Method and device for predicting residual capacity of lithium ion energy storage battery

Abstract

The invention discloses a lithium ion energy storage battery residual capacity prediction method and device, and the method comprises the steps: S1, obtaining the one-dimensional time series data and one-dimensional IC curve data of the voltage, current and temperature of a lithium battery based on a historical charging and discharging data set of an energy storage power station; s2, converting the one-dimensional time sequence data and the one-dimensional IC curve in the charge-discharge cycle process of the lithium battery into a visual two-dimensional feature image, and constructing a lithium ion energy storage battery feature image set; s3, constructing a two-dimensional convolutional neural network model, taking the battery feature image set as input, gradually extracting local to global features, capturing a potential nonlinear relationship between the image features and the capacity, and training to obtain a battery residual capacity prediction model; s4, charging and discharging data are collected in real time and input into the trained battery residual capacity prediction model, and a lithium ion energy storage battery residual capacity prediction result is obtained; according to the invention, accurate prediction of the residual capacity of the lithium ion energy storage battery of the smart energy station is realized.

Description

Technical Field

[0001] This invention relates to the field of energy system coordination and optimization technology, and more specifically to a method and apparatus for predicting the remaining capacity of lithium-ion energy storage batteries. Background Technology

[0002] Online intelligent prediction of the operating status of lithium-ion energy storage batteries is a core part of the research on safety early warning for energy storage power stations. By real-time monitoring and data analysis of lithium-ion energy storage batteries, the operating status of the batteries can be predicted, potential safety hazards can be identified in a timely manner, and strong support can be provided for the safety management of the power station. At the same time, once the energy storage power station is officially put into operation, it will generate a large amount of historical and real-time operating data. By mining and analyzing this data, the operating status of the batteries can be grasped in real time, the battery life can be predicted, and thus a scientific basis can be provided for battery maintenance and management.

[0003] In the current field of battery technology research, one-dimensional voltage, current, and temperature time series or simple statistics are generally used as inputs for prediction models. This makes it difficult to characterize the complex nonlinear coupling dynamics during battery aging and lacks visualization and mining of the key electrochemical feature of incremental capacity (IC) curve. As a result, the error increases sharply at the inflection point of capacity decay, and the generalization and robustness of the model are limited.

[0004] Therefore, how to overcome the limitations of one-dimensional time-series data, fully explore the aging information of incremental capacity curves, and integrate it with multi-physical quantity time-series data into a two-dimensional image with high characterization capabilities, so as to construct a lithium-ion battery remaining capacity prediction model that combines early sensitivity, inflection point accuracy and engineering ease of use, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and apparatus for predicting the remaining capacity of lithium-ion energy storage batteries to solve some of the technical problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for predicting the remaining capacity of a lithium-ion energy storage battery includes the following steps: S1. Based on the historical charge and discharge dataset of energy storage power stations, obtain one-dimensional time series data of lithium battery voltage, current and temperature, as well as one-dimensional IC curve data; S2. Transform the one-dimensional time series data and one-dimensional IC curve during the charge-discharge cycle of lithium batteries into visualized two-dimensional feature images to construct a feature image set for lithium-ion energy storage batteries; S3. Construct a two-dimensional convolutional neural network model, using a set of lithium-ion energy storage battery feature images as input, gradually extracting features from local to global, capturing the potential nonlinear relationship between image features and capacity, and training to obtain a battery remaining capacity prediction model. S4. Real-time charging and discharging data are collected and input into the trained battery remaining capacity prediction model to obtain the remaining capacity prediction result of the lithium-ion energy storage battery.

[0007] Preferably, in step S1, the battery voltage data is as follows: U k ={( t 1, u 1),( t 2, u 2),...,( t n , u n The current data is as follows: I k ={( t 1, i 1),( t 2, i 2),...,( t n , i n Temperature data is T k ={( t 1, T 1),( t 2, T 2),...,( t n , T n IC curve data are Y k ={( u 1,(dq / dv)1),( u 2,(dq / dv)2),...,( u n (dq / dv) n )},in, k The number of cycles. t For time, u For voltage, i For current, T dq is the temperature, dv is the capacitance difference, and dv is the voltage difference. n This represents the data length.

