Method and device for predicting maximum available capacity of battery cell and energy storage system

Abstract

The invention provides a method and device for predicting the maximum available capacity of a battery cell and an energy storage system, and relates to the technical field of batteries, and the method comprises the steps: obtaining a data set which comprises a plurality of pieces of first fragment data; inputting the first preset fragment data and the plurality of pieces of first fragment data into a preset model for first voltage prediction to obtain a plurality of pieces of first predicted voltage sequence data; when a predicted voltage corresponding to the first predicted voltage sequence data reaches a first cut-off voltage, stopping prediction, and obtaining a first moment and target fragment data; inputting the second preset fragment data and a plurality of pieces of first fragment data starting from the target fragment data into a preset model for second voltage prediction to obtain a plurality of pieces of second predicted voltage sequence data; when the voltage corresponding to the second predicted voltage sequence data reaches a second cut-off voltage, stopping prediction, and obtaining a second moment; and generating the predicted maximum available capacity of the battery cell according to the standard working condition information of discharging of the battery cell, the first moment and the second moment.

Description

Technical Field

[0001] The present application relates to the technical field of batteries, and in particular, to a method, device, and energy storage system for predicting the maximum available capacity of a battery cell. Background Art

[0002] The maximum available capacity of a battery cell refers to the maximum amount of electricity that the battery cell can provide under specific charge and discharge conditions. By predicting the maximum available capacity of the battery cell, the state of health (SOH) of the battery cell can be estimated, thereby predicting the performance, service life, and degradation state of the battery cell.

[0003] In related technologies, methods based on testing, empirical formulas, models, and machine learning are usually used to predict the maximum available capacity of a battery cell. However, there are problems such as low prediction accuracy and inability to apply to different working conditions of the battery cell. Summary of the Invention

[0004] The present application provides a method, device, and energy storage system for predicting the maximum available capacity of a battery cell, which is applicable to predicting the maximum available capacity of a battery cell under different working conditions and improves the prediction accuracy.

[0005] In a first aspect, the present application provides a method for predicting the maximum available capacity of a battery cell, including: Obtaining a data set, the data set includes a plurality of first segment data, the plurality of first segment data are arranged in a time series, and each first segment data includes current information, voltage information, and temperature information of the battery cell; Inputting the first preset segment data and the plurality of first segment data into a trained preset model for the first voltage prediction respectively to obtain a plurality of first predicted voltage sequence data, the first preset segment data includes standard working condition information of the battery cell charging; When the predicted voltage corresponding to the first predicted voltage sequence data reaches the first cut-off voltage, stop the first voltage prediction, obtain the first moment and the target segment data, the first moment is the moment corresponding to the first cut-off voltage, and the target segment data is the last first segment data input when stopping the first voltage prediction; Inputting the second preset segment data and the plurality of first segment data starting from the target segment data into a trained preset model for the second voltage prediction respectively to obtain a plurality of second predicted voltage sequence data, the second preset segment data includes standard working condition information of the battery cell discharging; When the predicted voltage corresponding to the second predicted voltage sequence data reaches the second cut-off voltage, stop the second voltage prediction, and obtain the second moment, the second moment is the moment corresponding to the second cut-off voltage; Generating the predicted maximum available capacity of the battery cell according to the standard working condition information of the battery cell discharging, the first moment, and the second moment; Estimate the state of health of the battery cell based on the predicted maximum available capacity of the battery cell.

[0006] In a second aspect, the present application provides a device for predicting the maximum available capacity of a battery cell, including: An acquisition module, configured to acquire a data set, the data set includes a plurality of first segment data, the plurality of first segment data are arranged in a time series, and each first segment data includes current information, voltage information, and temperature information of the battery cell; A first voltage prediction module, configured to input the first preset segment data and the plurality of first segment data into a trained preset model for the first voltage prediction respectively, to obtain a plurality of first predicted voltage sequence data, the first preset segment data includes standard operating condition information of the battery cell during charging; When the predicted voltage corresponding to the first predicted voltage sequence data reaches a first cut-off voltage, stop the first voltage prediction, and obtain a first moment and target segment data, the first moment is the moment corresponding to the first cut-off voltage, and the target segment data is the last first segment data input when the first voltage prediction is stopped; A second voltage prediction module, configured to input the second preset segment data and the plurality of first segment data starting from the target segment data into a trained preset model for the second voltage prediction respectively, to obtain a plurality of second predicted voltage sequence data, the second preset segment data includes standard operating condition information of the battery cell during discharging; When the predicted voltage corresponding to the second predicted voltage sequence data reaches a second cut-off voltage, stop the second voltage prediction, and obtain a second moment, the second moment is the moment corresponding to the second cut-off voltage; A generation module, configured to generate the predicted maximum available capacity of the battery cell according to the standard operating condition information of the battery cell during discharging, the first moment, and the second moment; An estimation module, configured to estimate the state of health of the battery cell based on the predicted maximum available capacity of the battery cell.

[0007] In a third aspect, the present application provides an energy storage system, the energy storage system includes a battery management system and a battery pack, the battery pack includes a plurality of battery cells, and the battery management system is used to implement the method for predicting the maximum available capacity of the battery cell as in the first aspect.

[0008] The beneficial effects of the present application are: The present application provides a method for predicting the maximum available capacity of a battery cell. By inputting the voltage information, current information, temperature information of the battery cell, and the standard operating condition information of the battery cell charging into a preset model for the first voltage prediction, simulating the charging process of the battery cell under the standard operating condition, and iteratively predicting until the predicted voltage of the battery cell reaches the full charge voltage, recording the moment when the full charge voltage is reached; then inputting the voltage information, current information, temperature information of the fully charged battery cell, and the standard operating condition information of the battery cell discharging into the preset model for the second voltage prediction, simulating the process of the battery cell discharging from the fully charged state under the standard operating condition, and iteratively predicting until the battery cell voltage reaches the full discharge voltage, recording the moment when the full discharge voltage is reached. Then, by integrating the current during the process of the battery cell from the full charge moment to the full discharge moment, the predicted maximum available capacity of the battery cell is obtained, and further the health state of the battery cell is evaluated.

