Battery charging and discharging curve prediction method and system based on capacity estimation

Abstract

The invention relates to a battery charge and discharge curve prediction method and system based on capacity estimation. The method comprises the following steps: obtaining discharge capacity, a complete charge curve and a complete discharge curve of a plurality of batteries of a to-be-detected type; discretizing the complete charging curve to obtain charging curve data, and dividing the charging curve data into low-voltage interval data and high-voltage interval data; discretizing the complete discharge curve to obtain discharge curve data; constructing a training set by using the low-voltage interval data and the discharge capacity, and training to obtain a capacity prediction model; inputting the low-voltage interval data into a capacity prediction model to obtain battery capacity estimation data; based on the low-voltage interval data, the high-voltage interval data, the battery capacity estimation data and the discharge curve data, constructing a charge and discharge curve prediction model through deep learning; and predicting a complete charge and discharge curve of the to-be-tested battery by using a total model formed by the capacity prediction model and the charge and discharge curve prediction model. The battery characteristics can be better captured, and the accuracy of the prediction result is high.

Description

Technical Field

[0001] The present invention relates to the technical field of battery performance evaluation, and in particular to a method and system for predicting battery charge and discharge curves based on capacity estimation. Background Art

[0002] The battery charge and discharge curve records the changing relationship between voltage, current and time during the battery charging process, reflecting the dynamic response characteristics and internal electrochemical reaction process of the battery during the charging process. By analyzing the charge and discharge curve, the battery's health status, charging efficiency, actual capacity and performance degradation can be evaluated, providing key data support for the battery management system. The traditional method usually collects data through a full charge and discharge test, that is, the battery is fully discharged and then the complete charge and discharge process is performed to obtain the charge and discharge curve. The disadvantages are: the test time is long, usually it takes several hours to complete a complete charge and discharge cycle, which seriously affects production efficiency. In addition, large-scale full charge and discharge tests on industrial production lines are not only time-consuming, but also increase energy consumption and testing costs. Frequent charging and discharging accelerates the degradation of battery performance and has an adverse effect on battery life.

[0003] In response to the above shortcomings, an artificial intelligence model is used to establish a mapping relationship between data fragments and complete charging curves, and the actual charging data fragments are input into the trained model to reconstruct the complete charging curve. This method based on an artificial intelligence model can greatly improve production efficiency and avoid the adverse effects of frequent charging and discharging on the battery. However, the following problems still exist: data under different charging methods and different aging conditions need to be collected, and the required data collection cost is very high. The model is established only based on a simple mapping relationship, and the training data required for model construction lacks significant features with prior knowledge characteristics. And the lack of verification and analysis leads to the model's limited understanding of the physical meaning of the charging curve, and the prediction results are not easy for engineers to understand and trust. Summary of the invention

[0004] In view of the deficiencies of the prior art, the present invention provides a battery charge and discharge curve prediction method and system based on capacity estimation, the purpose of which is to improve the prediction model so that it can better capture the battery characteristics, thereby improving the accuracy of the prediction results.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The present invention provides a battery charge and discharge curve prediction method based on capacity estimation, the method comprising:

[0007] S1. Obtain discharge capacity, complete charging curve and complete discharging curve of several batteries of the same type to be tested;

[0008] S2. Sampling the complete charging curve at set voltage intervals to obtain discrete charging curve data, dividing the charging curve data into low voltage interval data and high voltage interval data, wherein the voltage of the low voltage interval data is lower than the voltage of the high voltage interval data; sampling the complete discharge curve at set voltage intervals to obtain discrete discharge curve data;

[0009] S3, constructing a first data set with the low voltage interval data and the discharge capacity, and obtaining a capacity prediction model by machine learning using the first data set; inputting the low voltage interval data into the capacity prediction model to obtain battery capacity estimation data;

[0010] S4. Based on the low voltage interval data, high voltage interval data, battery capacity estimation data and discharge curve data, a charge and discharge curve prediction model is constructed through deep learning, which is used to establish a mapping relationship between the discharge capacity and low voltage interval data and the complete charge and discharge curve data;

[0011] S5, combining the capacity prediction model and the charge-discharge curve prediction model into an overall model;

[0012] S6. Using the overall model to predict the complete charge and discharge curve of the battery to be tested.

