

[0028] in, is the number of batteries in the vehicle battery pack;

[0029] The internal resistance variation coefficient is calculated according to the average value of the internal resistance data to evaluate the discreteness of the internal resistance data. The internal resistance variation coefficient satisfies the calculation formula:

[0030]

[0031] in, is the standard deviation of the internal resistance variation.

[0032] Furthermore, the method for calculating the capacity retention rate of the vehicle single battery includes:

[0033] The energy storage data of the constant current charging segment of the vehicle battery for more than 200s is obtained to obtain the charge capacity of the battery per unit time. The SOC of the constant current charging segment of the vehicle battery extracted must be in the range of 40% to 90%, and the charge capacity of the battery per unit time satisfies the formula:

[0034]

[0035] in, is the amount of electricity charged into the battery per unit time. for The current magnitude at that moment, is the time difference between two frames of data;

[0036] The required capacity when the SOC is 100% is obtained based on the ratio of the amount of electricity charged into the battery per unit time to the SOC change. , the capacity Satisfy the calculation formula:

[0037]

[0038] in, The SOC change value within the constant current charging segment of the vehicle battery;

[0039] According to the capacity The capacity retention rate of the vehicle single battery is calculated, and the capacity retention rate of the vehicle single battery satisfies the calculation formula:

[0040]

[0041] in, is the factory rated voltage of the vehicle battery, The total storage capacity of the vehicle battery set at the factory.

[0042] Furthermore, the index data is standardized in S2 to satisfy the formula:

[0043]

[0044] in, is the mean of the original data, is the standard deviation of the original data;

[0045] In S2, the index data is normalized to satisfy the formula:

[0046]

[0047] in, Represents the maximum and minimum values ​​of the data.

[0048] Furthermore, in S3, a loss function is set to perform pre-training and cross-training on the linear model, wherein the loss function satisfies the formula:

[0049]

[0050] in, For the The true value of the samples, For the The predicted value of samples, is the total number of samples, for the ultimate loss;

[0051] In order to make the To minimize, the loss function is optimized by gradient descent, wherein the gradient descent satisfies the formula:

[0052]

[0053] in, is the parameter vector, is the learning rate, which is used to determine the step size of each update of the linear model. For the The predicted value of samples, For the The true value of the samples, is the total number of samples.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention calculates the four-dimensional indicator data of vehicle battery voltage consistency, temperature consistency, internal resistance consistency, and capacity, and standardizes and normalizes the data, transforming the local data set into data for linear model training, thereby reducing data deviation caused by abnormality of a certain indicator, and more accurately reflecting the overall health status of new energy vehicle batteries during operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flowchart of an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the technical scheme and technical effect of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0058] The present invention aims to provide a new energy vehicle battery health status assessment method based on multidimensional data analysis. Figure 1 , the method comprises the following steps:

[0059] Step 1, initializing the structure and parameters of the linear model constructed by the linear regression method;

[0060] Specifically, the linear model constructed by the linear regression method satisfies the formula:

[0061]

[0062] in, is the dependent variable; is the independent variable; is the corresponding weight coefficient; is the intercept.

[0063] Step 2: Obtain the index data of the vehicle's single battery, normalize and standardize the index data, and divide it into a training set and a test set;

[0064] Specifically, the present invention needs to obtain the index data of the vehicle single battery, including the monthly voltage consistency, temperature consistency, internal resistance consistency and capacity retention rate of the vehicle single battery. In this way, the evaluation based on the multi-dimensional data of the battery can reduce the data deviation caused by the abnormality of a certain indicator.

[0065] Furthermore, the data of the four dimensions of the vehicle single battery are respectively calculated, as described below:

[0066] Step 2.1: Calculation method for monthly voltage consistency and temperature consistency of vehicle single battery

[0067] Obtain the voltage data set and temperature data set of the single battery of the vehicle at all times on the day; obtain the extreme values ​​of the voltage and temperature of the single battery at each time, and calculate the root mean square error of the extreme values ​​of the voltage and temperature. The root mean square error satisfies the calculation formula:

[0068]

[0069] in, They are the total number of voltage and temperature data of the single battery of the vehicle at all times on the day; They are Maximum voltage and temperature at a given moment; They are Minimum voltage and temperature at a given moment.