[0008] Preferably, the specific content of step S2, which constructs the feature image set of lithium-ion energy storage batteries, is as follows: S21. Normalize the one-dimensional time series data of voltage, current and temperature and the one-dimensional IC curve data, and use the zero-padding method to pad the data length. S22. Using a color mapping mechanism, the data values ​​of voltage, current and temperature are converted into color maps; S23. For the one-dimensional IC curve, the Gram angle field (GAF) is used to convert the time-amplitude of the time-series signal into radius-angle parameters through polar coordinate mapping, and a two-dimensional feature image is generated by calculating the cosine value between the angles. S24. The images generated in step S22 and step S23 are stitched together to form the final lithium-ion energy storage battery feature image set.

[0009] Preferably, in step S21, the method for normalizing the one-dimensional time series data of voltage, current, and temperature is as follows:

[0010] in, u k for t k The voltage value at that moment; i k for t k The current value at that moment; T k for t k Temperature value at any given time; for t k The voltage value after time normalization. for t k The current value after time normalization. for t k Temperature value normalized to time; The method for padding data length is as follows: select the data length of the first cycle as the baseline, and if the data length of other cycles is insufficient after normalization, padding is performed using zero padding to ensure that all samples have the same length. Step S22: The normalized voltage, current, and temperature time series are used as row vectors of the two-dimensional image. In the MATLAB environment, the imagesc function is used to map the one-dimensional current, voltage, and temperature signal values ​​into colors to generate color images, with each image corresponding to a physical quantity.

[0011] Preferably, in step S23, the calculation method for generating a two-dimensional feature image from a one-dimensional IC curve is as follows: the normalized IC curve is used as a one-dimensional signal, and the inverse cosine value of each data point is calculated to construct the Gram angle field matrix, which is a two-dimensional image. The size is kept the same, and the temporal characteristics and aging information of the IC curve are preserved. Specifically:

[0012]

[0013]

[0014] in, Y For the set of all voltage values, y k For voltage is u k The dq / dv value at that time For normalized y k value, For time, for The polar cosine value of the transformation, G is the Gram angle field (GAF), which is the matrix that transforms a one-dimensional time series into a two-dimensional image.

[0015] Preferably, the specific content of step S3 is as follows: Based on the previous m A set of characteristic images of lithium-ion energy storage batteries with -1 cycle count and corresponding capacity values. C m-1 ={(1, c 1),(2, u 2),...,( m- 1, c m-1 A two-dimensional convolutional neural network model is constructed using a 128×128×3 feature image as input. The model extracts features from local to global levels through convolutional layers, batch normalization layers, ReLU activation layers, and max pooling layers. The number of feature channels increases from 32 to 128. Fully connected layers integrate the features into a 256-dimensional vector, and a Dropout layer prevents overfitting. A loss function is constructed using the actual capacity value and the predicted capacity value, and the Adam algorithm is used to optimize the model parameters. The first m Voltage under cycle U m ={( t 1, u 1),( t 2, u 2),...,( t n ,u n )}, Current I m ={( t 1, i 1),( t 2, i 2),...,( t m , i m )},temperature T m ={( t 1, T 1),( t 2, T 2),...,( t m , T m IC curve Y m ={( u 1,(dq / dv)1),( u 2,(dq / dv)2),...,( u n (dq / dv) n The two-dimensional feature image of the data is used as the input to the test samples of the prediction model. m Number of cycles remaining capacity C m Make predictions.

[0016] A lithium-ion energy storage battery remaining capacity prediction system, based on the aforementioned lithium-ion energy storage battery remaining capacity prediction method, includes a data acquisition module, a feature map conversion module, a model construction and training module, and a prediction module; The data acquisition module is used to acquire one-dimensional time series data of lithium battery voltage, current and temperature, as well as one-dimensional IC curve data, based on historical charge and discharge datasets of energy storage power stations and real-time acquisition. The feature map conversion module is used to convert one-dimensional time series data and one-dimensional IC curves during the charge and discharge cycle of lithium batteries into visualized two-dimensional feature images, and to construct a feature image set for lithium-ion energy storage batteries. The model building and training module is used to build a two-dimensional convolutional neural network model. It takes a set of lithium-ion energy storage battery feature images as input, gradually extracts local to global features, captures the potential nonlinear relationship between image features and capacity, and trains to obtain a battery remaining capacity prediction model. The prediction module is used to input real-time collected charge and discharge data into a trained battery remaining capacity prediction model to obtain the predicted result of the lithium-ion energy storage battery remaining capacity. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for predicting the remaining capacity of a lithium-ion energy storage battery.