[0013] Further technical solutions are:

[0014] In step S4, the construction of the charge-discharge curve prediction model includes:

[0015] Using the low voltage interval data and the battery capacity estimation data as independent variables and the high voltage interval data as dependent variables, construct a second data set, and use the second data set to train a deep learning model to obtain a charging curve prediction model;

[0016] A third data set is constructed with the low voltage interval data and the battery capacity estimation data as independent variables and the discharge curve data as dependent variables, and a deep learning model is trained using the third data set to obtain a discharge curve prediction model.

[0017] In step S6, the overall model is used to predict the complete charge and discharge curve of the battery to be tested, including predicting the complete charge curve, which includes:

[0018] Obtaining low voltage interval data collected during the charging process of the battery to be tested;

[0019] Inputting the collected low voltage interval data into the capacity prediction model to obtain a capacity estimation value;

[0020] Inputting the collected low voltage interval data and the estimated capacity value into the charging curve prediction model to obtain predicted high voltage interval data;

[0021] The collected low voltage interval data and the predicted high voltage interval data are spliced ​​to obtain complete charging curve data.

[0022] In step S6, the overall model is used to predict the complete charge and discharge curve of the battery to be tested, including predicting the complete discharge curve, which includes:

[0023] Inputting the collected low voltage interval data into the capacity prediction model to obtain a capacity estimation value;

[0024] The collected low voltage interval data and capacity estimation value are input into the discharge curve prediction model to obtain complete discharge curve data.

[0025] The method further includes updating the model, wherein the updating the model includes:

[0026] The battery management system is used to collect the complete charge and discharge curve data and the discharge capacity of the battery during operation.

[0027] Merging the collected data with the first data set to form new training data;

[0028] The new training data is used to update the capacity prediction model, the discharge curve prediction model and the charge curve prediction model, thereby updating the overall model.

[0029] In step S1, the discharge capacity, complete charging curve and complete discharging curve of several batteries of the same type to be tested are obtained in the following manner:

[0030] Discharge all batteries at a constant current to the lower cut-off voltage to ensure that the batteries are completely discharged;

[0031] All batteries are fully charged and discharged. The charging process is constant current charging first, then constant voltage charging, and the discharging process is constant current discharging.

[0032] Record the current, voltage and time data of each battery charging and discharging process;

[0033] Extract the discharge curve of the complete discharge process of the battery;

[0034] Based on the current and time data of the complete discharge process of the battery, the discharge capacity of each battery is calculated using the ampere-hour integration method;

[0035] Extract the charging curve of the battery's complete charging process.

[0036] In step S2, the charging curve data is divided into low voltage interval data and high voltage interval data, including:

[0037] Set a voltage value V0;

[0038] Classify the interval less than or equal to V0 in the charging curve data as low voltage interval data;

[0039] The interval greater than V0 in the charging curve data is classified as high voltage interval data.

[0040] In step S3, random forest regression, support vector machine or neural network is used to perform the machine learning.

[0041] In step S4, a convolutional neural network or a long short-term memory network is used to perform the deep learning.

[0042] The present invention also provides a battery charge and discharge curve prediction system based on capacity estimation, and the system is used to execute the method described.

[0043] The beneficial effects of the present invention are as follows:

[0044] The prediction strategy of the method of the present invention is to first estimate the capacity, and then use the estimated capacity value as a display feature combined with the low voltage interval data to predict the complete curve. The estimated capacity value provides more valuable prior information to the charge and discharge curve prediction model, so that the model can better understand the physical meaning of the curve, so as to better capture the battery characteristics. The present invention has the following specific advantages:

[0045] (1) The method of the present invention only needs to collect charging fragment data in the low voltage range of the battery, such as data of about 30 minutes, and use the low voltage range data to predict the battery capacity. That is, the complete charging curve can be accurately reconstructed through the fragment data and capacity estimation strategy, which greatly improves the test efficiency, avoids the time-consuming full charge and discharge test and the loss of the battery, and is more suitable for industrial-scale application.

[0046] (2) Compared with the existing methods, the present invention is firstly simple and intuitive in conception (first predict the key capacity parameters, and then predict the complete curve based on them), has higher computational efficiency (only one overall model is needed), has lower model dependence (no need for multiple models to work together, reducing system complexity), and has better consistency in prediction results (predicting the complete charge and discharge curve based on the core parameter of capacity). Secondly, a capacity estimate is provided as an intermediate result for verification and analysis, and the prediction results of the curve model are easier for engineers to understand and trust. In addition, the key feature of capacity can be displayed and applied to reduce error propagation and obtain a more accurate charge and discharge curve. Thirdly, there is no need to collect data under different charging methods and different aging conditions, which greatly reduces the cost and time of data collection.