[0009] The present application uses a preset model to predict the voltage of the battery cell under the standard operating condition, simulates the full charge and full discharge processes of the battery cell, omits the work of laboratory testing and capturing special points; does not require prior calibration and parameter identification; and the present application uses a self-supervised deep learning model for training and prediction, omits the work of feature extraction and labeling, and uses the full amount of cloud data (i.e., the data set) for training and prediction. The full amount of cloud data includes the parameters of the battery cell under various operating conditions such as charging, discharging, and standing, enabling the model to adapt to different operating conditions of the battery cell, with strong adaptability, improving the generalization ability and prediction accuracy of the model. Description of the Drawings

[0010] Figure 1 It is a schematic flow chart of the method for predicting the maximum available capacity of the battery cell provided by the embodiment of the present application; Figure 2 It is a schematic flow chart of forming the data set provided by the embodiment of the present application; Figure 3 It is a schematic flow chart of the training method of the preset model provided by the embodiment of the present application; Figure 4 It is a schematic structural diagram of the device for predicting the maximum available capacity of the battery cell provided by the embodiment of the present application; Figure 5 It is a schematic structural diagram of the energy storage system provided by the embodiment of the present application. Detailed Embodiments

[0011] In the embodiments of the present application, unless otherwise specified, the character " / " indicates that the related objects before and after are in an "or" relationship. For example, A / B may represent A or B. "And / or" describes the association relationship of the related objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone these three situations.

[0012] It should be noted that the terms "first", "second", etc. involved in the embodiments of the present application are only for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features, nor can they be understood as indicating or implying an order.

[0013] In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. In addition, "at least one (item)" or its similar expression means any combination of these items, which can include any combination of a single item or plural items. For example, at least one (item) of A, B, or C can mean: A, B, C, A and B, A and C, B and C, or A, B, and C. Each of A, B, and C itself can be an element or a set containing one or more elements.

[0014] In the embodiments of the present application, terms such as "exemplary", "in some embodiments", "in another embodiment" are used to give examples, illustrations, or explanations. Any embodiment or design described as "exemplary" in the present application should not be construed as being more preferred or having more advantages than other embodiments or designs. Rather, the use of the term "exemplary" is intended to present concepts in a specific manner.

[0015] In the embodiments of the present application, "of", "corresponding", and "corresponding to" can sometimes be used interchangeably. It should be noted that when not emphasizing their differences, the intended meanings are the same. In the embodiments of the present application, communication and transmission can sometimes be used interchangeably. It should be noted that when not emphasizing their differences, the meanings they express are the same. For example, transmission can include sending and / or receiving, and can be a noun or a verb.

[0016] In the embodiments of the present application, the equality involved can be used in combination with greater than, applicable to the technical solutions adopted when it is greater than, or can also be used in combination with less than, applicable to the technical solutions adopted when it is less than. It should be noted that when equality is used in combination with greater than, it cannot be used in combination with less than; when equality is used in combination with less than, it is not used in combination with greater than.

[0017] In the related art, methods based on testing, empirical formula, model, and machine learning are usually used to predict the maximum available capacity of the battery cell. Among them, the method based on testing can only perform charge and discharge tests under standard working conditions in the experiment to detect the capacity of the battery cell, or rely on special points such as full charge and discharge and static state in actual use of the battery cell to correct the capacity. The available points are few and the coverage rate is low; the method based on empirical formula is often difficult to apply to various working conditions of the battery cell in actual situations due to a single working condition; for the method based on model, whether it is an equivalent circuit or an electrochemical model, it depends on a large amount of prior work of parameter calibration and parameter identification, and the accuracy is greatly affected by it, and repeated calibration and identification are required as the battery ages; the method based on machine learning requires a large number of labeled samples, otherwise the generalization ability of the model is very weak.

[0018] Based on the above problems, an embodiment of the present application proposes a method for predicting the maximum available capacity of a battery cell, which is applicable to predicting the maximum available capacity of a battery cell under different working conditions and improving the prediction accuracy.

[0019] Now in combination with Figures 1 - 3 the method for predicting the maximum available capacity of a battery cell provided by an embodiment of the present application will be described.

[0020] Figure 1 FIG. is a schematic flowchart of the method for predicting the maximum available capacity of a battery cell provided by an embodiment of the present application, which specifically includes the following steps: Step S11, obtain a data set, where the data set includes a plurality of first segment data.

[0021] Among them, the plurality of first segment data are arranged in a time series, and each first segment data includes the current information, voltage information, and temperature information of the battery cell.

[0022] Specifically, the data set can be obtained from a local database or a cloud database. Preferably, the edge terminal uploads signals such as the voltage, current, and temperature of the battery cell to the cloud and stores them in the cloud database. When predicting the maximum available capacity of the battery cell, the data set is obtained from the cloud database, and the full amount of cloud data of the battery cell is used to participate in model training to improve the training effect and generalization ability of the model.

[0023] Among them, as Figure 2 shown, Figure 2 FIG. is a schematic flowchart of forming a data set provided by an embodiment of the present application. The formation process of this data set includes: Step S21, sample the voltage, current, and temperature of the battery cell at a fixed sampling frequency to obtain the historical time series data of the battery cell.