[0017] A processing terminal includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the aforementioned method for predicting the remaining capacity of a lithium-ion energy storage battery.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and device for predicting the remaining capacity of lithium-ion energy storage batteries. It transforms data that originally presented and recorded the changes in voltage, current, temperature, and IC curves over time during the charge-discharge cycle of lithium batteries in a one-dimensional form into a feature image set with intuitive visual effects. Simultaneously, it captures temporal and spatial information, effectively characterizing nonlinear degradation patterns while maintaining the basic statistical characteristics of the data. This significantly improves the model's ability to detect and analyze complex signals of battery capacity decay. Furthermore, by combining the external feature image set with a neural network model, it delves into the application of information visualization and deep convolutional networks in predicting the remaining capacity of lithium-ion energy storage batteries. This better adapts to diverse datasets in the field of battery management, achieving faster and more accurate prediction results. It enables accurate prediction of the remaining capacity of lithium-ion energy storage batteries in smart energy stations, providing a reliable basis for battery life prediction and safety management. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a method for predicting the remaining capacity of a lithium-ion energy storage battery provided by the present invention; Figure 2 This is a visual diagram of battery voltage provided by the present invention; Figure 3 This invention provides a visual diagram of battery current. Figure 4 This invention provides a visual illustration of battery temperature. Figure 5 A schematic diagram of IC curve visualization provided by the present invention; Figure 6 This is a schematic diagram of the timing data visualization results of the IC curve in the first cycle provided by the present invention; Figure 7This is a schematic diagram of the timing data visualization results of the IC curve in the 30th cycle provided by the present invention. Figure 8 This is a schematic diagram of the timing data visualization results of the IC curve in the 60th cycle provided by the present invention. Figure 9 This is a schematic diagram of the timing data visualization results of the IC curve in the 90th cycle provided by the present invention; Figure 10 This is a schematic diagram illustrating the different battery capacity prediction results provided by the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] This invention discloses a method for predicting the remaining capacity of a lithium-ion energy storage battery, such as... Figure 1 This includes the following steps: S1. Based on the historical charge and discharge dataset of energy storage power stations, obtain one-dimensional time series data of lithium battery voltage, current and temperature, as well as one-dimensional IC curve data; S2. Transform the one-dimensional time series data and one-dimensional IC curve during the charge-discharge cycle of lithium batteries into visualized two-dimensional feature images to construct a feature image set for lithium-ion energy storage batteries; S3. Construct a two-dimensional convolutional neural network model, using a set of lithium-ion energy storage battery feature images as input, gradually extracting features from local to global, capturing the potential nonlinear relationship between image features and capacity, and training to obtain a battery remaining capacity prediction model. S4. Real-time charging and discharging data are collected and input into the trained battery remaining capacity prediction model to obtain the remaining capacity prediction result of the lithium-ion energy storage battery.

[0023] To further implement the above technical solution, in step S1, the one-dimensional time series data specifically includes: Battery voltage data is U n ={( t 1, u 1),( t 2, u 2),...,( t n , u n ),}, current data are I n ={(t 1, i 1),( t 2, i 2),...,( t n , i n Temperature data is T n ={( t 1, T 1),( t 2, T 2),...,( t n , T n IC curve data are Y n ={( u 1,dq / dv1),( u 2,dq / dv2),...,( u n ,dq / dv n )},in, t For time, u For voltage, i For current, T dq is the temperature, dv is the capacity difference, dv is the voltage difference, and n is the data length.

[0024] To further implement the above technical solution, the specific content of step S2, which constructs the feature image set of lithium-ion energy storage batteries, is as follows: S21. Normalize the one-dimensional time series data of voltage, current, and temperature, as well as the one-dimensional IC curve data, and use zero-padding to pad the data length, such as... Figures 2-4 ; S22. Using a color mapping mechanism, the data values ​​of voltage, current and temperature are converted into color maps; S23. For the one-dimensional IC curve, the Gram angle field (GAF) is used to convert the time-amplitude of the time-series signal into radius-angle parameters through polar coordinate mapping. A two-dimensional feature image is generated by calculating the cosine values ​​between the angles, such as... Figure 5 ; S24. The images generated in step S22 and step S23 are stitched together to form the final lithium-ion energy storage battery feature image set, which provides reliable input for the subsequent lithium-ion battery capacity prediction model, while significantly reducing the amount of training data and reducing the complexity of data preprocessing.