[0047] (3) The present invention collects complete charge and discharge curve data during the operation of the battery to be tested and updates the first data set of the discharge capacity of the battery through the battery management system, thereby updating the capacity prediction model, the discharge curve prediction model and the charging curve prediction model, realizing adaptive updating of the model according to the working state of the battery, and ensuring the prediction accuracy of the model.

[0048] (4) The present invention can integrate the two-stage models of capacity prediction and curve prediction into an end-to-end overall model by constructing a convolutional network, which predicts capacity in the early stage and predicts the curve in the later stage. The model structure is simple and easy to construct.

[0049] (5) The present invention is applicable to the prediction of charge and discharge curves of single cells and battery packs of various electrode materials such as lithium iron phosphate batteries, ternary lithium batteries, sodium ion batteries, lead-acid batteries, lithium sulfur batteries, etc. It can meet the needs of extending battery life, calculating driving range, and fault warning.

[0050] Other features and advantages of the present invention will be set forth in the following description or may be learned by practicing the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the process of the embodiment of the present invention.

[0052] Figure 2 This is a curve diagram of voltage variation during full charge and discharge test of a battery according to an embodiment of the present invention.

[0053] Figure 3 The diagram is a schematic diagram of the interval division in which the low voltage interval data and the high voltage interval data are respectively used as the input of the capacity prediction model and the charge and discharge curve prediction model, and the output of the charge curve prediction model in an embodiment of the present invention.

[0054] Figure 4 This is a comparison chart of the discharge capacities of eight batteries under test obtained based on full charge and discharge tests and the capacity estimation values ​​obtained based on the capacity prediction model in an embodiment of the present invention.

[0055] Figure 5 The figure is a comparison result between the complete charging curve predicted by the method of the embodiment of the present invention and the actual charging curve.

[0056] Figure 6 A schematic diagram of the voltage error between the complete charging curve predicted by the method of an embodiment of the present invention and the actual charging curve.

[0057] Figure 7 The figure is a comparison result between the complete discharge curve predicted by the method of the embodiment of the present invention and the actual discharge curve.

[0058] Figure 8Schematic diagram of the voltage error between the complete discharge curve predicted by the method of the embodiment of the present invention and the actual discharge curve.

[0059] Fig. 9 The figure shows the comparison of the prediction time required by the prediction method according to the embodiment of the present invention and the curve generation method based on full filling and discharge. DETAILED DESCRIPTION

[0060] The specific implementation of the present invention is described below with reference to the accompanying drawings.

[0061] See also Figure 1 The present application provides a battery charge and discharge curve prediction method based on capacity estimation, the method comprising:

[0062] S1. Conduct capacity tests based on full charge and discharge on the battery to obtain the discharge capacity, complete charge curve and complete discharge curve of several batteries of the same type to be tested. Specifically including:

[0063] Discharge all batteries at a constant current to the lower cut-off voltage to ensure that the batteries are completely discharged;

[0064] All batteries are fully charged and discharged. The charging process is constant current charging first, then constant voltage charging, and the discharging process is constant current discharging.

[0065] Record the current, voltage and time data of each battery charging and discharging process;

[0066] Extract the discharge curve of the complete discharge process of the battery;

[0067] Based on the current and time data of the complete discharge process of the battery, the discharge capacity of each battery is calculated using the ampere-hour integration method;

[0068] Extract the charging curve of the battery's complete charging process.

[0069] During the test, the cut-off voltage and cut-off current are determined by the battery manufacturer.

[0070] As a specific implementation, 100 lithium iron phosphate batteries with a rated capacity of 106Ah were selected as test objects, and the battery capacity test based on full charge and discharge was performed on the battery. The typical voltage change is as follows: Figure 2 shown.

[0071] S2. Sampling the complete charging curve at set voltage intervals to obtain discrete charging curve data, dividing the charging curve data into low voltage interval data and high voltage interval data, wherein the voltage of the low voltage interval data is less than the voltage of the high voltage interval data; sampling the complete discharge curve at set voltage intervals to obtain discrete discharge curve data.

[0072] Dividing the charging curve data into low voltage interval data and high voltage interval data includes:

[0073] Set a voltage value V0;

[0074] Classify the interval less than or equal to V0 in the charging curve data as low voltage interval data;

[0075] The interval greater than V0 in the charging curve data is classified as high voltage interval data.