[0024] Specifically, the voltage, current, and temperature of the battery cell are sampled at a fixed sampling frequency (e.g., once every 20 s). The voltage, current, and temperature of the battery cell are taken to form each frame of data of the battery cell. Each frame of data refers to a set of battery cell parameters (including voltage, current, and temperature) collected at fixed time intervals (e.g., 20 s) according to the fixed sampling frequency. Each frame of data is arranged in chronological order to form the historical time series data of the battery cell.

[0025] Optionally, the temperature information of the battery cell may include the surface temperature of the target battery cell (the battery cell for which the maximum available capacity is predicted), or may include the average temperature of multiple battery cells including the target battery cell.

[0026] In an actual application scenario, in a battery pack composed of multiple battery cells, there are many battery cells. Considering the installation cost and the space utilization rate of the battery pack, a separate temperature sensor will not be used to detect the surface temperature of the target battery cell. Therefore, in a battery pack composed of multiple battery cells, the temperature information of the battery cell includes the average temperature, and the average temperature is the average temperature of all battery cells in the battery pack.

[0027] Specifically, the method provided in the embodiment of the present application further includes: obtaining the highest battery cell temperature and the lowest battery cell temperature in the battery pack; calculating the average temperature according to the highest battery cell temperature and the lowest battery cell temperature, that is, average temperature = (highest battery cell temperature + lowest battery cell temperature) / 2.

[0028] Step S22: Use a frame-by-frame moving method to slice and split the historical time series data to obtain multiple first segment data.

[0029] Frame-by-frame moving (also known as a sliding window) is a commonly used method for processing time series data, which is used to split a long sequence of data (i.e., historical time series data) into multiple subsequences of fixed length (i.e., first segment data). The specific steps of frame-by-frame moving include: (1) Define the window size: The window size (which can also be called the slice length) refers to the length of each subsequence. For example, if the window size is 5, then each subsequence will contain 5 data points.

[0030] (2) Define the sliding step: The sliding step refers to the number of steps moved each time. For example, if the sliding step is 1, then one data point is moved each time.

[0031] (3) Initialize the data: Prepare the original time series data (i.e., historical time series data).

[0032] (4) Slide the window: Starting from the starting position of the historical time series data, move a distance of the sliding step each time, and extract data of the window size until the end of the data.

[0033] (5) Generate the first segment data: After each slide, save the extracted data as a first segment data.

[0034] For example, the historical time series data includes 9 frames of data, denoted as [x1, x2, x3, x4, x5, x6, x7, x8, x9], and each frame of data includes the voltage information, current information, and temperature information of the battery cell. Using the frame-by-frame moving method, define the window size as 3 and the sliding step as 1. The first segment data obtained by slicing and splitting is denoted as: [x1, x2, x3], [x2, x3, x4], [x3, x4, x5], [x4, x5, x6], [x5, x6, x7], [x6, x7, x8], [x7, x8, x9].

[0035] Using the frame-by-frame moving method to slice and split the historical time series data to generate multiple first segment data can increase the diversity of the data, thus providing more training samples for the model and helping to improve the generalization ability of the model. Moreover, the segment data generated by the frame-by-frame moving method can be used to train more complex models, such as models based on Transformer, which can capture the long-term dependencies in the long sequence data, thereby improving the accuracy and efficiency of prediction.

[0036] Step S23, perform data preprocessing on multiple first segment data, and the data preprocessing includes abnormal data clearing.

[0037] To ensure the accuracy of the prediction results and improve the stability and prediction accuracy of the model, the voltage, current, and temperature in the first segment data must all be valid values. Therefore, it is necessary to perform data preprocessing on multiple first segment data, clear the abnormal data in multiple first segment data, and retain the first segment data in which the voltage, current, and temperature are all valid values.

[0038] Step S24, form a data set from multiple first segment data after data preprocessing.

[0039] Specifically, form a data set from multiple first segment data obtained after processing in step S23. For example, as shown in the example of step S22, if multiple first segment data are all valid values, the data set D formed is denoted as: {[x1, x2, x3], [x2, x3, x4], [x3, x4, x5], [x4, x5, x6], [x5, x6, x7], [x6, x7, x8], [x7, x8, x9]}.

[0040] In some alternative embodiments, the data set D includes 98% training set, 1% validation set, and 1% test set. Training the model using a large amount of data in the data set can improve the performance and robustness of the model. Using a relatively small proportion of data to validate and test the model can quickly verify the model's performance and independently evaluate the model's generalization ability.

[0041] Step S12: Input the first preset segment data and multiple first segment data into the trained preset model for the first voltage prediction to obtain multiple first predicted voltage sequence data.

[0042] Among them, the first preset segment data includes the standard operating condition information of the battery cell charging.

[0043] In this step, the standard operating condition information of the battery cell charging includes the first standard temperature and a constant charging power, or the standard operating condition information of the battery cell charging includes the first standard temperature and a constant charging current. For example, under standard operating conditions, the battery cell is charged at a constant temperature and constant power at 25°C and -0.5P. To meet the data input length requirement of the preset model, the length of the first preset segment data is the same as that of the first segment data. Assuming the dimension of the first preset segment data is 1 3, expressed as [y1, y2, y3], that is, the first preset segment data includes 3 frames of data, each frame of data includes information of a temperature of 25°C and a power of -0.5P, and the voltage information in the first preset segment data is all 0.

[0044] Input the first preset segment data and multiple first segment data into the trained preset model for the first voltage prediction. Specifically, taking the data set D: {[x1, x2, x3], [x2, x3, x4], [x3, x4, x5], [x4, x5, x6], [x5, x6, x7], [x6, x7, x8], [x7, x8, x9]} as an example, the input of the preset model includes multiple groups of input data composed of the first preset segment data and multiple first segment data. The multiple groups of input data include {[y1, y2, y3], [x1, x2, x3]}, {[y1, y2, y3], [x2, x3, x4]}, {[y1, y2, y3], [x3, x4, x5]}, {[y1, y2, y3], [x4, x5, x6]}, {[y1, y2, y3], [x5, x6, x7]}, {[y1, y2, y3], [x6, x7, x8]}, {[y1, y2, y3], [x7, x8, x9]}. The output of the model is multiple first predicted voltage sequence data corresponding to the multiple groups of input data. The dimension of the first predicted voltage sequence data is 1 3, used to represent the predicted voltage of the battery cell charging under standard operating conditions.

