Method for predicting cycle life of mine sweeping type battery cell based on big data analysis

Abstract

The invention particularly relates to a mine sweeping type battery cell cycle life prediction method based on big data analysis in the technical field of battery cell performance prediction. The method comprises the following steps: S1, data reading: acquiring cycle data in an initial stage; s2, determining a control equation: fitting the loop data in the initial stage to obtain a preliminary attenuation model equation; s3, training a model: obtaining a training model equation by fitting coefficients; s4, data scanning: carrying out data point scanning on the whole period of the battery; s5, model prediction: adding a slow curve and a high-order term into various change rates of the early-stage data to obtain a prediction capacity model; s6, drawing a graph: drawing a prediction curve and a prediction range scatter diagram; and S7, performance evaluation: using the number of cycles corresponding to the capacity of 80% SOH as an evaluation factor to obtain a corresponding accuracy rate. The method has the beneficial effect that the cycle life of the battery cell can be accurately predicted through early-stage cycle data analysis of the battery cell.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery cell performance prediction, and specifically relates to a minesweeping battery cell cycle life prediction method based on big data analysis. Background Art

[0002] A battery cell usually refers to a single smallest electrical energy storage unit containing positive and negative electrodes. It is the basic electrochemical unit that makes up a battery. It is generally not used directly alone. In most cases, multiple battery cells are packaged to form a battery module. Several modules are then controlled or managed by the BMS and thermal management system to form a battery pack. Batteries usually refer to battery packs. Batteries have a certain cycle life. The cycle life of lithium battery cells is generally not less than 2,000 times. The cycle life of specially designed lithium battery cells can even reach tens of thousands of times. Cycle life prediction is of great significance to the production and use of lithium battery cells. It has long been an important research topic for technicians in the lithium battery field. At present, it is limited by the small amount of lithium battery cell cycle data, and the accuracy and breadth of cell cycle life model prediction still need to be improved. Summary of the invention

[0003] In view of the above situation, the present invention provides a minesweeping type battery cell cycle life prediction method based on big data analysis. Through the analysis of the early cycle data of the battery cell, it can more accurately predict the number of cycles corresponding to the battery cell capacity reaching 80% SOH charge state.

[0004] In order to achieve the above objectives, the present invention provides a minesweeping type battery cell cycle life prediction method based on big data analysis, comprising the following steps:

[0005] S1. Reading data: Obtaining 20-200 cycle data of the battery, the cycle data including the number of cycles and the discharge capacity, and obtaining the cycle data of the initial stage after cleaning, screening and averaging;

[0006] S2. Determine the control equation: Use a polynomial to fit the cycle data of the initial stage in S1 to obtain the polynomial equation of the capacity decay model in the initial stage:

[0007] f(x)=ax 3 +bx 2 +cx+d (1)

[0008] Where: a, b, c are polynomial coefficients, x is the number of cycles, and d is the initial capacity of the battery cell;

[0009] S3. Training model: By fitting the coefficients of the cycle number and cell capacity of the data in S1 using the equation in S2, the training model equation is obtained:

[0010] f′(x)=a′x 3+b′x 2 +c′x+d′ (2)

[0011] The fitting accuracy was controlled by adding the root mean square error (RMSE) and R-square (R2);

[0012] S4. Data scanning: Scan the data points of the entire battery cycle to ensure that the predicted data of each stage falls within the following range: a′* early stage, lower limit 0.4, upper limit 0.8, step length 0.05; b′* mid-term, lower limit 0.1, upper limit 0.105, step length 0.001; c′* late stage, lower limit 0.01, upper limit 0.2, step length 0.01;

[0013] S5. Prediction model: Use various change rates of previous data, add slow curves and high-order terms for prediction. The specific method is as follows:

[0014] S5-1. First, make a difference on the data in S1 to obtain the rate of change of capacity with the number of cycles, and take the maximum value qmax, the minimum value qmin, and the average value qmean;

[0015] S5-2. Then make it into a one-variable multi-time equation:

[0016] y=Q+q max *x 0.99 +q mean *x 0.1 (3)

[0017] Among them, y is the predicted capacity, Q is the initial cycle capacity, and x is the number of cycles after the initial cycle;

[0018] S6. Drawing a graph: Through the above steps in S5, a prediction curve and a prediction range scatter plot can be drawn;

[0019] S7. Performance evaluation: Using the number of cycles corresponding to the capacity of 80% SOH as an evaluation factor, the corresponding accuracy can be obtained.

[0020] Preferably, in step S1, the cycle data of the initial stage can be further reduced to 50-100 cycles to further optimize the acquisition of the cycle data of the initial stage.

