Lithium ion battery available capacity estimation method and system based on ultrasonic signal

Abstract

The invention discloses a lithium ion battery available capacity estimation method and system based on an ultrasonic signal, and relates to the technical field of battery management, and the method comprises the steps: collecting the current, capacity, ultrasonic feedback signals and corresponding sampling time data of a lithium ion battery in a charge-discharge cycle constant-current charge-discharge stage; processing the ultrasonic feedback signal to obtain the amplitude of the ultrasonic feedback signal, and obtaining the ultrasonic flight time from the ultrasonic feedback signal; calculating the correlation between the amplitude of the ultrasonic signal and the capacity information, calculating the correlation between the flight time information of the ultrasonic signal and the capacity information, collecting the ultrasonic signal, extracting the characteristics of the amplitude and the flight time, and carrying out the capacity estimation in combination with a multi-characteristic data set and an optimization model. The technical problem that the traditional electrical parameter monitoring method cannot directly reflect the physical and chemical state change in the battery is solved, and the method has the advantages that the physical state change in the battery is directly reflected through the ultrasonic signal, the capacity estimation precision is improved, and the equipment complexity is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery management, in particular to a lithium ion battery available capacity estimation method and system based on ultrasonic signals. BACKGROUND

[0002] The available capacity of the battery to some extent reflects the aging condition and health degree of the battery, and is an important parameter reflecting the service life of the battery. With the increase of use time and the increase of charge and discharge cycle times, the available capacity of the battery will gradually decrease. In the application of electric vehicles, accurate evaluation of the capacity can help predict the remaining life of the battery, plan the replacement cycle of the battery, and effectively manage the energy distribution of the battery, thereby prolonging the service life of the equipment.

[0003] Under the current battery design concept of advocating high efficiency, safety and sustainable development, how to accurately and reliably monitor and evaluate the working state of lithium battery has become a key link for optimizing charge and discharge strategy, avoiding battery abuse and improving operation efficiency. At present, although the traditional method based on current, voltage and temperature and other electrical signals has been widely recognized in practical application, its estimation accuracy and reliability decrease significantly when facing battery aging and complex environmental temperature fluctuations. In addition, these methods usually obtain information through indirect monitoring, mainly relying on electrochemical characteristics or electrical parameters such as impedance spectrum or coulomb counting. These methods are often limited by idealized assumptions, require complete charge and discharge cycles, and have insufficient generalization ability for different battery systems. At the same time, it is difficult to fully reflect the internal physical and chemical state of the battery, and more complex and accurate detection methods are limited by equipment deployment difficulties, high detection cost, irreversible damage to the battery and other problems. Therefore, it is particularly important to introduce a new battery available capacity estimation method. SUMMARY

[0004] The purpose of the present application is to provide a lithium ion battery available capacity estimation method and system based on ultrasonic signals, in order to overcome the technical problems in the prior art that the feature extraction of external characteristic parameters is insufficient to represent the internal state of the battery, simple and efficient non-destructive detection is lacking, and accurate battery available capacity estimation is difficult to achieve.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is as follows: A lithium ion battery available capacity estimation method based on ultrasonic signals, comprising the following steps: S1, collecting current, capacity, ultrasonic feedback signal and corresponding sampling time data of the lithium ion battery in the constant current charge and discharge phase of the charge and discharge cycle; S2, processing the ultrasonic feedback signal based on wavelet packet transform to obtain the amplitude of the ultrasonic feedback signal, and obtaining the time of flight of the ultrasonic signal from the ultrasonic feedback signal based on the cross-correlation algorithm; S3, calculating the correlation degrees of the amplitude of the ultrasonic signal, the time-of-flight information of the ultrasonic signal, and the capacity information respectively based on a Pearson correlation algorithm, and grouping the amplitude of the ultrasonic signal and the time-of-flight information of the ultrasonic signal to generate a multi-feature data set; S4, establishing a capacity estimation model based on CNN-SE-BILSTM based on the multi-feature data set, as shown in the model algorithm framework Figure 3 as shown, using the sparrow optimization algorithm shown in Figure 4 to optimize the model parameters of the capacity estimation model, and using the capacity estimation model after the model parameter optimization to estimate the available capacity of the battery.

