A multi-scale lithium-ion battery state-of-charge estimation method based on swelling force

Abstract

The present application relates to the technical field of lithium ion battery state evaluation, in particular to a multi-scale lithium ion battery state of charge evaluation method based on expansion force, comprising the following steps: collecting current, voltage and expansion force signals during discharging; normalizing the collected data to eliminate the influence of different types of data dimensions; inputting the processed data into a multi-scale SOC estimation algorithm; performing reverse normalization on the output data, and finally outputting the estimated value; the beneficial effects are that: the multi-scale lithium ion battery state of charge evaluation method based on expansion force proposed by the present application selects a thin film type expansion force sensor, does not need to occupy additional space, and can be integrated without additional modification of the battery pack; under various sampling intervals, the method can ensure an SOC evaluation accuracy within about 1%, and the prediction accuracy can be higher than that of existing methods.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery state assessment technology, specifically to a multi-scale lithium-ion battery state-of-charge assessment method based on expansion force. Background Technology

[0002] Due to the energy crisis and global commitments to reduce greenhouse gas emissions, electric vehicles have become a viable alternative to traditional automobiles. Lithium-ion batteries, characterized by high energy density and long cycle life, are widely recognized as the primary energy storage device for electric vehicles. Therefore, effectively monitoring battery status, including state of charge (SOC), is crucial for ensuring the safe, reliable, and efficient operation of electric vehicles.

[0003] In existing technologies, lithium-ion battery SOC assessment methods mainly include the ampere-hour integration method or open-circuit voltage method (OCV), model-based methods, and data-driven methods. The ampere-hour integration method calculates battery SOC by integrating the current; however, even small errors in measurement can lead to significant estimation biases. The open-circuit voltage method establishes a mapping relationship between OCV and SOC, but OCV can only be obtained after the battery has been resting for tens of minutes to reach equilibrium, which hinders the application of these methods in electric vehicles. Model-based methods can provide accurate SOC estimates, but determining the various parameters in the model presents a significant challenge. Furthermore, the development of battery models requires a deep understanding of the electrochemical characteristics of the battery under test, which reduces the universality of the established model. Data-driven methods can provide accurate SOC estimates, but their accuracy largely depends on the data sampling interval. In rapidly changing scenarios such as electric vehicles, large sampling intervals may fail to capture the dynamic characteristics of battery SOC, thus affecting the accuracy of SOC estimation. Therefore, current methods cannot provide accurate estimates over large data sampling intervals. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-scale lithium-ion battery state-of-charge assessment method based on expansion force, in order to solve the problem that existing technologies mentioned in the background art, such as the ampere-hour integration method, open-circuit voltage method, model-based methods, or data-driven methods, can only be used for assessment at small sampling intervals and are difficult to meet the needs of practical applications.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multi-scale lithium-ion battery state-of-charge assessment method based on expansion force, the method comprising the following steps:

[0006] Acquire current, voltage, and expansion force signals during the discharge process;

[0007] The collected data is normalized to eliminate the influence of different data units.

[0008] The processed data is then input into the multi-scale SOC estimation algorithm;

[0009] The output data is inversely normalized, and finally the estimated value is output.

[0010] Preferably, when collecting current, voltage, and expansion force signals during the discharge process, a sliding window is used to smooth the data, while continuous data within the time window is analyzed to capture the trend of battery SOC change over time and make an estimate accordingly.

[0011] Preferably, the multi-scale SOC estimation algorithm uses the least squares method to adjust the weights of the combined prediction model to minimize the prediction error:

[0012]

[0013] Where ω k As the weight, x i For a true SOC, This is the estimated SOC value.

[0014] Preferably, the input is normalized using the min-max normalization method to eliminate the influence of different dimensions between parameters and ensure the comparability of data indicators. The formula is as follows:

[0015]

[0016] Where x t x represents the battery parameters at time t. t,max and x t,min These are the maximum and minimum values, respectively. These are the normalized eigenvalues.