[0045] Step S13: When the predicted voltage corresponding to the first predicted voltage sequence data reaches the first cut-off voltage, stop the first voltage prediction and obtain the first moment and the target segment data.

[0046] Among them, the first moment is the moment corresponding to the first cut-off voltage, and the target segment data is the last first segment data input when the first voltage prediction is stopped.

[0047] Specifically, the first cut-off voltage is expressed as the full charge voltage of the battery cell. When the predicted voltage corresponding to the first predicted voltage sequence data reaches the first cut-off voltage, it indicates that the battery cell has reached the full charge state. At this time, stop the data input and prediction of the model, record the moment when the predicted voltage reaches the first cut-off voltage (i.e., the first moment), and the last first segment data input by the model (i.e., the target segment data), indicating that when the preset model inputs the target segment data for prediction, the predicted voltage of the corresponding battery cell reaches the first cut-off voltage.

[0048] It should be noted that the predicted voltage corresponding to the first predicted voltage sequence data reaching the first cut-off voltage means that the predicted voltage is greater than or equal to the first cut-off voltage. At this time, it is considered that the battery cell has reached the full charge state.

[0049] For example, in the trained preset model, input the data composed of the first preset segment data and the first segment data one by one. When the input is {[y1,y2,y3],[x4,x5,x6]}, the output first predicted voltage sequence data is [u1,u2,u3], where u2 is equal to the first cut-off voltage (such as 3.6V). Then stop the data input and prediction of the model, record the moment T1 corresponding to the voltage u2 (i.e., the first moment), and the target segment data [x4,x5,x6]. It can be understood that the first predicted voltage sequence data is used to characterize the voltage predicted according to the first segment data under the standard working conditions of the battery cell charging. The first predicted voltage sequence data corresponds to the first segment data one by one. The moment corresponding to the voltage u2 is also the moment corresponding to the data of this frame x5. Specifically, the first moment can be calculated by multiplying the length from x1 to x5 by a fixed time interval. For example, let x1 be the starting moment, the fixed time interval be 20s, and the length from x1 to x5 be 4. The first moment is 4 × 20s = 80s.

[0050] In this application, by inputting the voltage information, current information, and temperature information of the battery cell, as well as the standard working condition information of the battery cell charging into the preset model, simulating the charging process of the battery cell under the standard working conditions, and iteratively predicting repeatedly until the predicted voltage of the battery cell reaches the full charge voltage, and recording the moment when the full charge voltage is reached, the work of laboratory testing and capturing special points is omitted.

[0051] Step S14: Input the second preset segment data and multiple first segment data starting from the target segment data into the trained preset model for the second voltage prediction to obtain multiple second predicted voltage sequence data.

[0052] Among them, the second preset segment data includes the standard working condition information of the battery cell discharging.

[0053] In this step, the standard working condition information of the battery cell discharging includes the second standard temperature and the constant discharging power, or the standard working condition information of the battery cell discharging includes the second standard temperature and the constant discharging current. For example, under the standard working condition, the battery cell discharges at a constant temperature and constant power of 25°C and 0.5P. To meet the data input length requirement of the preset model, the length of the second preset segment data is the same as that of the first segment data. Assuming the dimension of the second preset segment data is 1 3, expressed as [y4, y5, y6], that is, the second preset segment data includes 3 frames of data. Each frame of data includes the information of a temperature of 25°C and a power of 0.5P, and the voltage information in the second preset segment data is all 0.

[0054] Input the second preset segment data and multiple first segment data starting from the target segment data into the trained preset model for the second voltage prediction. Specifically, taking the data set D: {[x1, x2, x3], [x2, x3, x4], [x3, x4, x5], [x4, x5, x6], [x5, x6, x7], [x6, x7, x8], [x7, x8, x9]} as an example, the multiple first segment data starting from the target segment data [x4, x5, x6] includes {[x4, x5, x6], [x5, x6, x7], [x6, x7, x8], [x7, x8, x9]}. Then the multiple groups of input data of the preset model include {[y4, y5, y6], [x4, x5, x6]}, {[y4, y5, y6], [x5, x6, x7]}, {[y4, y5, y6], [x6, x7, x8]}, {[y4, y5, y6], [x7, x8, x9]}. The output of the model is multiple second predicted voltage sequence data corresponding to the multiple groups of input data. The dimension of the second predicted voltage sequence data is 1 3, used to represent the predicted voltage of the battery cell discharging under the standard working condition.

[0055] Step S15: When the predicted voltage corresponding to the second predicted voltage sequence data reaches the second cut-off voltage, stop the second voltage prediction and obtain the second moment.

[0056] Among them, the second moment is the moment corresponding to the second cut-off voltage.

[0057] Specifically, the second cut-off voltage is expressed as the full discharge voltage of the battery cell. When the predicted voltage corresponding to the second predicted voltage sequence data reaches the second cut-off voltage, it indicates that the battery cell has reached the full discharge state. At this time, the data input and prediction of the model are stopped, and the moment when the predicted voltage reaches the second cut-off voltage (i.e., the second moment) is recorded.

[0058] It should be noted that the predicted voltage corresponding to the second predicted voltage sequence data reaching the second cut-off voltage means that the predicted voltage is less than or equal to the second cut-off voltage. At this time, it is considered that the battery cell has reached the full discharge state.