[0006] Preferably, the current collected in the constant-current charging phase of the lithium ion battery in the charge-discharge cycle is specifically a sequence of current data in the constant-current charging phase of the kth cycle , wherein and are the initial current and the terminal current in the constant-current charging phase, respectively. The capacity collected in the constant-current charging phase of the lithium ion battery in the charge-discharge cycle is specifically a sequence of charging capacity data in the constant-current charging phase , , wherein is the constant-current charging capacity of the kth cycle battery, and are the current and the sampling time of the kth cycle, respectively. The ultrasonic feedback signal collected in the constant-current charging phase of the lithium ion battery in the charge-discharge cycle is specifically a sequence of constant-current charging ultrasonic data of the kth cycle , wherein and are the initial ultrasonic signal at the constant-current charging phase and the terminal ultrasonic signal at the constant-current discharging phase, respectively. The sampling time data collected in the constant-current charging phase of the lithium ion battery in the charge-discharge cycle is specifically a sequence of constant-current charging sampling time data of the kth cycle , wherein and are the initial sampling time and the terminal sampling time in the constant-current charging phase, respectively.

[0007] Preferably, the amplitude of the ultrasonic feedback signal obtained by processing the ultrasonic feedback signal based on the wavelet packet transform includes: Based on the wavelet packet transform, the low-frequency and high-frequency components of the ultrasonic feedback signal are recursively decomposed to obtain a multi-resolution frequency representation of the ultrasonic feedback signal, i.e., the amplitude of the ultrasonic feedback signal.

[0008] Preferably, the mother wavelet function of the wavelet packet transform has the following expression form:

[0009] In the formula It is the energy normalization factor; a For scale parameters, b For translation parameters; for ultrasonic feedback signals x(n) The formula for wavelet packet transform is shown below:

[0010] In the formula The result of wavelet transform, The raw data of the ultrasonic feedback signal, in its discrete form, has the following low-frequency component, with the low-frequency component sub-nodes as shown below. 2n :

[0011] Its high-frequency component is shown in the following formula, and the child nodes of the high-frequency component are: 2n+1 :

[0012] In the formula j The wavelet packet decomposition level is [number of layers]. n For node position, k and m This refers to discrete time or location. x j,n [k] For the first j Layer, First n A discrete signal sequence of sub-bands; h[m] and g[m] These are the coefficients of the low-pass and high-pass filters.

[0013] Preferably, the method of obtaining the time of flight of ultrasound from the ultrasound feedback signal based on the cross-correlation algorithm specifically includes: comparing one signal with another signal through a series of different time delays, thereby finding the moment when the two signals are most similar, and extracting the time difference at this moment as the time of flight of ultrasound.

[0014] Preferably, the method for obtaining the time-of-flight of ultrasound from the ultrasound feedback signal based on the cross-correlation algorithm is specifically shown in the following formula:

[0015] The transmitted signal is The received signal is ,in It is the time delay, representing the relative displacement between the transmitted and received signals. When the value is at its maximum, the correlation between the two signals is at its maximum. For ultrasonic time of flight, a discrete representation is made, as shown in the following formula:

[0016] In the formula represents the value of the signal X at the discrete time point n , represents the value of the signal Y at the discrete time point n+ k , k is a lag quantity, and represents the time offset of the signal Y relative to the signal X .

[0017] Preferably, the Pearson correlation coefficient is used to measure the degree of linear correlation between two variables, and the value range is [-1, 1], which is used to judge the strength and direction of the linear relationship between two variables. The calculation formula is as follows:

[0018] In the formula X i , Y i is the value of the ith sample point, , is the sample mean, X , Y are two related variables.

[0019] Preferably, the sparrow optimization algorithm is used to optimize the model parameters of the capacity estimation model, which includes the following steps: g, set the hyperparameter space, determine the optimized hyperparameters and their ranges; h, initialize the sparrow population, and the position of each sparrow represents a set of hyperparameters; i, establish a fitness function to minimize the error; j, train the model using the current sparrow and output the fitness; k, assign roles and positions according to the fitness value, and update the hyperparameter combination; l, if the maximum number of iterations has not been reached or the convergence condition is met, return to d, otherwise select the optimal hyperparameters and output the hyperparameter combination with the optimal fitness value; According to the optimized hyperparameters, the model is trained, and the battery available capacity is estimated.