[0017] Preferably, the root mean square error (RMSE), mean absolute error (MAE), and the maximum absolute value (MaxAE) are calculated:

[0018]

[0019]

[0020] Where N represents the number of data points, y i Represents real data. This represents the predicted data.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] The proposed multi-scale lithium-ion battery state-of-charge (SOC) assessment method based on expansion force uses a thin-film expansion force sensor, which does not require additional space and can be integrated into the battery pack without additional modifications. Under various sampling intervals, this method can guarantee an SOC assessment accuracy of less than 1%, and its prediction accuracy can reach a level higher than that of existing methods. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the LSTM unit structure of the present invention;

[0024] Figure 2 This is a schematic diagram of the multi-scale SOC estimation algorithm structure of the present invention;

[0025] Figure 3 This is a flowchart of the method of the present invention;

[0026] Figure 4 This is a schematic diagram of the battery testing platform structure of the present invention;

[0027] Figure 5 This is a schematic diagram of the dynamic operating conditions of the present invention;

[0028] Figure 6 The following are the SOC estimation results of the present invention under NEDC cycles with different sampling intervals: (a) SOC for 1 second; (b) SOC error for 1 second; (c) SOC for 5 seconds; (d) SOC error for 5 seconds; (e) SOC for 10 seconds; (f) Schematic diagram of SOC error for 10 seconds;

[0029] Figure 7 The following are the SOC estimation results of the present invention under UDDS cycles with different sampling intervals: (a) SOC for 1 second; (b) SOC error for 1 second; (c) SOC for 5 seconds; (d) SOC error for 5 seconds; (e) SOC for 10 seconds; (f) Schematic diagram of SOC error for 10 seconds;

[0030] Figure 8 Statistical analysis of the SOC evaluation method of this invention: (a) RMSE; (b) MaxAE; (c) MAE schematic diagram. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit 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.

[0032] Example 1

[0033] Please see Figure 3 This invention provides a technical solution: a multi-scale lithium-ion battery state-of-charge assessment method based on expansion force, the method comprising the following steps:

[0034] Acquire current, voltage, and expansion force signals during the discharge process;

[0035] The collected data is normalized to eliminate the influence of different data units.

[0036] The processed data is then input into the multi-scale SOC estimation algorithm;

[0037] The output data is inversely normalized, and finally the estimated value is output.

[0038] The long-term estimate of SOC (i.e., the global trend) is established by extracting information from expansion force and voltage data using a Long Short-Term Memory (LSTM) network, while the short-term estimate of SOC (i.e., local variations) is obtained by extracting information from current and voltage measurements using Support Vector Regression (SVR). The long-term and short-term (i.e., multi-scale) SOC estimates are then fused to provide the final SOC estimate. Both long-term and short-term estimates use voltage measurements because they contain valuable information about both the global trend and local variations of SOC. This approach allows the long-term and short-term estimates to complement each other, effectively mitigating the dependence on smaller data sampling intervals.

[0039] LSTM is a recurrent neural network (RNN) architecture used in deep learning to process and predict time series data. Compared to traditional RNN architectures, LSTM introduces a method called a "gating mechanism," which can more effectively handle long-term dependencies and the vanishing gradient problem. Therefore, we choose LSTM to estimate long-term SOC using long-term variables such as dilatational forces; the structure of an LSTM unit is as follows... Figure 1 As shown. It contains a forget gate (f t ), an input gate (i t ) and an output gate (o t The forget gate determines the state from the cell (c). t The input gate determines which information to discard from the current input (x). t ) and the hidden state of the previous time step (l t-1 The hidden state is determined by which new information is retrieved from the cell state and then updated to the cell state. The output gate determines how much information is output from the cell state to the hidden state at the next time step. The specific formula is as follows:

[0040]

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] SVR (Structural Risk Minimization) is a technique for solving regression problems. It is an algorithm based on statistical learning theory and the principle of structural risk minimization. The core idea of ​​SVR is to map the input vector to a high-dimensional space and construct an optimal separating hyperplane, then use convex optimization to obtain the global optimum. SVR uses kernel functions and regression algorithms to transform nonlinear models into linear models. Based on SVR's advantages in handling nonlinear data, we choose SVR to estimate short-term SOC (State of Charge) using short-term variables such as current.