[0025] To further implement the above technical solution, the method for normalizing the one-dimensional time series data of voltage, current, and temperature in step S21 is as follows:

[0026] in, u k for t k The voltage value at that moment; i k for t k The current value at that moment; T k for t k Temperature value at any given time; for t k The voltage value after time normalization. for t k The current value after time normalization. for t k Temperature value normalized to time; The method for padding data length is as follows: select the data length of the first cycle as the baseline, and if the data length of other cycles is insufficient after normalization, padding is performed using zero padding to ensure that all samples have the same length. Step S22: The normalized voltage, current, and temperature time series are used as row vectors of the two-dimensional image. In the MATLAB environment, the imagesc function is used to map the one-dimensional current, voltage, and temperature signal values ​​into colors to generate color images, with each image corresponding to a physical quantity.

[0027] To further implement the above technical solution, step S23, the calculation method for generating a two-dimensional feature image from a one-dimensional IC curve is as follows: the normalized IC curve is used as a one-dimensional signal, and the inverse cosine value of each data point is calculated to construct the Gram angle field matrix, which is a two-dimensional image. The size is kept the same, and the temporal characteristics and aging information of the IC curve are preserved. Specifically:

[0028]

[0029]

[0030] in, Y For the set of all voltage values, y k For voltage is u k The dq / dv value at that time For normalized y k value, For time, for The polar cosine value of the transformation, G is the Gram angle field GAF, which is the matrix that transforms a one-dimensional time series into a two-dimensional image; GAF extracts time series features based on polar coordinate mapping. It achieves time correlation encoding by linearly increasing the polar angle parameter with the timestamp. The polar angle parameter is generated by the inverse cosine operation of the normalized amplitude. The angular difference preserves the time series fluctuation characteristics. The key advantage of using the GAF method in this embodiment is that when the polar angle θ is constrained to the interval [0,π], the cosine function is strictly monotonic, forming a bijective mapping between the time series and the polar coordinate parameter. The original signal can be accurately reconstructed based on the main diagonal of the Gramian matrix. The polar radius parameter is strictly positively correlated with the timestamp, and the absolute time series relationship is explicitly preserved in two-dimensional space.

[0031] In this embodiment, compared with methods such as S-transform time-frequency analysis, relative displacement moment and recursion graph, GAF not only preserves the structural features of the signal, but also enhances its periodicity. The resulting two-dimensional image shows the global features and local changes of the signal, and has the advantages of reversible reconstruction and high feature resolution.

[0032] To further implement the above technical solution, the specific content of step S3 is as follows: Based on the previous m A set of characteristic images of lithium-ion energy storage batteries with -1 cycle count and corresponding capacity values. C m-1 ={(1, c 1),(2, u 2),...,( m- 1, c m-1 A two-dimensional convolutional neural network model is constructed using a 128×128×3 feature image as input. The model extracts features from local to global levels through convolutional layers, batch normalization layers, ReLU activation layers, and max pooling layers. The number of feature channels increases from 32 to 128. Fully connected layers integrate the features into a 256-dimensional vector, and a Dropout layer prevents overfitting. A loss function is constructed using the actual capacity value and the predicted capacity value, and the Adam algorithm is used to optimize the model parameters.

[0033] in, This is the actual capacity value. Here, M represents the predicted capacity, and M represents the batch size. The first m Voltage under cycle U m ={( t 1, u 1),( t 2, u2),...,( t n , u n )}, Current I m ={( t 1, i 1),( t 2, i 2),...,( t m , i m )},temperature T m ={( t 1, T 1),( t 2, T 2),...,( t m , T m IC curve Y m ={( u 1,(dq / dv)1),( u 2,(dq / dv)2),...,( u n (dq / dv) n The two-dimensional feature image of the data is used as the input to the test samples of the prediction model. m Number of cycles remaining capacity C m Make predictions.

[0034] In another embodiment, taking the aging data of three batteries (BNCA-0.5C-1#, BNCA-0.5C-2#, and BNCA-0.5C-3#) from the Tongji University dataset as an example, IC curves within a voltage range of 3.6 to 3.9V at different cycle counts are selected. The cosine value between the angles is calculated to generate a two-dimensional feature image, thus constructing a lithium-ion energy storage battery feature image set, such as... Figures 6-9 As shown, the two-dimensional convolutional neural network takes a 128×128×3 feature image as input, and extracts features from local to global levels step by step through convolutional layers, batch normalization layers, ReLU activation layers, and max pooling layers, with the number of feature channels increasing from 32 to 128. The fully connected layer integrates the features into a 256-dimensional vector, and the Dropout layer prevents overfitting. Then, the Adam algorithm is used to optimize the model parameters based on the training set, and finally, the healthy features of the test set are input to predict the capacity value.