[0020] An ultrasonic signal-based lithium-ion battery available capacity estimation system includes a data acquisition module, a data preprocessing module, a feature calculation module, and a capacity estimation module: A data acquisition module acquires current, capacity, ultrasonic feedback signal and corresponding sampling time data of the lithium ion battery in a constant current charging and discharging phase of a charging and discharging cycle. A data preprocessing module processes the ultrasonic feedback signal based on wavelet packet transform to obtain an amplitude of the ultrasonic feedback signal, and obtains a time of flight of the ultrasonic signal from the ultrasonic feedback signal based on a cross-correlation algorithm. A feature calculation module calculates a correlation degree of the amplitude of the ultrasonic signal and the time of flight information of the ultrasonic signal with the capacity information respectively based on a Pearson correlation algorithm, and groups the amplitude of the ultrasonic signal and the time of flight information of the ultrasonic signal to generate a multi-feature data set. A capacity estimation module establishes a capacity estimation model based on CNN-SE-BILSTM based on the multi-feature data set as shown in Figure 3 A sparrow optimization algorithm shown in Figure 4 is used to optimize the model parameters of the capacity estimation model, and the capacity estimation model after the model parameter optimization is used for battery available capacity estimation.

[0021] Preferably, processing the ultrasonic feedback signal based on wavelet packet transform to obtain the amplitude of the ultrasonic feedback signal specifically includes: The low-frequency and high-frequency components of the ultrasonic feedback signal are recursively decomposed based on wavelet packet transform to obtain a multi-resolution frequency representation of the ultrasonic feedback signal, that is, the amplitude of the ultrasonic feedback signal.

[0022] Compared with the prior art, the present application has the following beneficial technical effects: The present application discloses a kind of based on ultrasonic signal's lithium ion battery available capacity estimation method, acquisition lithium ion battery in the constant current charging and discharging phase of charging and discharging cycle current, capacity, ultrasonic feedback signal and corresponding sampling time data;Ultrasonic feedback signal is processed to obtain the amplitude of the ultrasonic feedback signal, and the time of flight of the ultrasonic signal is obtained from the ultrasonic feedback signal;The correlation degree of the amplitude of the ultrasonic signal and the time of flight information of the ultrasonic signal with the capacity information is calculated respectively, the amplitude and the time of flight feature are extracted by collecting ultrasonic signal, and capacity estimation is carried out by combining multi-feature data set and optimization model, solve the technical problems that traditional electrical parameter monitoring method cannot directly reflect the change of battery internal physical and chemical state, with the advantages of directly reflecting the change of battery internal physical state through ultrasonic signal, improving capacity estimation precision, reducing equipment complexity. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 It is the flow chart of the lithium ion battery available capacity estimation method based on ultrasonic signal in the embodiment of the present application.

[0024] Figures 2a-2f It is the ultrasonic experimental data graph in the embodiment of the present application. Figure 2aultrasound signal amplitude change under different aging cycle numbers; Figure 2b time of flight change under different aging cycle numbers; Figure 2c initial signal amplitude change with cycle number; Figure 2d maximum signal amplitude change with cycle number; Figure 2e minimum signal amplitude change with cycle number; Figure 2f initial time of flight change with cycle number.

[0025] Figure 3 is a structural schematic diagram of the CNN-SE-BILSTM model in the embodiment of the application.

[0026] Figure 4 is a sparrow optimization flowchart in the embodiment of the application.

[0027] Figure 5a is a prediction result graph of the lithium ion battery available capacity data in the application; Figure 5b is an available capacity estimation error value. DETAILED DESCRIPTION

[0028] In order for those skilled in the art to better understand the application scheme, the technical solutions in the embodiments of the application will be clearly and completely described below in conjunction with the drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0030] In the prior art, battery capacity evaluation mainly relies on traditional methods of electrical signals such as current, voltage and temperature. These methods have a significant decrease in estimation accuracy when the battery ages or the ambient temperature fluctuates, and need to rely on complete charge and discharge cycle data, making it difficult to reflect the internal physical and chemical state of the battery. For example, methods based on impedance spectroscopy or coulomb counting have limitations of idealized assumptions, while high-precision detection methods face problems of complex equipment, high cost or damage to the battery. In the context of electric vehicle battery management, the prior art cannot obtain the internal state changes of the battery in real time under non-invasive conditions, resulting in a deviation between the capacity estimation result and the actual aging degree.

[0031] To solve the above problems, it is found through analysis that the amplitude change of the ultrasonic signal may reflect the micro-deformation of the electrode material, and the time-of-flight difference may represent the change in ion migration rate of the electrolyte. Further thinking finds that a single feature cannot fully represent the capacity attenuation, and a multi-dimensional feature fusion and deep learning model are needed, but the existing model parameter optimization efficiency is low, affecting the estimation accuracy. Therefore, it is necessary to construct a capacity estimation framework with multi-feature fusion, and introduce an intelligent optimization algorithm to improve the model performance.