[0070] The following are more specific calculation methods for the monthly voltage consistency and temperature consistency of vehicle single batteries:

[0071] Step 2.1.1: Calculation method for monthly voltage consistency of vehicle single battery

[0072] Obtain the single battery voltage data set of the vehicle at all times of the day and find the maximum voltage at each time , minimum , and the average value of the single cell voltage data at all times of the day , calculate the root mean square error of the maximum and minimum voltage values, satisfying the formula:

[0073]

[0074] in, is the total number of voltage data on that day, for The maximum voltage value at the time, for Minimum voltage at a given moment.

[0075] The voltage variation coefficient is calculated based on the average value of the voltage data to evaluate the discreteness of the voltage data. The voltage variation coefficient satisfies the calculation formula:

[0076]

[0077] in, The root mean square error of the voltage extreme value, It is the average value of the vehicle's voltage data at all times of the day.

[0078] Step 2.1.2: Calculation method for monthly temperature consistency of vehicle single battery

[0079] Get the temperature data set of the single battery at all times of the day, and find the highest temperature value at each time , minimum value , calculate the root mean square error value of the temperature curve, the formula is as follows:

[0080]

[0081] in is the total number of data items for the day, for The highest temperature value at that time, for The lowest temperature at the moment.

[0082] Step 2.2: Calculation method for monthly internal resistance consistency of vehicle single battery

[0083] Obtain the first and last data of the vehicle's complete charging period; compare the voltage data difference and current data difference at the time corresponding to the first and last data to obtain the internal resistance change set of the single battery The average value of internal resistance change , calculate the standard deviation of the internal resistance change of the single battery, the standard deviation of the internal resistance change satisfies the calculation formula:

[0084]

[0085] Where nc is the number of batteries in the vehicle battery pack, is the average value of the internal resistance change between two moments, It is a set of internal resistance change values ​​between each single battery at two moments.

[0086] The internal resistance variation coefficient is calculated based on the average value of the internal resistance data to evaluate the discrete situation of the internal resistance data. The internal resistance variation coefficient satisfies the calculation formula:

[0087]

[0088] in, is the standard deviation of the internal resistance variation.

[0089] Step 2.3: Calculation method for capacity retention rate of vehicle single battery

[0090] The on-time measurement method is used to calculate the capacity retention rate of vehicle single batteries. Compared with other methods, the on-time measurement method is less restricted by the battery's own conditions. The calculation method is simple and reliable, and can perform real-time evaluation of the battery's state of charge.

[0091] Specifically, the energy storage data of the constant current charging segment of the vehicle battery for more than 200s is obtained to obtain the battery charge per unit time. The SOC of the extracted constant current charging segment of the vehicle battery must be in the range of 40% to 90%. SOC is the state of charge of the battery, which refers to the ratio of the remaining charge in the battery to the total available charge. The battery charge per unit time satisfies the formula:

[0092]

[0093] in, is the amount of electricity charged into the battery per unit time. for The current magnitude at that moment, It is the time difference between two frames of data.

[0094] The required capacity when the SOC is 100% is obtained based on the ratio of the amount of electricity charged into the battery per unit time to the SOC change. ,capacity Satisfy the calculation formula:

[0095]

[0096] in, It is the SOC change value during the constant current charging segment of the vehicle battery.

[0097] According to capacity Calculate the capacity retention rate of the vehicle's single battery. The capacity retention rate of the vehicle's single battery satisfies the calculation formula:

[0098]

[0099] in, The unit of is wh; The factory rated voltage of the vehicle battery, in V; It is the total storage capacity of the vehicle battery set at the factory, and its unit is Kwh.

[0100] Step 2.4: Standardize and normalize the above four-dimensional indicator data respectively.

[0101] Step 2.4.1: Standardize the indicator data to meet the formula:

[0102]

[0103] in, is the mean of the original data, is the standard deviation of the original data.

[0104] Standardizing the indicator data can achieve proportional scaling of the original data. By using the maximum, minimum or maximum value of the variable, the original data is converted into data within a certain range, thereby eliminating the impact of dimension and order of magnitude, and changing the weight of the variable in the analysis to solve different measurement problems.