[0047] For a given sample D = {(x1, y1), (x2, y2), ..., (x...} n y n )},y i ∈R, the goal is to obtain a regression function that makes f(x) as close as possible to y.

[0048]

[0049] w is the weight vector. It is a mapping function that maps the input feature vector to a high-dimensional feature space, and b is the bias term.

[0050] Equation (7) is transformed into the objective function, where C is the regularization parameter used to balance the complexity of the regression function and the weight of the fitting error, and ε is the tolerance, representing the maximum difference between the predicted value and the true value.

[0051]

[0052]

[0053] Introducing slack variable ξ i , representing the allowable functional margin deviation for the i-th sample, can be transformed into the objective function.

[0054]

[0055]

[0056]

[0057]

[0058] By introducing the Lagrange function and taking its partial derivatives, we obtain the regression function.

[0059]

[0060]

[0061] α is the Lagrange coefficient, and K is the kernel function.

[0062] Example 2

[0063] Building upon Example 1, as the data sampling interval increases, the battery current and voltage data collected under dynamic conditions may fail to capture the dynamic characteristics of the battery's State of Charge (SOC), which will severely affect the stability of the SOC evaluation. Therefore, this patent selects an appropriate sliding window width. Assuming the sliding window size is λ, that is, the composition of the t-th input vector is {x... t-λ+1 ,…,x t-1 ,x t Using a sliding window allows for data smoothing, while simultaneously analyzing continuous data within the time window to capture trends in battery SOC over time and make estimations accordingly.

[0064] Figure 2 The proposed multi-scale SOC estimation method is described. A multi-scale SOC estimation method is constructed by combining long-term and short-term SOC estimations, and the weights of the combined prediction model are adjusted using the least squares method to minimize prediction error. The long-term SOC estimate (i.e., the global trend) is established by extracting information from expansion force and voltage data using a Long Short-Term Memory (LSTM) network, while the short-term SOC estimate (i.e., local variations) is obtained by extracting information from current and voltage measurements using Support Vector Regression (SVR). Then, the long-term and short-term (i.e., multi-scale) SOC estimation results are fused to provide the final SOC estimation result. Both long-term and short-term estimates use voltage measurements because they contain valuable information about the global trend and local variations of SOC. Using this method, long-term and short-term estimates can complement each other, effectively alleviating the dependence on small data sampling intervals.

[0065]

[0066] Where ω k As the weight, x i For a true SOC, This is the estimated SOC value.

[0067] Meanwhile, in order to eliminate the influence of different dimensions between parameters and ensure the comparability of data indicators, the min-max normalization method is used to normalize the input. The specific method is as follows:

[0068]

[0069] Where x t x represents the battery parameters at time t. t,max and x t,min These are the maximum and minimum values, respectively. These are the normalized eigenvalues.

[0070] To evaluate the effectiveness of the proposed method, we use formulas (18-19) to calculate the root mean square error (RMSE), mean absolute error (MAE), and the maximum absolute value (MaxAE):

[0071]

[0072]

[0073] Where N represents the number of data points, y i Represents real data. This represents the predicted data.

[0074] Example 3

[0075] Based on Example 2, the lithium battery test bench, as follows: Figure 4 As shown, the battery is fixed on the expansion force testing device and placed in a constant temperature chamber. The battery is charged and discharged by the power supply cabinet, and the computer collects the current, voltage, and expansion force data. Table 1 lists the detailed parameters of the battery.

[0076] Table 1. Test Battery Parameters

[0077]

[0078]

[0079] To comprehensively understand the changes in expansion force during the charging and discharging of lithium-ion batteries, it is necessary to test the batteries under different dynamic testing conditions. Specific testing conditions are shown in Table 2. The expansion force testing device applies a preload of 25 kg to the battery and charges it to a cutoff voltage of 4.2V and a cutoff current of 0.05C in constant current and constant voltage mode. During the test, the following procedures are executed: Figure 5 The New European Driving Cycle (NEDC) and Urban Dynamometer Driving Schedule (UDDS) dynamic conditions are shown.