[0032] As shown in Figure 1 , the present application provides a lithium ion battery available capacity estimation method based on ultrasonic signals, which can non-invasively obtain key information such as electrode swelling and internal concentration changes by analyzing the propagation characteristics (sound speed, attenuation, reflection spectrum) of sound waves inside the battery, providing multi-dimensional physical parameters for lithium ion battery available capacity estimation, improving the efficiency and accuracy of battery capacity estimation, specifically including the following steps: S1, collecting current, capacity, ultrasonic feedback signal and corresponding sampling time data of lithium ion battery in constant current charge and discharge phase of charge and discharge cycle, the ultrasonic signal data collection result is as shown in Figures 2a to 2f ; In the specific embodiments of the present application, the current collected in the constant current charge and discharge phase of the lithium ion battery in the charge and discharge cycle is specifically: the current data of the kth cycle in the constant current charge and discharge phase is a sequence , and are the initial current and the terminal current in the constant current charging phase, respectively; The capacity collected in the constant current charge and discharge phase of the lithium ion battery in the charge and discharge cycle is specifically: the charging capacity data in the constant current charging phase is a sequence , , is the constant current charging capacity of the kth cycle battery, and are the current and sampling time of the kth cycle, respectively; The ultrasonic feedback signal collected by the lithium ion battery in the constant current charging and discharging stage of the charge-discharge cycle is specifically: the constant current charging and discharging ultrasonic data of the kth cycle is a sequence , wherein , and are the initial charging ultrasonic signal and the terminal discharging ultrasonic signal in the constant current charging stage, respectively. The sampling time data collected by the lithium ion battery in the constant current charging and discharging stage of the charge-discharge cycle is specifically: the constant current charging and discharging sampling time data of the kth cycle is a sequence , wherein , and are the initial sampling time and the terminal sampling time in the constant current charging stage, respectively.

[0033] S2, based on wavelet packet transform, the amplitude of the ultrasonic feedback signal is obtained, and the time of flight of the ultrasonic signal is obtained from the ultrasonic feedback signal based on the cross-correlation algorithm; In the specific embodiment of the present application, the amplitude of the ultrasonic feedback signal obtained by processing the ultrasonic feedback signal based on wavelet packet transform specifically includes: Based on wavelet packet transform, the low-frequency and high-frequency components of the ultrasonic feedback signal are recursively decomposed to obtain a multi-resolution frequency representation of the ultrasonic feedback signal, i.e. the amplitude of the ultrasonic feedback signal.

[0034] The expression form of the mother wavelet function of the wavelet packet transform is as follows:

[0035] In the formula, is an energy normalization factor to keep the energy exchange unchanged; a is a scale parameter that determines the width of the wavelet (i.e. time resolution and frequency resolution), b is a translation parameter that determines the position of the wavelet on the time axis; the wavelet mother function is expanded by the recursive formula to cover the entire frequency range. The formula of the ultrasonic feedback signal x(n) The wavelet packet transform formula is as follows:

[0036] In the formula, is the result of wavelet transform, also known as wavelet coefficient, is the original data of the ultrasonic feedback signal, in practical application, the data is often discrete, usually realized by a filter bank. Therefore, its discrete form is used, and its low-frequency component (subnode is 2n ) is as follows:

[0037] The high-frequency component (subnode is 2n+1 ) is as follows:

[0038] In the formula j is the wavelet packet decomposition layer number, n is the node position, k and m is discrete time or position, x j,n [k] is the first j layer, the first n discrete signal sequence of the subband; h[m] and g[m] are low-pass and high-pass filter coefficients.

[0039] In order to determine the displacement of the waveform when the ultrasound propagates (i.e. the time of flight), the time of flight of the ultrasound is obtained from the ultrasound feedback signal based on a cross-correlation algorithm, which is a mathematical tool for measuring the similarity or correlation between two signals; specifically, one signal is compared with another signal through a series of different time delays, so as to find the most similar time of the two signals, and the time difference at this time is extracted as the time of flight of the ultrasound. As shown in the following formula:

[0040] In the formula, the transmitted signal is , the received signal is , wherein is the time delay, indicating the relative displacement between the transmitted signal and the received signal, when is the maximum value, the correlation between the two signals is the largest, at this time is the time of flight of the ultrasound, which needs to be discretely expressed in practical application, as shown in the following formula:

[0041] In the formula, represents the value of the signal X at the discrete time point n , represents the value of the signal Y at the discrete time point n+ k , k is the lag amount, indicating the time offset of the signal Y relative to the signal X .