[0076] The voltage value V0 for dividing the low voltage interval and the high voltage interval can be selected from any voltage point on the complete charging curve. As a specific implementation, the selected voltage value V0 = 3.30V divides the data in the voltage value interval of 2.50V-3.30V in the complete charging curve data into low voltage interval data, and divides the data in the voltage value interval of 3.31V-3.65V into high voltage interval data.

[0077] S3. Construct a first data set with the low voltage interval data and the discharge capacity, and use the first data set to obtain a capacity prediction model through machine learning; input the low voltage interval data into the capacity prediction model to obtain battery capacity estimation data.

[0078] Specifically, the low voltage interval data of each battery in the first data set is used as the dependent variable, and the corresponding discharge capacity is used as the dependent variable. Models such as random forest regression, support vector machine or neural network are used for training, learning and testing to obtain a capacity prediction model.

[0079] S4. Based on the low voltage interval data, high voltage interval data, battery capacity estimation data and discharge curve data, a charge and discharge curve prediction model is constructed through deep learning, which is used to establish a mapping relationship between the discharge capacity and low voltage interval data and the complete charge and discharge curve data.

[0080] The construction of the charge-discharge curve prediction model includes:

[0081] Using the low voltage interval data and the battery capacity estimation data as independent variables and the high voltage interval data as dependent variables, construct a second data set, and use the second data set to train a deep learning model to obtain a discharge curve prediction model;

[0082] A third data set is constructed with the low voltage interval data and the battery capacity estimation data as independent variables and the discharge curve data as dependent variables, and a deep learning model is trained using the third data set to obtain a charging curve prediction model.

[0083] The deep learning model may adopt a convolutional neural network or a long short-term memory network, etc.

[0084] like Figure 3 As shown, the low voltage interval data is used as the input of the capacity prediction model and the charge / discharge curve prediction model, and the output of the charge curve prediction model is the high voltage interval data.

[0085] S5. The capacity prediction model and the discharge curve prediction model are combined into an overall model.

[0086] S6. Predicting the complete charge and discharge curve of the battery to be tested using the overall model, including:

[0087] S61, predicting the complete charging curve, which includes:

[0088] (1) Obtaining low voltage interval data collected during the charging process of the battery to be tested; it can be understood that the voltage interval of the collected low voltage interval data is consistent with the voltage interval of the low voltage interval data used for training when constructing the model training, and the voltage interval is also 2.5V-3.3V.

[0089] (2) Inputting the collected low voltage interval data into the capacity prediction model to obtain a capacity estimation value.

[0090] (3) The collected low voltage interval data and capacity estimation value are input into the charging curve prediction model to obtain predicted high voltage interval data.

[0091] Figure 4 The comparison chart of the discharge capacity obtained based on the full charge and discharge test (i.e., "full charge and discharge" in the figure) and the capacity estimation value obtained based on the capacity prediction model (i.e., "proposed method" in the figure) of 8 batteries under test is shown. It can be seen that the capacity estimation value is very close to the capacity value obtained based on full charge and discharge. The average absolute percentage error is 0.45%, and the maximum absolute percentage error is 1.33%, achieving a high capacity estimation accuracy.

[0092] The present invention first predicts the most critical capacity parameter of the battery, and then adds the capacity parameter as a display feature to the charge and discharge curve prediction, providing more valuable prior information to the charge and discharge curve prediction model.

[0093] (4) splicing the collected low voltage interval data and the predicted high voltage interval data to obtain complete charging curve data.

[0094] S62, predicting the complete discharge curve, which includes:

[0095] Inputting the collected low voltage interval data into the capacity prediction model to obtain a capacity estimation value;

[0096] The collected low voltage interval data and capacity estimation value are input into the discharge curve prediction model to obtain complete discharge curve data.

[0097] It can be understood that in this embodiment, the capacity prediction model and the charge and discharge curve prediction model constitute a total model. When the total model is applied, it is only necessary to input the low voltage interval data collected during the charging process of the battery to be tested into the total model, and the complete charging curve data and the complete discharge curve data can be automatically output. The intermediate data including the capacity estimation value is circulated through the automatic operation of the model, and the intermediate process including the splicing of the low voltage interval data and the high voltage interval data can be automatically realized through the model algorithm.