[0035] Using datasets with different proportions (50%, 70%) as training sets and the remaining data as test sets, the battery capacity prediction results are as follows: Figure 10As shown, Figure 10 Figures (a), (b), and (c) show the capacity prediction results for batteries BNCA-0.5C-1#, BNCA-0.5C-2#, and BNCA-0.5C-3#, respectively. The error indices MAE and RMSE are both at a low level.

[0036] A lithium-ion energy storage battery remaining capacity prediction system, based on a lithium-ion energy storage battery remaining capacity prediction method, includes a data acquisition module, a feature map conversion module, a model building and training module, and a prediction module; The data acquisition module is used to acquire one-dimensional time series data of lithium battery voltage, current and temperature, as well as one-dimensional IC curve data, based on historical charge and discharge datasets of energy storage power stations and real-time acquisition. The feature map conversion module is used to convert one-dimensional time series data and one-dimensional IC curves during the charge and discharge cycle of lithium batteries into visualized two-dimensional feature images, and to construct a feature image set for lithium-ion energy storage batteries. The model building and training module is used to build a two-dimensional convolutional neural network model. It takes a set of lithium-ion energy storage battery feature images as input, gradually extracts local to global features, captures the potential nonlinear relationship between image features and capacity, and trains to obtain a battery remaining capacity prediction model. The prediction module is used to input real-time collected charge and discharge data into a trained battery remaining capacity prediction model to obtain the predicted result of the lithium-ion energy storage battery remaining capacity. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for predicting the remaining capacity of a lithium-ion energy storage battery.

[0037] A processing terminal includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements a method for predicting the remaining capacity of a lithium-ion energy storage battery.

[0038] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0039] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the remaining capacity of a lithium-ion energy storage battery, characterized in that, Includes the following steps: S1. Based on the historical charge and discharge dataset of energy storage power stations, obtain one-dimensional time series data of lithium battery voltage, current and temperature, as well as one-dimensional IC curve data; S2. Transform the one-dimensional time series data and one-dimensional IC curve during the charge-discharge cycle of lithium batteries into visualized two-dimensional feature images to construct a feature image set for lithium-ion energy storage batteries; S3. Construct a two-dimensional convolutional neural network model, using a set of lithium-ion energy storage battery feature images as input, gradually extracting features from local to global, capturing the potential nonlinear relationship between image features and capacity, and training to obtain a battery remaining capacity prediction model. S4. Real-time charging and discharging data are collected and input into the trained battery remaining capacity prediction model to obtain the remaining capacity prediction result of the lithium-ion energy storage battery.

2. The method for predicting the remaining capacity of a lithium-ion energy storage battery according to claim 1, characterized in that, In step S1, the battery voltage data is as follows: U k ={( t 1, u 1),( t 2, u 2),...,( t n , u n The current data is as follows: I k ={( t 1, i 1),( t 2, i 2),...,( t n , i n Temperature data is T k ={( t 1, T 1),( t 2, T 2),...,( t n , T n IC curve data are Y k ={( u 1,(dq / dv)1),( u 2,(dq / dv)2),...,( u n (dq / dv) n )},in, k The number of cycles. t For time, u For voltage, i For current, T dq is the temperature, dv is the capacitance difference, and dv is the voltage difference. n This represents the data length.

3. The method for predicting the remaining capacity of a lithium-ion energy storage battery according to claim 1, characterized in that, The specific content of step S2, constructing the feature image set of lithium-ion energy storage batteries, is as follows: S21. Normalize the one-dimensional time series data of voltage, current and temperature and the one-dimensional IC curve data, and use the zero-padding method to pad the data length. S22. Using a color mapping mechanism, the data values ​​of voltage, current and temperature are converted into color maps; S23. For the one-dimensional IC curve, the Gram angle field (GAF) is used to convert the time-amplitude of the time-series signal into radius-angle parameters through polar coordinate mapping, and a two-dimensional feature image is generated by calculating the cosine value between the angles. S24. The images generated in step S22 and step S23 are stitched together to form the final lithium-ion energy storage battery feature image set.