[0042] S3, based on the Pearson correlation algorithm, the correlation degrees of the amplitude of the ultrasound signal, the time of flight information of the ultrasound signal and the volume information are calculated, and the amplitude of the ultrasound signal and the time of flight information of the ultrasound signal are grouped to generate a multi-feature data set; In a specific embodiment of the present invention, the Pearson correlation coefficient is used to measure the degree of linear correlation between two variables, with a value range of [-1, 1]. It is used to determine the strength and direction of the linear relationship between the two variables, and the calculation formula is as follows:

[0043] In the formula X i , Y i Let i be the value of the i-th sample point. , The sample mean. X , Y These are two related variables.

[0044] Pearson correlation coefficients were used to obtain the Pearson coefficients for initial flight time and battery capacity, initial ultrasonic amplitude and battery capacity, maximum signal amplitude and battery capacity, and minimum signal amplitude and battery capacity, respectively.

[0045] S4, based on a multi-feature dataset, establishes... Figure 3 The capacity estimation model of CNN-SE-BILSTM shown uses... Figure 4 The Sparrow Optimization Algorithm optimizes the parameters of the capacity estimation model and uses the optimized capacity estimation model to estimate the available battery capacity.

[0046] The ultrasonic feedback signal of this invention is the reflected signal of the internal structure of the battery collected by an ultrasonic sensor. Specifically, it can be realized by using a piezoelectric transducer to transmit and receive ultrasonic waves. Its waveform changes reflect the deformation of the electrode material and the distribution state of the electrolyte.

[0047] During the constant current charging and discharging phase, current, capacity, and ultrasonic signal data were simultaneously acquired. Wavelet packet decomposition was performed on the ultrasonic signals to extract the energy amplitude of each frequency band as a characterization of electrode structure changes. The time delay of transmitted and received signals was calculated using a cross-correlation algorithm to obtain the propagation time of ultrasonic waves inside the battery, reflecting the electrolyte ion transport characteristics. Pearson coefficients were used to screen ultrasonic features strongly correlated with capacity, constructing a multi-feature dataset including amplitude, time-of-flight, and current parameters. These multi-features were input into a CNN-SE-BILSTM model, where convolutional layers extract spatial features, the SE module adaptively adjusts channel weights, and a bidirectional LSTM captures the temporal patterns of the charging and discharging process. The sparrow algorithm was used to optimize hyperparameters such as the learning rate and the number of convolutional kernels. The optimal parameter combination was found by iteratively updating the population position, ultimately outputting a high-precision capacity estimation result. Figure 5a and Figure 5b As shown.

[0048] The application automatically searches for optimal parameters through an intelligent optimization algorithm to improve the model generalization capability. In addition, the prior art requires complete charging and discharging data, which can be collected in real time during the charging and discharging stage to realize dynamic capacity monitoring. Through the above technical solutions, the application realizes non-invasive battery available capacity estimation, avoiding damage to the battery. The ultrasonic signal directly reflects the electrode deformation and electrolyte state, overcoming the defect that the electrical signal is easily disturbed by the environment. The multi-feature fusion and hybrid neural network model effectively capture the multi-factor coupling relationship of capacity attenuation, and the pidgeon algorithm optimization improves the model prediction accuracy. The method does not require complex detection equipment, and can obtain data in real time under the normal working state of the battery, and is suitable for dynamic application scenarios such as electric vehicles.

[0049] The CNN-SE-BILSTM network used in the application is specifically:

[0050]

[0051] In the formula: represents the weight matrix of the 1D CNN; represents the feature vector after the 1D CNN; represents the activation function; represents the input feature vector; The SE attention mechanism is:

[0052] In the formula: represents the input vector of the first layer network; represents the feature vector after the activation function; represents the input vector of the first layer network; The output of the BILSTM is determined by the forward and backward LSTMs:

[0053] Wherein: : the hidden state of the forward LSTM at time step t ; : the hidden state of the backward LSTM at time step t; : the final output of the BiLSTM (usually the concatenation of the two hidden states); The principle of the SE attention mechanism is as follows: Transformation: the feature map X becomes the feature map U. F tr tr ​It can be seen as a standard convolution operator.

[0054] Compression: compress the global spatial information of each channel into a channel descriptor, and perform global average pooling. The feature map containing global information WxHxC is directly compressed into a 1x1xC feature vector Z, so that the channel features of C feature maps are compressed into a numerical value, which makes the generated channel-level statistical data Z contain context information, and alleviates the problem of channel dependence. The formula is as follows:

[0055] u c is the c-th channel of the input feature map. Z c is the compressed channel descriptor.