[0098] In specific implementation, the two-stage models of capacity prediction and curve prediction can be integrated into an end-to-end overall model. For example, a convolutional network can be used to build an overall model to predict capacity in the early stage and curve in the later stage.

[0099] Figure 5 The comparison results of the predicted complete charging curve and the actual charging curve in the charging stage are shown. It can be seen that the complete charging curve obtained based on capacity estimation in this embodiment ("Charging Curve 1" in the figure) has a higher degree of fit with the actual charging curve than the predicted curve obtained by the common prediction method without capacity estimation ("Charging Curve 2" in the figure). And in the stage of rapid voltage rise, the trend of charging curve 1 is consistent with the actual curve. It shows that the method of this embodiment captures the characteristic changes of battery charging well.

[0100] Figure 6 It is a schematic diagram of the error between the predicted complete charging curve and the actual charging curve. At most time points, the average voltage error of charging curve 2 obtained by the common prediction method without capacity estimation is 10mV, and the maximum absolute error is 190mV. However, the average voltage error of charging curve 1 obtained by the method of this embodiment is only 3mV, and the maximum absolute error is only 20mV, which further proves the accuracy of the method of this embodiment.

[0101] Figure 7 The comparison results of the predicted complete discharge curve and the actual discharge curve in the discharge stage are shown. It can be seen that the complete discharge curve obtained by the embodiment based on capacity estimation ("discharge curve 1" in the figure) has a higher degree of fit with the actual discharge curve than the predicted curve obtained by the ordinary prediction method without capacity estimation ("discharge curve 2" in the figure). And in the stage of rapid voltage drop, the trend of discharge curve 1 is consistent with that of the actual discharge curve. This shows that the method of this embodiment captures the characteristic changes of battery discharge well.

[0102] Figure 8It is a schematic diagram of the error between the predicted complete discharge curve and the actual discharge curve. At most time points, the average voltage error of the discharge curve 2 obtained by the common prediction method without capacity estimation is 13mV, and the maximum absolute error is 193mV. However, the average voltage error of the discharge curve 1 obtained by the method of this embodiment is only 9mV, and the maximum absolute error is only 68mV, which further proves the accuracy of the method of this embodiment.

[0103] Time comparison of the charge and discharge curve prediction method and the curve generation method based on full charge and discharge in the embodiment of the present invention Fig. 9 As shown. The method based on full charge and discharge takes an average of 190 minutes, while the curve prediction method proposed in the embodiment of the present invention takes an average of 39 minutes, which is more than 4 times more efficient. Therefore, the prediction method in the embodiment of the present invention can greatly shorten the battery detection time while achieving high-precision curve prediction.

[0104] As a further improvement, the battery charge and discharge curve prediction method based on capacity estimation in this embodiment further includes:

[0105] S7. Update model:

[0106] The battery management system is used to collect the complete charge and discharge curve data and the discharge capacity of the battery during operation.

[0107] Merging the collected data with the first data set to form new training data;

[0108] The new training data is used to update the capacity prediction model, the charging curve prediction model and the discharging curve prediction model, thereby updating the overall model.

[0109] This enables the model to be adaptively updated according to the battery operating status.

[0110] It is understood that the capacity prediction model is not limited to any machine learning or deep learning model. The charge and discharge curve prediction model is not limited to any machine learning or deep learning model.

[0111] The method of this embodiment only uses fragmentary information of the battery charging process to predict the charge and discharge curve of the entire process, which greatly improves the test efficiency and reduces the loss of the battery. It solves the problem that the traditional charge and discharge curve prediction method requires the battery to be fully charged and discharged, which takes a long time, accelerates the degradation of battery performance, and shortens the battery life.

[0112] The method of this embodiment is of great significance for battery performance evaluation and management. By analyzing the charge and discharge curves, the capacity state and health of the battery can be evaluated to help optimize the charging strategy to balance the charging speed and battery life. These data provide key support for the development of battery management systems, enabling them to more accurately estimate battery status, predict life, and detect potential problems in a timely manner. And it is not only suitable for predicting lithium iron phosphate batteries, but also for ternary lithium batteries or sodium ion batteries, lead-acid batteries, lithium-sulfur batteries, etc. It can meet the needs of various application scenarios such as mileage calculation, fault warning, and extended battery life.

[0113] This embodiment also provides a battery charge and discharge curve prediction system based on capacity estimation, and the system is used to execute the battery charge and discharge curve prediction method based on capacity estimation.