[0105] Step 2.4.2: Normalize the indicator data to satisfy the formula:

[0106]

[0107] in, Represents the maximum and minimum values ​​of the data.

[0108] Different features often have different dimensional units, which will affect the results of data analysis. In order to eliminate the errors caused by inconsistent dimensional units between indicator data, it is necessary to normalize the indicator data of the four dimensions to achieve comparability between data indicators. After the original data is normalized, each indicator is a decimal between [0, 1], which is suitable for comprehensive comparative evaluation.

[0109] Furthermore, the data set is divided into training set: test set = 1:4.

[0110] Step 3, pre-training and cross-training the linear model to update the linear model;

[0111] Before training a linear model, you first need to define a loss function, which is used to measure the difference between the model's predicted value and the actual value. By setting the loss function, the linear model is pre-trained and cross-trained. The purpose is to minimize the loss function by optimizing the model parameters, that is, to reduce the overfitting phenomenon of the linear model. The loss function satisfies the formula:

[0112]

[0113] in, For the The true value of the samples, For the The predicted value of samples, is the total number of samples, for the ultimate loss;

[0114] In order to To minimize the loss, the loss function needs to be optimized. Common optimization methods are normal equations and gradient descent. Specifically:

[0115] The normal equation satisfies the formula:

[0116]

[0117] in, is the eigenvalue matrix, is the target value matrix, The result is an n*1 vector, in which each element represents the weight of the corresponding feature in the model. The best result can be obtained directly, but when the feature complexity is high, the solution speed will be too slow to obtain the result.

[0118] Compared with the normal equation, gradient descent is more suitable for large-scale data sets, can be used to solve the minimum value of complex nonlinear functions, and can automatically adjust the learning rate to improve the convergence speed. Among them, gradient descent satisfies the formula:

[0119]

[0120] in, is the parameter vector, is the learning rate, which is used to determine the step size of each update of the linear model. For the The predicted value of samples, For the The true value of the samples, is the total number of samples. In each iteration, the loss function is calculated for the parameter The gradient of , updates the parameters in the opposite direction of the gradient This is because the gradient points to the direction in which the loss function increases fastest, and the goal is to minimize the loss function. Controls the step size of each update. If Too large may lead to skipping the optimal solution; if If it is too small, the convergence speed will be very slow. Through the gradient descent algorithm, the parameters can be gradually adjusted. , so that the loss function Reach the minimum value, so that the model's predicted value is closer to the actual value.

[0121] Step 4: Use the data of the test set to evaluate the linear model and obtain the evaluation results.

[0122] Step 5: Compare the evaluation result with the preset threshold. If the evaluation result is less than or equal to the preset threshold, repeat steps 3 and 4. If the evaluation result is greater than the preset threshold, the linear model outputs the evaluation result.

[0123] Those skilled in the art further explained that the "preset threshold" is based on the indicator data in step 2, and the real health status score of the new energy battery is obtained through statistical deduction. Furthermore, considering the deviation between the real data and the linear model prediction result, the preset threshold of the model will be determined by adjusting the score down by 0.2% based on the score result.

[0124] To ensure the prediction accuracy of the model, when the model prediction result is less than or equal to the preset threshold, it means that the model training error is too large and it is necessary to adjust the cross-training parameters and re-train the model; if the model prediction result is greater than the preset threshold, it means that the model has achieved the expected accuracy.

[0125] Step 6: Complete the evaluation model output.

[0126] 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.

Claims

1. A new energy vehicle battery health status assessment method based on multidimensional data analysis, characterized in that: The method comprises the following steps: S1. Initialize the structure and parameters of a linear model constructed by a linear regression method; wherein the linear model satisfies the formula: in, is the dependent variable; is the independent variable; is the corresponding weight coefficient; is the intercept; S2. Calculate and obtain the index data of the vehicle single battery, normalize and standardize the index data, and divide the index data into a training set and a test set; wherein the index data includes the monthly voltage consistency, temperature consistency, internal resistance consistency and capacity retention rate of the vehicle single battery; S3, using the indicator data of the training set to pre-train and cross-train the linear model to update the linear model; S4, using the indicator data of the test set to evaluate the linear model and obtain an evaluation result; S5. Compare the evaluation result with a preset threshold value. If the evaluation result is less than or equal to the preset threshold value, repeat S3 and S4. If the evaluation result is greater than the preset threshold value, the linear model outputs the evaluation result, and then outputs an evaluation report on the battery health status.