[0056] Again, the incentive: learn the dependence between channels and generate channel weights. Usually take two fully connected layers, the first fully connected layer compresses C channels into C / r channels to reduce the amount of calculation, and then passes through a RELU nonlinear activation layer, the second fully connected layer restores the channel number to C channels, and then obtains the weight s through the Sigmoid activation, and finally the dimension of this s is 1x1xC, and the formula is as follows:

[0057] The attention weight obtained in the foregoing is weighted to the feature of each channel, and through the SE module, the network can adaptively enhance the feature response of key channels and suppress redundant information, thereby significantly improving the model performance, especially the robustness performance in complex scenes: .

[0058] The sparrow population can be divided into discoverers, followers and sentinels in terms of categories. By calculating the fitness value, the position of individuals in the population is constantly updated to obtain the optimal solution of the problem. The update method of the discoverer is shown in equation 5.1.

[0059]

[0060] In the formula, is the c-th channel of the input feature map. Z t is the c-th channel of the input feature map. Z i is the c-th channel of the input feature map. Z j is the c-th channel of the input feature map. Z T is the c-th channel of the input feature map. Z LIt is an identity matrix with elements of 1, ST is the set safety threshold, and R2 represents the alarm value, taking values ​​in [0,1]. When R2 is less than ST, it indicates a safe environment, and the discoverer conducts a broad search to obtain a better fitness. Other situations indicate a dangerous environment, and the discoverer moves to a safe zone to avoid a decrease in fitness. Followers in the population follow the discoverer, search in a local area, and update their positions according to Equation 5.2.

[0061] .

[0062] In the formula, This represents the worst position in the t-th iteration. Indicates the first t The optimal position in the next iteration. N It refers to the sparrow population size. A It is a matrix with elements of ±1. When i Greater than N When / 2, it means the follower is not following the discoverer and will trigger autonomous search behavior, expanding the search area. Conversely, when i Less than or equal to N At / 2, followers obtain food by following the discoverer and conduct precise searches within the high-quality solution area. The watchdogs in the population are randomly selected from other sparrows, and their positions are updated according to Equation 5.3.

[0063]

[0064] In the formula, β For random numbers that follow a normal distribution, K The value range is [-1, 1], which controls the individual's movement direction. ε It is a very small constant. It is the current individual fitness level. This is the optimal fitness. When the two differ, meaning the current individual is not globally optimal, a refined search is performed within the neighborhood of the optimal solution. Otherwise, a controlled perturbation is applied to explore unknown regions.

[0065] In a specific embodiment of this application, the sparrow optimization algorithm is used to optimize the model parameters of the capacity estimation model, specifically including the following steps: m, define the hyperparameter space, and determine the hyperparameters to be optimized and their range; n represents the initial sparrow population, where the position of each sparrow represents a set of hyperparameters. o. Establish a fitness function and minimize the error; p, using the current sparrow training model, outputs the fitness; q, Assign roles and positions based on fitness values, and update hyperparameter combinations; r, if the maximum number of iterations is not reached or the convergence condition is met, return to d, otherwise select the optimal hyperparameter, output the hyperparameter combination with the optimal fitness value; s, training the model according to the optimized hyperparameters, and estimating the battery available capacity.

[0066] To demonstrate the feasibility and accuracy of the ultrasonic signal-based lithium-ion battery available capacity estimation method based on ultrasonic detection and SAA-CNN-SE-BILSTM model of the present application, the following examples are used to illustrate, and the MAE, R 2 and RMSE of various algorithms are compared to illustrate the accuracy of the present application.

[0067] Table 1

[0068] The present application uses ultrasonic experimental data from the laboratory of the research group, and the correlation between the characteristic parameters and the battery capacity is shown in Table 1. As can be seen from the correlation analysis, the Pearson coefficient of the initial flight time and the initial ultrasonic amplitude is the highest, while the correlation between the maximum signal amplitude and the minimum signal amplitude is lower, but also greater than 0.8, indicating that the correlation is low. Therefore, the initial signal amplitude and flight time data in the ultrasonic data are used to estimate the capacity of the lithium-ion battery. The first 70% of the data is used as the training set, and the last 30% of the data is used as the test set. At the same time, the interpolation method is used to increase the data volume without changing the trend. The training set data is used to train the model, and the test set is used to verify the estimation effect of the capacity, and the accuracy of the designed model is evaluated.