[0059] For example, in the trained preset model, data composed of the second preset segment data and multiple first segment data starting from the target segment data are input one by one. When the input is {[y4,y5,y6],[x7,x8,x9]}, the output second predicted voltage sequence data is [u4,u5,u6], where u5 is equal to the second cut-off voltage (such as 2.8V). Then, the data input and prediction of the model are stopped, and the moment T2 corresponding to the voltage u5 (i.e., the second moment) is recorded. It can be understood that the second predicted voltage sequence data is used to characterize the voltage predicted according to the first segment data under the standard working condition of the battery cell discharging. The second predicted voltage sequence data corresponds one by one to the first segment data. The moment corresponding to the voltage u5 is also the moment corresponding to the x8th frame of data. Specifically, the second moment can be calculated by multiplying the length from x1 to x8 by a fixed time interval. For example, let x1 be the starting moment, the fixed time interval be 20s, and the length from x1 to x8 be 7, then the second moment is 7 20s = 140s.

[0060] In this application, the voltage information, current information, and temperature information of the battery cell after full charge, as well as the standard working condition information of the battery cell discharging, are input into the preset model to simulate the discharging process of the battery cell starting from the full charge state under the standard working condition, and iteratively predict repeatedly until the predicted voltage of the battery cell reaches the full discharge voltage, and record the moment when the full discharge voltage is reached, saving the work of laboratory testing and capturing special points.

[0061] In one possible embodiment, the preset model includes a self-supervised deep learning model. The self-supervised deep learning model is a Transformer model. The Transformer model includes an encoder (Encoder) and a decoder (Decoder). The multiple first segment data include continuous second segment data and third segment data. For example, the multiple first segment data include A0,A1,A2,A3,A4,A5. Each first segment data includes m frames of data, and each frame of data includes the voltage, current, and temperature information of the battery cell. Then the dimension of the first segment data is expressed as 3 m. The continuous second segment data and third segment data can be [A0, A1], [A1, A2], [A2, A3], [A3, A4], [A4, A5]. Taking [A0, A1] as an example (i.e., the second segment data is A0 and the third segment data is A1), the training method of the preset model will be described below.

[0062] Figure 3 It is a schematic flowchart of the training method of the preset model provided by the embodiment of the present application. As Figure 3 shown, the training method of the preset model includes: Step S31, set the voltage information of the battery cells in the third segment data to zero to obtain the fourth segment data.

[0063] Specifically, the third segment data A1 includes m frames. Set the voltage information of the battery cells in each frame to 0 to obtain the fourth segment data A1'.

[0064] Step S32, splice the second segment data and the fourth segment data to obtain the fifth segment data.

[0065] Specifically, splice the second segment data A0 and the fourth segment data A1' to obtain the fifth segment data [A0, A1'].

[0066] Step S33, input the second segment data into the encoder, input the fifth segment data into the decoder, and predict the voltage information of the battery cells in the third segment data to obtain the predicted voltage sequence data.

[0067] Specifically, input the second segment data A0 into the encoder Encoder, input the fifth segment data [A0, A1'] into the decoder Decoder, predict the voltage information of the battery cells in the third segment data A1 to obtain the predicted voltage sequence data, that is, obtain the predicted voltage of the battery cells in each frame of the third segment data A1. The dimension of the predicted voltage sequence data is 1 m.

[0068] Furthermore, the predicted voltage sequence data can be compared with the original voltage information in the third segment data A1, and the parameters of the model can be optimized by using an optimization algorithm to improve the prediction accuracy of the model.

[0069] In the present application, a deep learning model using self-supervised learning is adopted, which eliminates the work of labeling and does not require prior calibration and parameter identification. The prediction accuracy of the model is high, and the full-scale cloud data can be used for training. The full-scale cloud data includes data of the battery cells under different working conditions (such as charging, discharging, or standing). The voltage of the battery cells can be predicted under different working conditions, and the training model can adapt to different working conditions of the battery cells to improve the generalization ability of the model.

[0070] Step S16: Generate the predicted maximum available capacity of the battery cell based on the standard operating condition information of the battery cell during discharge, the first moment, and the second moment.

[0071] In one embodiment, the standard operating condition information of the battery cell during charging includes the first standard temperature and the constant charging power, and the standard operating condition information of the battery cell during discharge includes the second standard temperature and the constant discharge power. Then, generating the predicted maximum available capacity of the battery cell based on the standard operating condition information of the battery cell during discharge, the first moment, and the second moment includes: obtaining the discharge current information of the battery cell between the first moment and the second moment according to the constant discharge power and the second predicted voltage sequence data; integrating the discharge current information of the battery cell between the first moment and the second moment to obtain the predicted maximum available capacity of the battery cell.

[0072] For example, the standard operating condition information of the battery cell during charging is 25°C and -0.5P, and the standard operating condition information of the battery cell during discharge is 25°C and 0.5P. According to the above steps S11 - S15, the first moment T1 and the second moment T2, as well as the second predicted voltage sequence data, are obtained. Since the battery cell discharges at a constant power, the discharge current information of the battery cell during the discharge process can be obtained according to the second predicted voltage sequence data, that is, the discharge current information of the battery cell between the first moment T1 and the second moment T2 is obtained. By integrating the discharge current information using the current integration method (also known as the Coulomb counting method), the predicted maximum available capacity of the battery cell can be obtained.

[0073] It should be noted that 0.5P represents that the charging power is 0.5 times the rated power of the battery. For example, if the rated power of the battery is 100W, the charging power of 0.5P is expressed as 50W (0.5 × 100 = 50). -0.5P represents that the discharge power is 0.5 times the rated power of the battery. For example, if the rated power of the battery is 100W, the discharge power of -0.5P is expressed as -50W (-0.5 × 100 = -50). The positive and negative signs of the power value only serve to distinguish the charging and discharging processes, and usually, charging is positive and discharging is negative.