2. The method for evaluating the health status of new energy vehicle batteries based on multidimensional data analysis according to claim 1 is characterized in that: The calculation method of the monthly voltage consistency and temperature consistency of the vehicle single battery includes: Obtaining a voltage data set and a temperature data set of the single battery of the vehicle at all times on the day; The extreme values ​​of the voltage and temperature of the single battery at each moment are obtained respectively, and the root mean square error values ​​of the extreme values ​​of the voltage and temperature are calculated respectively. The root mean square error value satisfies the calculation formula: in, The total number of voltage and temperature data of the single battery at all times of the vehicle on that day; They are Maximum voltage and temperature at a given moment; They are Minimum voltage and temperature at a given moment.

3. The method for evaluating the health status of new energy vehicle batteries based on multidimensional data analysis according to claim 2 is characterized in that: The method for calculating the monthly voltage consistency of the vehicle single battery also includes: Calculate the average value of the vehicle's voltage data at all times of the day; The voltage variation coefficient is calculated according to the average value of the voltage data to evaluate the discrete situation of the voltage data. The voltage variation coefficient satisfies the calculation formula: in, The root mean square error of the voltage extreme value, It is the average value of the vehicle's voltage data at all times of the day.

4. The method for evaluating the health status of a new energy vehicle battery based on multidimensional data analysis according to claim 3 or 1, characterized in that: The calculation method of the monthly internal resistance consistency of the vehicle single battery includes: Get the first and last data of the vehicle's complete charging period; Compare the voltage data difference and current data difference at the time corresponding to the first and last data to obtain the internal resistance change set of the single battery The average value of internal resistance change , calculate the standard deviation of the internal resistance change of the single battery, and the standard deviation of the internal resistance change satisfies the calculation formula: in, is the number of batteries in the vehicle battery pack; The internal resistance variation coefficient is calculated according to the average value of the internal resistance data to evaluate the discreteness of the internal resistance data. The internal resistance variation coefficient satisfies the calculation formula: in, is the standard deviation of the internal resistance variation.

5. The method for evaluating the health status of new energy vehicle batteries based on multidimensional data analysis according to claim 1 is characterized in that: The method for calculating the capacity retention rate of the vehicle single battery includes: The energy storage data of the constant current charging segment of the vehicle battery for more than 200s is obtained to obtain the charge capacity of the battery per unit time. The SOC of the constant current charging segment of the vehicle battery extracted must be in the range of 40% to 90%, and the charge capacity of the battery per unit time satisfies the formula: in, is the amount of electricity charged into the battery per unit time. for The current magnitude at that moment, is the time difference between two frames of data; The required capacity when the SOC is 100% is obtained based on the ratio of the amount of electricity charged into the battery per unit time to the SOC change. , the capacity Satisfy the calculation formula: in, is the SOC change value within the constant current charging segment of the vehicle battery; According to the capacity The capacity retention rate of the vehicle single battery is calculated, and the capacity retention rate of the vehicle single battery satisfies the calculation formula: in, is the factory rated voltage of the vehicle battery, The total storage capacity of the vehicle battery set at the factory.

6. The method for evaluating the health status of new energy vehicle batteries based on multidimensional data analysis according to claim 1 is characterized in that: In S2, the index data is standardized to satisfy the formula: in, is the mean of the original data, is the standard deviation of the original data; In S2, the index data is normalized to satisfy the formula: in, Represents the maximum and minimum values ​​of the data.

7. The method for evaluating the health status of new energy vehicle batteries based on multidimensional data analysis according to claim 1 is characterized in that: In S3, a loss function is set to perform pre-training and cross-training on the linear model, wherein the loss function satisfies the formula: in, For the The true value of the samples, For the The predicted value of samples, is the total number of samples, for the ultimate loss; In order to make the To minimize, the loss function is optimized by gradient descent, wherein the gradient descent satisfies the formula: in, is the parameter vector, is the learning rate, which is used to determine the step size of each update of the linear model. For the The predicted value of samples, For the The true value of the samples, is the total number of samples.

Images