[0069] BILSTM, CNN-BILSTM, CNN-SE-BILSTM model and SSA-CNN-SE-BILSTM model are selected for comparison. The battery available capacity estimation results are shown in 2a as the battery available capacity estimation results, Figure 5b The absolute error box plot of the battery available capacity estimation results is shown in 2b. It can be seen that compared with BILSTM, CNN-BILSTM and CNN-SE-BILSTM model, the estimation performance of SAA-CNN-SE-BILSTM model proposed in the present application is better, the estimation curve fluctuates less, and the fitting with the actual SOH curve is closer. The trend of capacity change is captured through the ultrasonic signal. In contrast, the estimation curves of BILSTM, CNN-BILSTM and CNN-SE-BILSTM model become worse with the change of capacity.

[0070] As shown in Table 2, the RMSE of BILSTM model is 0.0014559, the MAE is 0.0011351, the R 20.8724, the RMSE of the CNN-BILSTM model is 0.0010599, the MAE is 0.0007706, and the R 2 0.9324, the RMSE of the CNN-SE-BILSTM model is 0.0007919, the MAE is 0.0007296, and the R 2 0.9623, the RMSE of the SSA-CNN-SE-BILSTM model is 0.00038531, the MAE is 0.0003286, and the R 2 0.9911, compared with the BILSTM model, the RMSE is reduced by 73.53%, the MAE is reduced by 71.05%, and the R 2 is increased by 13.61%.

[0071] Table 2 Error indicators of model comparison experiments

[0072] The application can accurately identify the ultrasonic signal features closely related to the capacity attenuation of the battery, improve the input quality of the model by constructing a high-correlation multi-feature data set, reduce noise interference, thereby improving the prediction accuracy and calculation efficiency of the capacity estimation model, and enhancing the generalization ability of the model under different battery aging states. The application can effectively solve the prediction deviation problem caused by improper parameter setting of the traditional capacity estimation model, and improve the generalization ability of the model under different battery aging states, environmental temperature fluctuations and other complex working conditions through adaptive parameter optimization. For example, in the capacity attenuation acceleration stage of the later stage of battery charging and discharging, the optimized model can more accurately capture the nonlinear relationship between the ultrasonic signal features and the capacity, reduce the cumulative error caused by fixed parameters, and thereby improve the stability and reliability of the capacity estimation result.

Claims

1. A method for estimating the available capacity of a lithium-ion battery based on ultrasonic signals, characterized in that, The method comprises the following steps: S1, collecting current, capacity, ultrasonic feedback signal and corresponding sampling time data of the lithium ion battery in the constant current charging and discharging stage of the charging and discharging cycle; S2, processing the ultrasonic feedback signal based on wavelet packet transform to obtain the amplitude of the ultrasonic feedback signal, and obtaining the time of flight of the ultrasonic signal from the ultrasonic feedback signal based on the cross-correlation algorithm; S3, calculating the correlation degree of the amplitude of the ultrasonic signal, the time of flight information of the ultrasonic signal and the capacity information based on the Pearson correlation algorithm, and grouping the amplitude of the ultrasonic signal and the time of flight information of the ultrasonic signal to generate a multi-feature data set; S4, establishing a capacity estimation model based on CNN-SE-BILSTM based on the multi-feature data set, optimizing the model parameters of the capacity estimation model by using the sparrow optimization algorithm, and estimating the available capacity of the battery by using the capacity estimation model after the model parameter optimization.

2. The method of claim 1, wherein, The current of the lithium-ion battery during the constant-current charging and discharging stage of the charging and discharging cycle is specifically: the current data of the constant-current charging and discharging stage of the kth cycle is a sequence wherein and are the initial current and the terminal current of the constant-current charging stage, respectively. The capacity of the lithium ion battery collected in the constant current charging and discharging stage of the charge and discharge cycle is specifically: the charging capacity data of the constant current charging stage is a sequence , wherein is the constant current charging capacity of the kth cycle battery, and is the current and sampling time of the kth cycle, respectively; The ultrasonic feedback signal collected from the lithium ion battery in the constant current charging and discharging stage of the charging and discharging cycle is specifically: the constant current charging and discharging ultrasonic data of the kth cycle is a sequence wherein and are respectively the initial charging ultrasonic signal and the terminal discharging ultrasonic signal in the constant current charging stage. The sampling time data of the lithium ion battery in the constant current charging and discharging phase of the charging and discharging cycle is specifically: the constant current charging and discharging sampling time data of the kth cycle is a sequence , wherein , and are the initial sampling time and the terminal sampling time of the constant current charging phase, respectively. 3.The method of claim 1, wherein, The amplitude of the ultrasonic feedback signal is obtained by processing the ultrasonic feedback signal based on wavelet packet transform, and specifically includes: The low-frequency and high-frequency components of the ultrasonic feedback signal are recursively decomposed based on wavelet packet transform to obtain a multi-resolution frequency representation of the ultrasonic feedback signal, i.e. the amplitude of the ultrasonic feedback signal.