[0074] In another embodiment, the standard operating condition information of the battery cell during charging includes the first standard temperature and the constant charging current, and the standard operating condition information of the battery cell during discharge includes the second standard temperature and the constant discharge current. Then, generating the predicted maximum available capacity of the battery cell based on the standard operating condition information of the battery cell during discharge, the first moment, and the second moment includes: integrating the constant discharge current between the first moment and the second moment to obtain the predicted maximum available capacity of the battery cell.

[0075] For example, the standard operating condition information for charging the battery cell is 25°C and 0.3C, and the standard operating condition information for discharging the battery cell is 25°C and -0.3C. According to the above steps S11 - S15, the first moment T1 and the second moment T2 are obtained. By integrating the constant discharge current (-0.3C) between the first moment T1 and the second moment T2 using the current integration method (also known as Coulomb counting method), the predicted maximum available capacity of the battery cell can be obtained.

[0076] It should be noted that 0.3C means the charging current is 0.3 times the rated capacity of the battery. For example, if the rated capacity of the battery is 100Ah, the charging current of 0.3C is expressed as 30A (0.3 × 100 = 30). -0.3C means the discharge current is 0.3 times the rated capacity of the battery. For example, if the rated capacity of the battery is 100Ah, the discharge current of -0.3C is expressed as -30A (-0.3 × 100 = -30). The plus and minus signs of the current value only serve to distinguish the charging and discharging processes, usually with charging being positive and discharging being negative.

[0077] Step S17, estimate the state of health of the battery cell based on the predicted maximum available capacity of the battery cell.

[0078] Specifically, based on the predicted maximum available capacity of the battery cell obtained in the above step S16, divide the predicted maximum available capacity of the battery cell by the initial maximum available capacity of the battery cell to obtain the state of health of the battery cell.

[0079] To improve the accuracy of the predicted maximum available capacity of the battery cell (or the estimated state of health of the battery cell), multiple full charge and discharge cycle tests can be performed using a preset model, that is, repeat steps S11 - S16. For example, repeat three times, and take the average value of the predicted maximum available capacities of the battery cell obtained three times as the final prediction result. By taking the average value through multiple predictions in this application, the stability and accuracy of the prediction can be improved.

[0080] The method proposed in the embodiment of this application inputs the voltage information, current information, and temperature information of the battery cell, as well as the standard operating condition information for charging the battery cell, into a preset model for the first voltage prediction, simulating the charging process of the battery cell under standard operating conditions, and iteratively predicting repeatedly until the predicted voltage of the battery cell reaches the full charge voltage, and record the moment when the full charge voltage is reached; then input the voltage information, current information, and temperature information of the fully charged battery cell, as well as the standard operating condition information for discharging the battery cell, into the preset model for the second voltage prediction, simulating the discharging process of the battery cell starting from the fully charged state under standard operating conditions, and iteratively predicting repeatedly until the battery cell voltage reaches the full discharge voltage, and record the moment when the full discharge voltage is reached. Then, by integrating the current during the process of the battery cell from the full charge moment to the full discharge moment, the predicted maximum available capacity of the battery cell is obtained, and further the state of health of the battery cell is evaluated.

[0081] This application uses a preset model to predict the voltage of the battery cell under standard operating conditions, simulates the full charge and full discharge processes of the battery cell, eliminating the need for laboratory testing and the work of capturing special points; prior calibration and parameter identification are not required, and the battery cell parameters can be automatically updated as the battery ages; moreover, this application uses a self-supervised deep learning model for training and prediction, eliminating the work of feature extraction and labeling, and uses all cloud data (i.e., the dataset) for training and prediction. The all cloud data includes the parameters of the battery cell under various operating conditions such as charging, discharging, and standing, enabling the model to adapt to different operating conditions of the battery cell, with strong adaptability, improving the generalization ability and prediction accuracy of the model.

[0082] Based on the same idea, an embodiment of this application also provides a device for predicting the maximum available capacity of a battery cell. Figure 4 As shown in Figure 4 the structure diagram of the device for predicting the maximum available capacity of a battery cell provided by an embodiment of this application, the device 40 for predicting the maximum available capacity of a battery cell may include: An acquisition module 41, configured to acquire a dataset. The dataset includes a plurality of first segment data, and the plurality of first segment data are arranged in a time series. Each first segment data includes the current information, voltage information, and temperature information of the battery cell; A first voltage prediction module 42, configured to input the first preset segment data and the plurality of first segment data into the trained preset model for the first voltage prediction respectively, to obtain a plurality of first predicted voltage sequence data. The first preset segment data includes the standard operating condition information of the battery cell during charging; When the predicted voltage corresponding to the first predicted voltage sequence data reaches the first cut-off voltage, stop the first voltage prediction, and obtain the first moment and the target segment data. The first moment is the moment corresponding to the first cut-off voltage, and the target segment data is the last first segment data input when the first voltage prediction is stopped; A second voltage prediction module 43, configured to input the second preset segment data and the plurality of first segment data starting from the target segment data into the trained preset model for the second voltage prediction respectively, to obtain a plurality of second predicted voltage sequence data. The second preset segment data includes the standard operating condition information of the battery cell during discharging; When the predicted voltage corresponding to the second predicted voltage sequence data reaches the second cut-off voltage, stop the second voltage prediction, and obtain the second moment. The second moment is the moment corresponding to the second cut-off voltage; A generation module 44, configured to generate the predicted maximum available capacity of the battery cell according to the standard operating condition information of the battery cell during discharging, the first moment, and the second moment; An estimation module 45, configured to estimate the health state of the battery cell according to the predicted maximum available capacity of the battery cell.