4. The method for predicting the remaining capacity of a lithium-ion energy storage battery according to claim 3, characterized in that, In step S21, the method for normalizing the one-dimensional time series data of voltage, current, and temperature is as follows: in, u k for t k The voltage value at that moment; i k for t k The current value at that moment; T k for t k Temperature value at any given time; for t k The voltage value after time normalization. for t k The current value after time normalization. for t k Temperature value normalized to time; The method for padding data length is as follows: select the data length of the first cycle as the baseline, and if the data length of other cycles is insufficient after normalization, padding is performed using zero padding to ensure that all samples have the same length. Step S22: The normalized voltage, current, and temperature time series are used as row vectors of the two-dimensional image. In the MATLAB environment, the imagesc function is used to map the one-dimensional current, voltage, and temperature signal values ​​into colors to generate color images, with each image corresponding to a physical quantity.

5. The method for predicting the remaining capacity of a lithium-ion energy storage battery according to claim 3, characterized in that, Step S23, the calculation method for generating a two-dimensional feature image from a one-dimensional IC curve is as follows: the normalized IC curve is used as a one-dimensional signal, and the inverse cosine value of each data point is calculated to construct the Gram angle field matrix, which is a two-dimensional image. The size is kept the same, and the temporal characteristics and aging information of the IC curve are preserved. Specifically: in, Y For the set of all voltage values, y k For voltage is u k The dq / dv value at that time For normalized y k value, For time, for The polar cosine value of the transformation, G is the Gram angle field (GAF), which is the matrix that transforms a one-dimensional time series into a two-dimensional image.

6. The method for predicting the remaining capacity of a lithium-ion energy storage battery according to claim 1, characterized in that, The specific content of step S3 is as follows: Based on the previous m A set of characteristic images of lithium-ion energy storage batteries with -1 cycle count and corresponding capacity values. C m-1 ={(1, c 1),(2, u 2),...,( m- 1, c m-1 A two-dimensional convolutional neural network model is constructed using a 128×128×3 feature image as input. The model extracts features from local to global levels through convolutional layers, batch normalization layers, ReLU activation layers, and max pooling layers. The number of feature channels increases from 32 to 128. Fully connected layers integrate the features into a 256-dimensional vector, and a Dropout layer prevents overfitting. A loss function is constructed using the actual capacity value and the predicted capacity value, and the Adam algorithm is used to optimize the model parameters. The first m Voltage under cycle U m ={( t 1, u 1),( t 2, u 2),...,( t n , u n )}, Current I m ={( t 1, i 1),( t 2, i 2),...,( t m , i m )},temperature T m ={( t 1, T 1),( t 2, T 2),...,( t m , T m IC curve Y m ={( u 1,(dq / dv)1),( u 2,(dq / dv)2),...,( u n (dq / dv) n The two-dimensional feature image of the data is used as the input to the test samples of the prediction model. m Number of cycles remaining capacity C m Make predictions.

7. A lithium-ion energy storage battery remaining capacity prediction system, characterized in that, A method for predicting the remaining capacity of a lithium-ion energy storage battery according to any one of claims 1-6 includes a data acquisition module, a feature map conversion module, a model building and training module, and a prediction module. The data acquisition module is used to acquire one-dimensional time series data and one-dimensional IC curve data of lithium battery voltage, current and temperature based on historical charge and discharge datasets of energy storage power stations and real-time acquisition. The feature map conversion module is used to convert one-dimensional time series data and one-dimensional IC curves during the charge and discharge cycle of lithium batteries into visualized two-dimensional feature images, and to construct a feature image set for lithium-ion energy storage batteries. The model building and training module is used to build a two-dimensional convolutional neural network model. It takes a set of lithium-ion energy storage battery feature images as input, gradually extracts local to global features, captures the potential nonlinear relationship between image features and capacity, and trains to obtain a battery remaining capacity prediction model. The prediction module is used to input real-time collected charge and discharge data into the trained battery remaining capacity prediction model to obtain the prediction result of the remaining capacity of the lithium-ion energy storage battery.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements a method for predicting the remaining capacity of a lithium-ion energy storage battery as described in any one of claims 1-6.

9. A processing terminal, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes a computer program, it implements a method for predicting the remaining capacity of a lithium-ion energy storage battery as described in any one of claims 1-6.

Images