4. The method of claim 3, wherein the method is based on ultrasonic signals. The mother wavelet function of the wavelet packet transform is expressed as follows: wherein is an energy normalization factor; a is a scale parameter, b is a translation parameter; for the ultrasound feedback signal x(n) The wavelet packet transform formula is as follows: wherein is the result of the wavelet transform, is the raw data of the ultrasound feedback signal, using its discrete form, whose low frequency components are shown below, with the sub-nodes of the low frequency components being 2n : The high frequency component is shown in the following formula, and the child node of the high frequency component is 2n+1 : wherein j is the number of wavelet packet decomposition levels, n is the node position, k and m is for discrete time or position, x j,n [k] is the discrete signal sequence of the j layer, the n subband; h[m] and g[m] are low pass and high pass filter coefficients.

5. The method of claim 1, wherein, The time of flight of the ultrasonic signal is obtained from the ultrasonic feedback signal based on the cross-correlation algorithm, which specifically includes comparing one signal with another signal through a series of different time delays to find the most similar time of the two signals, and extracting the time difference at this time as the time of flight of the ultrasonic signal.

6. The method of claim 5, wherein the method further comprises: The time of flight of the ultrasonic signal is obtained from the ultrasonic feedback signal based on the cross-correlation algorithm, which is specifically shown in the following formula: where the transmitted signal is and the received signal is where is a time delay representing the relative displacement between the transmitted and received signals, and when is at a maximum, the correlation between the two signals is at a maximum, and when is the time of flight of the ultrasound, which is expressed discretely as follows: wherein the signal X at discrete time points n , the values the signal Y at discrete time points n+k , the values k is a lag quantity, the signal Y is time offset with respect to the signal X .

7. The method of claim 1, wherein, The Pearson correlation coefficient is used to measure the linear correlation degree between two variables, and the value range is [-1, 1], which is used to judge the linear relationship strength and direction between two variables, and the calculation formula is as follows: wherein X i , Y i is the value of the ith sample point, , is the sample mean, X , Y is the correlation between the two variables.

8. The method of claim 1, wherein, The sparrow optimization algorithm is used to optimize the model parameters of the capacity estimation model, which specifically includes the following steps: a, set the hyperparameter space and determine the optimized hyperparameters and their ranges; b, initialize the sparrow population, and the position of each sparrow represents a set of hyperparameters; c, establish a fitness function to minimize the error; d, train the model using the current sparrow and output the fitness; e, assign roles and positions according to the fitness value, and update the hyperparameter combination; f, if the maximum number of iterations is not reached or the convergence condition is met, return to d, otherwise select the optimal hyperparameters and output the hyperparameter combination with the optimal fitness value; Train the model according to the optimized hyperparameters and estimate the available capacity of the battery.

9. A system for estimating the available capacity of a lithium-ion battery based on ultrasonic signals, the system comprising: It comprises a data acquisition module, a data preprocessing module, a feature calculation module and a capacity estimation module: The data acquisition module collects the current, capacity, ultrasonic feedback signal and corresponding sampling time data of the lithium ion battery in the constant current charging and discharging stage of the charging and discharging cycle; The data preprocessing module processes the ultrasonic feedback signal based on wavelet packet transform to obtain the amplitude of the ultrasonic feedback signal, and obtains the time of flight of the ultrasonic signal from the ultrasonic feedback signal based on the cross-correlation algorithm; The feature calculation module calculates the correlation degrees of the amplitude of the ultrasonic signal, the time-of-flight information of the ultrasonic signal and the capacity information respectively based on a Pearson correlation algorithm, and groups the amplitude of the ultrasonic signal and the time-of-flight information of the ultrasonic signal to generate a multi-feature data set; The capacity estimation module establishes a capacity estimation model based on CNN-SE-BILSTM based on the multi-feature data set, optimizes the model parameters of the capacity estimation model by using a sparrow optimization algorithm, and estimates the available capacity of the battery by using the capacity estimation model after the model parameter optimization.

10. The system for estimating the available capacity of a lithium-ion battery based on ultrasonic signals according to claim 9, wherein, The amplitude of the ultrasonic feedback signal obtained by processing the ultrasonic feedback signal based on the wavelet packet transform specifically includes: The amplitude of the ultrasonic feedback signal is obtained by recursively decomposing the low-frequency and high-frequency components of the ultrasonic feedback signal based on the wavelet packet transform to obtain a multi-resolution frequency representation of the ultrasonic feedback signal.

Images