[0083] In one possible implementation, the preset model includes a self-supervised deep learning model, the self-supervised deep learning model is a Transformer model, the Transformer model includes an encoder and a decoder, and the multiple first segment data includes consecutive second segment data and third segment data. The training method of the preset model includes: Set the voltage information of the battery cells in the third segment data to zero to obtain fourth segment data; Concatenate the second segment data and the fourth segment data to obtain fifth segment data; Input the second segment data into the encoder, input the fifth segment data into the decoder, and predict the voltage information of the battery cells in the third segment data to obtain predicted voltage sequence data.

[0084] In one possible implementation, the standard operating condition information for battery cell charging includes a first standard temperature and a constant charging power, and the standard operating condition information for battery cell discharging includes a second standard temperature and a constant discharging power.

[0085] In one possible implementation, the generating module 44 can also be used for: Obtain the discharging current information of the battery cells between the first moment and the second moment according to the constant discharging power and the second predicted voltage sequence data; Integrate the discharging current information of the battery cells between the first moment and the second moment to obtain the predicted maximum available capacity of the battery cells.

[0086] In one possible implementation, the standard operating condition information for battery cell charging includes a first standard temperature and a constant charging current, and the standard operating condition information for battery cell discharging includes a second standard temperature and a constant discharging current.

[0087] In one possible implementation, the generating module 44 can also be used for: Integrate the constant discharging current between the first moment and the second moment to obtain the predicted maximum available capacity of the battery cells.

[0088] In one possible implementation, the estimating module 45 can also be used for: Divide the predicted maximum available capacity of the battery cells by the initial maximum available capacity of the battery cells to obtain the health state of the battery cells.

[0089] In one possible implementation, the formation process of the data set includes: Sample the voltage, current, and temperature of the battery cells at a fixed sampling frequency to obtain the historical time series data of the battery cells; Use the frame-by-frame moving method to slice and split the historical time series data to obtain multiple first segment data; Perform data preprocessing on multiple first segment data, where the data preprocessing includes abnormal data clearing; A data set is composed of multiple first segment data after data preprocessing.

[0090] In one possible implementation, the data set includes a 98% training set, a 1% validation set, and a 1% test set.

[0091] In one possible implementation, the temperature information includes the average temperature. A battery pack is composed of multiple battery cells, and the acquisition module 41 is further configured to acquire the highest battery cell temperature and the lowest battery cell temperature in the battery pack; the above-mentioned device 40 for predicting the maximum available capacity of the battery cell further includes: A calculation module, configured to calculate the average temperature according to the highest battery cell temperature and the lowest battery cell temperature.

[0092] Figure 4 The device 40 for predicting the maximum available capacity of the battery cell provided by the illustrated embodiment can be used to execute the technical solution of the method embodiment shown in the present application, and its implementation principle and technical effect can be further referred to the relevant description in the method embodiment.

[0093] It should be understood that the division of each module of the above Figure 4 The device 40 for predicting the maximum available capacity of the battery cell shown is only a logical function division. In actual implementation, it can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also all be implemented in the form of hardware; or some modules can be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the acquisition module can be a separately established processing element, or can be integrated in a certain chip of an electronic device. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together or can be independently implemented. In the implementation process, each step of the above method or each of the above modules can be completed by the integrated logic circuit in the processor element or the instruction in the form of software.

[0094] For example, the above-mentioned modules may be one or more integrated circuits configured to implement the above methods, such as: one or more Application Specific Integrated Circuits (ASICs), or, one or more Digital Signal Processors (DSPs), or, one or more Field Programmable Gate Arrays (FPGAs), etc. Additionally, these modules may be integrated together and implemented in the form of a System-On-a-Chip (SOC).

[0095] In the above embodiments, the involved processors may include, for example, CPUs, DSPs, microcontrollers or digital signal processors, and may also include GPUs, Neural-network Process Units (NPUs), and Image Signal Processors (ISPs). The processor may further include necessary hardware accelerators or logic processing hardware circuits, such as ASICs, or one or more integrated circuits for controlling the execution of the technical solution programs of this application. In addition, the processor may have the function of operating one or more software programs, and the software programs may be stored in a storage medium.

[0096] The embodiments of this application also provide an energy storage system, which includes a Battery Management System (BMS) and a battery pack. The battery pack includes multiple battery cells, and the battery management system is used to implement the methods provided in the embodiments shown in this application.

[0097] Figure 5 It is a schematic structural diagram of the energy storage system provided in the embodiments of this application. As Figure 5 shown, the energy storage system includes a battery pack 51 and a BMS 52. The battery pack 51 includes multiple battery cells, and the BMS 52 is used to implement the method for predicting the maximum available capacity of the battery cells provided in the embodiments shown in this application. Figure 5 The above-mentioned energy storage system is only for illustrative purposes, and the specific structure is not limited. The positions and quantities of the battery pack and BMS included in the energy storage system are also not limited.

[0098] As mentioned above, the above are only the specific implementation manners of this application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. The protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A method for predicting the maximum available capacity of a battery cell, characterized in that: The method comprises: Acquire a data set, the data set comprising a plurality of first segment data, the plurality of first segment data being arranged in time series, each first segment data comprising current information, voltage information and temperature information of a battery cell; Inputting the first preset segment data and the plurality of first segment data into the trained preset model respectively to perform a first voltage prediction, to obtain a plurality of first predicted voltage sequence data, wherein the first preset segment data includes standard operating condition information of the battery cell charging; When the predicted voltage corresponding to the first predicted voltage sequence data reaches the first cut-off voltage, the first voltage prediction is stopped, and a first moment and target segment data are acquired, wherein the first moment is the moment corresponding to the first cut-off voltage, and the target segment data is the last first segment data input when the first voltage prediction is stopped; Inputting the second preset segment data and a plurality of first segment data starting from the target segment data into the trained preset model for a second voltage prediction to obtain a plurality of second predicted voltage sequence data, wherein the second preset segment data includes standard operating condition information of the battery cell discharge; When the predicted voltage corresponding to the second predicted voltage sequence data reaches the second cut-off voltage, stopping the second voltage prediction and acquiring a second moment, where the second moment is the moment corresponding to the second cut-off voltage; Generate a predicted maximum available capacity of the battery cell according to the standard operating condition information of the battery cell discharge, the first moment and the second moment; The health state of the battery cell is estimated based on the predicted maximum available capacity of the battery cell.