[0021] Specifically, the training model equation (2) in step S3 can be understood as the prediction model equation obtained by adjusting equation (1) by scaling factor z:

[0022] f′(x)=z×(ax 3 +bx 2 +cx+d) (4)

[0023] By changing the z value, the curve changes in different stages can be adjusted to predict the capacity attenuation range.

[0024] Furthermore, based on the prediction model equation (4), the target number of cycles can be input into the prediction model to output the capacity data predicted by the target number of cycles.

[0025] Furthermore, the model prediction accuracy can be output by comparing the predicted capacity data with the actual data.

[0026] Furthermore, the progress is improved by modifying the scaling factor library in the prediction model equation, and repeated iterations are performed to increase the number of prediction target cycles, draw a complete cycle curve, and then form a range of predicted capacity decay, and on this basis, seek the diving point and inflection point of battery cell capacity decay.

[0027] The present invention also includes other components that enable it to be used normally, all of which are conventional means in the art. In addition, devices or components not limited in the present invention all adopt the existing technology in the art.

[0028] The beneficial effects of the present invention are as follows:

[0029] The minesweeping type battery cell cycle life prediction method based on big data analysis provided by the present invention can adjust the accuracy of the prediction model through a scaling factor, and can quantitatively consider the pros and cons of the prediction model; it can have a certain expected effect on the later data in the early stage of the battery cell cycle, especially for the battery cell with poor cycle performance. The identification effect is better and the research and development cycle is shortened; as the amount of battery cell cycle life data increases, a basic database for the battery cell can be established, which is of great significance to the later research and development. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the flow chart of the method for predicting the cycle life of a battery cell in the present invention.

[0031] Figure 2 It is a curve diagram of the battery capacity retention rate corresponding to the number of cycles in the present invention.

[0032] Figure 3 This is a curve chart of the prediction results of the battery cell cycle life attenuation model in the present invention. DETAILED DESCRIPTION

[0033] The technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0034] Example 1

[0035] like Figure 1 As shown, the minesweeping type battery cell cycle life prediction method based on big data analysis includes the following steps:

[0036] S1. Reading data: Obtaining 20-200 cycle data of the battery, the cycle data including the number of cycles and the discharge capacity, and obtaining the cycle data of the initial stage after cleaning, screening and averaging;

[0037] S2. Determine the control equation: Use a polynomial to fit the cycle data of the initial stage in S1 to obtain the polynomial equation of the capacity decay model in the initial stage:

[0038] f(x)=ax 3 +bx 2 +cx+d (1)

[0039] Where: a, b, c are polynomial coefficients, x is the number of cycles, and d is the initial capacity of the battery cell;

[0040] S3. Training model: By fitting the coefficients of the cycle number and cell capacity of the data in S1 using the equation in S2, the training model equation is obtained:

[0041] f′(x)=a′x 3 +bx 2 ++c′x+d′ (2)

[0042] The fitting accuracy was controlled by adding the root mean square error (RMSE) and R-square (R2);

[0043] S4. Data scanning: Scan the data points of the entire battery cycle to ensure that the predicted data of each stage falls within the following range: a′* early stage, lower limit 0.4, upper limit 0.8, step length 0.05; b′* mid-term, lower limit 0.1, upper limit 0.105, step length 0.001; c′* late stage, lower limit 0.01, upper limit 0.2, step length 0.01;

[0044] S5. Prediction model: Use various change rates of previous data, add slow curves and high-order terms for prediction. The specific method is as follows:

[0045] S5-1. First, make a difference on the data in S1 to obtain the rate of change of capacity with the number of cycles, and take the maximum value qmax, the minimum value qmin, and the average value qmean;

[0046] S5-2. Then make it into a one-variable multi-time equation:

[0047] y=Q+q max *x 0.99 +q mean *x 0.1 (3)

[0048] Where y is the predicted capacity, Q is the capacity at the 50th cycle, and x is the number of cycles after 50 cycles;

[0049] S6. Drawing graphs: Through the above steps in S5, a prediction curve and a prediction range scatter plot can be drawn, such as Figure 2 As shown;

[0050] S7. Performance evaluation: Using the number of cycles corresponding to the capacity of 80% SOH as an evaluation factor, the corresponding accuracy can be obtained.

[0051] Example 2

[0052] Based on Example 1, step S1 preferably obtains 50-100 cycles of cycle data in the initial stage to further optimize the acquisition of cycle data in the initial stage.

[0053] Specifically, the training model equation (2) in step S3 can be understood as the prediction model equation obtained by adjusting equation (1) by scaling factor z:

[0054] f'(x)=z×(ax 3 +bx 2 +cx+d) (4), by changing the z value, the curve changes in different stages can be adjusted to predict the capacity attenuation range.