[0080] Table 2. Battery charging and discharging test scheme

[0081]

[0082] To further evaluate the performance of the proposed method, we introduce some existing methods for comparison, including LSTM (defined as FLSTM) with inputs of expansion force, voltage, and current, and SVR with inputs of voltage and current. The results are as follows: Figure 6 , Figure 7 and Figure 8 As shown.

[0083] Figure 6 The performance comparison of the three methods for SOC estimation under NEDC conditions is shown. It can be seen that, with a data sampling interval of 1 second, all three methods can achieve reliable battery SOC estimation, while the proposed method has the lowest RMSE, MaxAE, and MAE at 0.44%, 1.58%, and 0.32%, respectively. Furthermore, due to the inclusion of expansion force measurement, the SOC estimation results of FLSTM and the proposed method show less fluctuation compared to SVR (see...). Figure 6 (a) and (b)).

[0084] As the sampling interval increased from 1 second to 5 seconds, the estimation accuracy of FLSTM and SVR decreased sharply, while the RMSE and MAE of SVR nearly doubled. Conversely, the estimation accuracy of the method proposed in the patent remained almost unchanged, with its RMSE, MaxAE, and MAE increasing from 0.44%, 1.58%, and 0.32% to 0.46%, 1.86%, and 0.34%, respectively. Furthermore, when the sampling interval was further increased to 10 seconds, the RMSE values ​​of FLSTM and SVR were 1.4% and 2%, respectively, and their MAE values ​​were 1.21% and 1.61%, respectively, while the RMSE and MAE values ​​of the method proposed in the patent were 0.64% and 0.48%, respectively—less than half and a third of those of FLSTM and SVR. Regarding MaxAE, FLSTM and SVR achieved 4.89% and 7.45%, respectively, while the method proposed in the patent could control it to around 2.84%. Figure 7 and Figure 8 The estimation results for the UDDS operating condition shown are similar to those for the NEDC operating condition. Based on the above results, it can be concluded that the method proposed in the patent can provide better SOC estimation accuracy while maintaining high stability.

[0085] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-scale method for assessing the state of charge (SOC) of lithium-ion batteries based on expansion force, characterized in that: The method includes the following steps: Acquire current, voltage, and expansion force signals during the discharge process; The collected data is normalized to eliminate the influence of different data units. The processed data is input into a multi-scale SOC estimation algorithm. A multi-scale SOC estimation method is constructed by combining long-term and short-term SOC estimations, and the weights of the combined prediction model are adjusted using the least squares method to minimize the prediction error. The long-term SOC estimate is established by extracting information from expansion force and voltage data using a Long Short-Term Memory (LSTM) network, while the short-term SOC estimate is obtained by extracting information from current and voltage measurements using Support Vector Regression (SVR). Then, the long-term and short-term SOC estimation results are fused to provide the final SOC estimation result. The output data is inversely normalized, and finally the estimated value is output.

2. The method for assessing the state of charge of a lithium-ion battery based on expansion force according to claim 1, characterized in that: When collecting current, voltage, and expansion force signals during the discharge process, a sliding window is used to smooth the data. At the same time, continuous data within the time window is analyzed to capture the trend of battery SOC change over time and to make an estimate based on this.

3. The method for assessing the state of charge of a lithium-ion battery based on expansion force according to claim 1, characterized in that: The multi-scale SOC estimation algorithm uses the least squares method to adjust the weights of the combined prediction model to minimize the prediction error. in As weight, For a true SOC, This is the estimated SOC value.

4. The method for assessing the state of charge of a lithium-ion battery based on expansion force according to claim 3, characterized in that: The input is normalized using the min-max normalization method to eliminate the influence of different dimensions between parameters and ensure the comparability of data indicators. The formula is as follows: in express Battery parameters at specific times and These are the maximum and minimum values, respectively. These are the normalized eigenvalues.

5. The method for assessing the state of charge of a lithium-ion battery based on expansion force according to claim 1, characterized in that: Calculate the root mean square error (RMSE), mean absolute error (MAE), and maximum absolute value (MaxAE): in Indicates the number of data points. Represents real data. This represents the predicted data.

Images