2. The method for predicting the maximum available capacity of a battery cell according to claim 1, characterized in that: The preset model includes a self-supervised deep learning model, the self-supervised deep learning model is a Transformer model, the Transformer model includes an encoder and a decoder, the plurality of first segment data include continuous second segment data and third segment data, The training method of the preset model includes: Setting the voltage information of the battery cell in the third segment data to zero to obtain fourth segment data; Concatenate the second fragment data and the fourth fragment data to obtain fifth fragment data; The second segment data is input into the encoder, the fifth segment data is input into the decoder, and the voltage information of the battery cell in the third segment data is predicted to obtain predicted voltage sequence data.

3. The method for predicting the maximum available capacity of a battery cell according to claim 1, characterized in that: The standard operating condition information of the battery cell charging includes a first standard temperature and a constant charging power, and the standard operating condition information of the battery cell discharging includes a second standard temperature and a constant discharging power.

4. The method for predicting the maximum available capacity of a battery cell according to claim 3, characterized in that: Generating a predicted maximum available capacity of the battery cell according to the standard operating condition information of the battery cell discharge, the first moment, and the second moment includes: Obtaining discharge current information of the battery cell between the first moment and the second moment according to the constant discharge power and the second predicted voltage sequence data; The discharge current information of the battery cell between the first moment and the second moment is integrated to obtain the predicted maximum available capacity of the battery cell.

5. The method for predicting the maximum available capacity of a battery cell according to claim 1, characterized in that: The standard operating condition information of the battery cell charging includes a first standard temperature and a constant charging current, and the standard operating condition information of the battery cell discharging includes a second standard temperature and a constant discharging current.

6. The method for predicting the maximum available capacity of a battery cell according to claim 5, characterized in that: Generating a predicted maximum available capacity of the battery cell according to the standard operating condition information of the battery cell discharge, the first moment, and the second moment includes: The constant discharge current between the first moment and the second moment is integrated to obtain a predicted maximum available capacity of the battery cell.

7. The method for predicting the maximum available capacity of a battery cell according to any one of claims 1 to 6, characterized in that: The health status of the battery cell is estimated based on the predicted maximum available capacity of the battery cell, including: The predicted maximum available capacity of the battery cell is divided by the initial maximum available capacity of the battery cell to obtain the health status of the battery cell.

8. The method for predicting the maximum available capacity of a battery cell according to claim 1, characterized in that: The process of forming the data set includes: The voltage, current and temperature of the battery cell are sampled at a fixed sampling frequency to obtain the historical time series data of the battery cell; Slicing and splitting the historical time series data in a frame-by-frame manner to obtain the plurality of first segment data; Performing data preprocessing on the plurality of first fragment data, wherein the data preprocessing includes removing abnormal data; The data set is composed of a plurality of first segment data after the data preprocessing.

9. The method for predicting the maximum available capacity of a battery cell according to claim 8, characterized in that: The dataset consists of 98% training set, 1% validation set and 1% test set.

10. The method for predicting the maximum available capacity of a battery cell according to claim 1, characterized in that: The temperature information includes an average temperature, and a battery pack is composed of a plurality of battery cells. The method further includes: Obtaining the highest cell temperature and the lowest cell temperature in the battery pack; The average temperature is calculated according to the maximum battery cell temperature and the minimum battery cell temperature.

11. A device for predicting the maximum available capacity of a battery cell, characterized in that: include: An acquisition module, used to acquire a data set, wherein the data set includes a plurality of first segment data, the plurality of first segment data are arranged in time series, and each first segment data includes current information, voltage information, and temperature information of a battery cell; a first voltage prediction module, used for inputting the first preset segment data and the plurality of first segment data into the trained preset model respectively to perform a first voltage prediction, and obtain a plurality of first predicted voltage sequence data, wherein the first preset segment data includes standard operating condition information of battery cell charging; When the predicted voltage corresponding to the first predicted voltage sequence data reaches the first cut-off voltage, the first voltage prediction is stopped, and a first moment and target segment data are acquired, wherein the first moment is the moment corresponding to the first cut-off voltage, and the target segment data is the last first segment data input when the first voltage prediction is stopped; a second voltage prediction module, used for inputting the second preset segment data and a plurality of first segment data starting from the target segment data into the trained preset model for second voltage prediction, to obtain a plurality of second predicted voltage sequence data, wherein the second preset segment data includes standard operating condition information of the battery cell discharge; When the predicted voltage corresponding to the second predicted voltage sequence data reaches the second cut-off voltage, stopping the second voltage prediction and acquiring a second moment, where the second moment is the moment corresponding to the second cut-off voltage; A generating module, configured to generate a predicted maximum available capacity of the battery cell according to the standard operating condition information of the battery cell discharge, the first moment and the second moment; An estimation module is used to estimate the health state of the battery cell according to the predicted maximum available capacity of the battery cell.

12. An energy storage system, characterized in that: The energy storage system includes a battery management system and a battery pack, the battery pack includes a plurality of battery cells, and the battery management system is used to implement the method for predicting the maximum available capacity of a battery cell as described in any one of claims 1-10.

Images