[0055] On the basis of the prediction model equation (4), the target number of cycles can be input into the prediction model to output the capacity data predicted by the target number of cycles; and then the predicted capacity data can be compared with the actual data to output the model prediction accuracy.

[0056] Then, the progress is improved by modifying the scaling factor library in the prediction model equation (4), and repeated iterations are performed to increase the number of prediction target cycles, draw a complete cycle curve, and then form a range of predicted capacity decay. On this basis, the cell capacity decay diving point and inflection point are sought, such as Figure 3 shown.

[0057] The accuracy of the prediction model can be controlled by the scaling factor, and the pros and cons of the prediction model can be quantitatively considered. In the early stage of the battery cell cycle, it can have a certain expected effect on the later data, especially for the identification of batteries with poor cycle performance, thus shortening the research and development cycle. As the amount of battery cell data increases, a basic database for the battery cells can be established, which is of great significance to the later research and development.

[0058] The technical solution of the present invention is not limited to the above-mentioned specific embodiments. Without departing from the scope and spirit of the described embodiments, many modifications and changes are obvious to ordinary technicians in this technical field. Any technical deformation made within the spirit and principles of the present invention falls within the protection scope of the present invention.

Claims

1. A minesweeping battery cell cycle life prediction method based on big data analysis, characterized in that: The following steps are involved: S1. Reading data: Obtaining 20-200 cycle data of the battery, the cycle data including the number of cycles and the discharge capacity, and obtaining the cycle data of the initial stage after cleaning, screening and averaging; S2. Determine the control equation: Use a polynomial to fit the cycle data of the initial stage in S1 to obtain the polynomial equation of the capacity decay model in the initial stage: f(x)=ax 3 +bx 2 +cx+d (1) Where: a, b, c are polynomial coefficients, x is the number of cycles, and d is the initial capacity of the battery cell; S3. Training model: By fitting the coefficients of the cycle number and cell capacity of the data in S1 using the equation in S2, the training model equation is obtained: f′(x)=a′x 3 +b′x 2 +c′x+d′ (2) The fitting accuracy was controlled by adding the root mean square error and R-square; S4. Data scanning: Scan the data points of the entire battery cycle to ensure that the predicted data of each stage falls within the following range: a′* early stage, lower limit 0.4, upper limit 0.8, step length 0.05; b′* mid-term, lower limit 0.1, upper limit 0.105, step length 0.001; c′* late stage, lower limit 0.01, upper limit 0.2, step length 0.01; S5. Prediction model: Use various change rates of previous data, add slow curves and high-order terms for prediction. The specific method is as follows: S5-1. First, make a difference on the data in S1 to obtain the rate of change of capacity with the number of cycles, and take the maximum value qmax, the minimum value qmin, and the average value qmean; S5-2. Then make it into a one-variable multi-time equation: y=Q+q max *x 0.99 +q mean *x 0.01 (3) Among them, y is the predicted capacity, Q is the initial cycle capacity; S6. Drawing a graph: Through the above steps in S5, a prediction curve and a prediction range scatter plot can be drawn; S7. Performance evaluation: Using the number of cycles corresponding to the capacity of 80% SOH as an evaluation factor, the corresponding accuracy can be obtained.

2. The minesweeping type battery cell cycle life prediction method based on big data analysis according to claim 1 is characterized in that: In step S1, 50-100 cycles of cycle data in the initial stage are optimized to further optimize the cycle data in the initial stage.

3. The minesweeping type battery cell cycle life prediction method based on big data analysis according to claim 1 is characterized in that: The training model equation (2) in step S3 can be understood as the prediction model equation obtained by adjusting equation (1) by scaling factor z: f′(x)=z×(ax 3 +bx 2 +cx+d) (4) By changing the z value, the curve changes in different stages can be adjusted to predict the capacity attenuation range.

4. The minesweeping type battery cell cycle life prediction method based on big data analysis according to claim 3 is characterized in that: On the basis of obtaining the prediction model equation, the target number of cycles can be input into the prediction model to output the capacity data predicted by the target number of cycles.

5. The minesweeping type battery cell cycle life prediction method based on big data analysis according to claim 4 is characterized in that: The model prediction accuracy can be output by comparing the predicted capacity data with the actual data.

6. The minesweeping type battery cell cycle life prediction method based on big data analysis according to claim 5 is characterized in that: The progress is improved by modifying the scaling factor library in the prediction model equation, and repeated iterations are used to refine the prediction target number of cycles, draw a complete cycle curve, and then form a range of predicted capacity attenuation, and on this basis, find the diving point and inflection point of battery cell capacity attenuation.

Images