[0114] Those skilled in the art can understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention is described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A battery charge and discharge curve prediction method based on capacity estimation, characterized in that: The method comprises: S1. Obtain discharge capacity, complete charging curve and complete discharging curve of several batteries of the same type to be tested; S2. Sampling the complete charging curve at set voltage intervals to obtain discrete charging curve data, dividing the charging curve data into low voltage interval data and high voltage interval data, wherein the voltage of the low voltage interval data is lower than the voltage of the high voltage interval data; sampling the complete discharge curve at set voltage intervals to obtain discrete discharge curve data; S3, constructing a first data set with the low voltage interval data and the discharge capacity, and obtaining a capacity prediction model by machine learning using the first data set; inputting the low voltage interval data into the capacity prediction model to obtain battery capacity estimation data; S4. Based on the low voltage interval data, high voltage interval data, battery capacity estimation data and discharge curve data, a charge and discharge curve prediction model is constructed through deep learning, which is used to establish a mapping relationship between the discharge capacity and low voltage interval data and the complete charge and discharge curve data; S5, combining the capacity prediction model and the charge-discharge curve prediction model into an overall model; S6. Using the overall model to predict the complete charge and discharge curve of the battery to be tested.

2. The method according to claim 1, characterized in that In step S4, the construction of the charge-discharge curve prediction model includes: Using the low voltage interval data and the battery capacity estimation data as independent variables and the high voltage interval data as dependent variables, construct a second data set, and use the second data set to train a deep learning model to obtain a charging curve prediction model; A third data set is constructed with the low voltage interval data and the battery capacity estimation data as independent variables and the discharge curve data as dependent variables, and a deep learning model is trained using the third data set to obtain a discharge curve prediction model.

3. The method according to claim 2, characterized in that In step S6, the overall model is used to predict the complete charge and discharge curve of the battery to be tested, including predicting the complete charge curve, which includes: Obtaining low voltage interval data collected during the charging process of the battery to be tested; Inputting the collected low voltage interval data into the capacity prediction model to obtain a capacity estimation value; Inputting the collected low voltage interval data and the estimated capacity value into the charging curve prediction model to obtain predicted high voltage interval data; The collected low voltage interval data and the predicted high voltage interval data are spliced ​​to obtain complete charging curve data.

4. The method according to claim 3, characterized in that In step S6, the overall model is used to predict the complete charge and discharge curve of the battery to be tested, including predicting the complete discharge curve, which includes: Inputting the collected low voltage interval data into the capacity prediction model to obtain a capacity estimation value; The collected low voltage interval data and capacity estimation value are input into the discharge curve prediction model to obtain complete discharge curve data.

5. The method according to claim 2, characterized in that: The method further includes updating the model, wherein the updating the model includes: The battery management system is used to collect the complete charge and discharge curve data and the discharge capacity of the battery during operation. Merging the collected data with the first data set to form new training data; The new training data is used to update the capacity prediction model, the charging curve prediction model and the discharging curve prediction model, thereby updating the overall model.

6. The method according to claim 1, characterized in that In step S1, the discharge capacity, complete charging curve and complete discharging curve of several batteries of the same type to be tested are obtained in the following manner: Discharge all batteries at a constant current to the lower cut-off voltage to ensure that the batteries are completely discharged; All batteries are fully charged and discharged. The charging process is constant current charging first, then constant voltage charging, and the discharging process is constant current discharging. Record the current, voltage and time data of each battery charging and discharging process; Extract the discharge curve of the complete discharge process of the battery; Based on the current and time data of the complete discharge process of the battery, the discharge capacity of each battery is calculated using the ampere-hour integration method; Extract the charging curve of the battery's complete charging process.

7. The method according to claim 1, characterized in that In step S2, the charging curve data is divided into low voltage interval data and high voltage interval data, including: Set a voltage value V0; Classify the interval less than or equal to V0 in the charging curve data as low voltage interval data; The interval greater than V0 in the charging curve data is classified as high voltage interval data.

8. The method according to claim 1, characterized in that In step S3, random forest regression, support vector machine or neural network is used to perform the machine learning.

9. The method according to claim 1, characterized in that: In step S4, a convolutional neural network or a long short-term memory network is used to perform the deep learning.

10. A battery charge and discharge curve prediction system based on capacity estimation, characterized in that: The system is used to execute the method according to any one of claims 1 to 9.

Images