Quasi-solid-state battery SOH-SOC joint estimation method oriented to energy storage scene

Abstract

The invention discloses a quasi-solid-state battery SOH-SOC joint estimation method for an energy storage scene, and the method comprises the steps: obtaining a quasi-solid-state battery data set which comprises an all-voltage data set and a specific voltage data set; an SOH-SOC joint estimation model is established; the specific voltage data set is sent to a charging encoder, and charging encoding data is obtained; the charging coded data are sent to an SOH encoder, and specific SOH decoded data are obtained; sending the specific SOH decoding data and all the voltage data sets into an SOH encoder to obtain all the SOH encoding data; all the SOH coded data and the charging coded data are sent into an SOC decoder, and SOC decoded data are obtained; taking the specific SOH decoding data and the SOC decoding data as estimated values of the battery health state and the battery charge state respectively; the bidirectional jump TCN-BiGRU is utilized to effectively reduce data redundancy input, and the purpose of solving the joint estimation problem of quasi-solid-state battery data SOH-SOC by using one model is achieved.

Description

Technical Field

[0001] The present invention relates to the field of joint estimation of the state of health and state of charge of a quasi-solid-state battery, and in particular to a quasi-solid-state battery SOH-SOC joint estimation method for energy storage scenarios. Background Art

[0002] With the rapid development of new energy vehicles and energy storage systems, quasi-solid-state batteries have become an important development direction for the next generation of power batteries due to their high energy density, excellent safety performance and low risk of electrolyte leakage. Accurately estimating the state of health (SOH) and state of charge (SOC) of quasi-solid-state batteries is the core challenge of the new generation of technology.

[0003] As key indicators for characterizing the battery state, SOH and SOC directly affect the battery charge and discharge control, life prediction and system safety. Traditional SOC estimation methods rely on the linear assumption of the battery model, but the solid-liquid mixed electrolyte characteristics of quasi-solid-state batteries lead to more significant polarization effects and interface impedance changes, and traditional models are difficult to accurately capture the nonlinear response under dynamic conditions. In addition, existing SOH estimation is mostly based on single-factor analysis of capacity or internal resistance, while the aging mechanism of quasi-solid-state batteries involves multiple coupling factors such as electrode-electrolyte interface degradation and solid electrolyte cracking. It is difficult for a single parameter model to fully reflect the capacity decay law. Moreover, the SOH estimation of quasi-solid-state batteries during constant current and constant voltage charging depends only on specific charging data within a predetermined range, while SOC estimation needs to be evaluated throughout the charging process. Previous studies used the estimation results of the SOH model as the input of the SOC model and used two independent models for estimation. However, these methods ignore the potential inherent correlation in the SOH and SOC charging data, resulting in increased computational complexity and reduced estimation accuracy. In addition, current convolutional models and deep network models ignore the intrinsic connection between local and global features, hindering the ability to capture time series aging information. Therefore, there is an urgent need for a SOH-SOC joint estimation method that takes into account both model accuracy and computational efficiency and adapts to the characteristics of quasi-solid-state batteries in order to break through the bottleneck of existing technology. Summary of the invention

[0004] In response to the problems in the prior art, the present invention provides a quasi-solid-state battery SOH-SOC joint estimation method for energy storage scenarios, aiming to reduce redundant data input and improve the estimation efficiency of battery health status and battery state of charge.

[0005] The SOH-SOC joint estimation method of quasi-solid-state batteries for energy storage scenarios includes the following steps:

[0006] Step 1: Obtain a quasi-solid-state battery dataset, which includes a full voltage dataset and a specific voltage dataset;

[0007] Step 2: Establish a SOH-SOC joint estimation model, which includes a charging encoder, a SOH encoder, a SOH decoder, and a SOC decoder;

[0008] Step 3: Send the specific voltage data set to the charging encoder and obtain charging encoding data;

[0009] Step 4: Send the charge encoding data to the SOH encoder and obtain specific SOH decoding data;

[0010] Step 5: Send the specific SOH decoded data and the entire voltage data set to the SOH encoder to obtain the entire SOH encoded data;

[0011] Step 6: Send all SOH encoding data and charging encoding data to the SOC decoder to obtain SOC decoding data;

[0012] Step 7: Using the specific SOH decoded data and the SOC decoded data as estimated values ​​of the battery state of health and the battery state of charge, respectively.

[0013] Further: the quasi-solid-state battery dataset is a self-built dataset and includes the capacity and charge changes of the quasi-solid-state battery during the charging process of the entire life cycle; the dataset is the data collected by the ArbinBT2000 battery experimental system in a continuous charging experiment, and all experimental quasi-solid-state batteries follow the same charging curve; the experiment was carried out at a controlled ambient temperature of 25°C, and the battery was charged using a constant current and constant voltage charging protocol; during the constant current charging process, data was collected every 60s; four batteries were selected from the self-built quasi-solid-state battery dataset, named L11, L12, L13 and L14, all of which use lithium iron phosphate as the positive electrode material, with a nominal capacity of 2AH and a nominal voltage of 4.2V; the dataset contains comprehensive information about the aging process, including voltage, current, and charging time; L11, L12 and L13 were selected as training datasets, while L14 was selected as a test dataset. In a complete constant current and constant voltage charging process, the initial charging current is 0.25C (0.5A) and 2C (4A), respectively, and then it transitions to a constant voltage of 4.2V to charge to the cut-off current.

[0014] Furthermore, the specific voltage data set includes data on voltage changes within a voltage range of [3.5, 4.2V] during the constant current charging stage.

[0015] Further, the charging encoder includes a one-dimensional convolution module (1D-CNN), a bidirectional jump TCN-BiGRU network module, a fully connected layer and a dense layer. Step 3 specifically includes the following steps:

[0016] Step 3.1: The specific voltage dataset is converted by the 1D-CNN layer to obtain local feature data;

[0017] Step 3.2: The local feature data is passed through the bidirectional jump TCN-BiGRU network module to obtain the model data;

[0018] Step 3.3: The model data is passed through the fully connected layer and the dense layer in sequence to obtain the charge encoding data.

[0019] Further: the bidirectional jump TCN-BiGRU network module includes a forward temporal convolutional network, a reverse temporal convolutional network, a fully connected layer and a bidirectional gated recurrent unit. The forward temporal convolutional network uses an expanded causal convolution to process local feature data in chronological order to ensure that the temporal dependency conforms to the causal relationship; the reverse temporal convolutional network flips the local feature data along the time axis and then performs an expanded causal convolution to capture reverse time information; one output branch of the forward temporal convolutional network and the reverse temporal convolutional network are both sent to the bidirectional gated recurrent unit through the fully connected layer, and the other output branch of the two is sent to the bidirectional gated recurrent unit through two jump connections respectively, and when the jump connection conditions are met, the outputs of the forward temporal convolutional network and the reverse temporal convolutional network are directly sent to the bidirectional gated recurrent unit through the jump connection.

[0020] Further, the skip connection condition is that the output of the forward temporal convolutional network or the output of the reverse temporal convolutional network is inconsistent with the input dimension of the bidirectional gated recurrent unit, and the input dimension is the number of channels or the time step.

[0021] Further, the skip connection condition is that the input value of a specific voltage data set is very large or very small, and the depth of the bidirectional gated recurrent unit exceeds 10 layers.

[0022] Further: the jump connection condition is: the weight operation in the forward temporal convolutional network and the reverse temporal convolutional network determines whether the specific voltage data set input is a high-value input, wherein when the weight of the specific voltage data set input is greater than 0.5, the specific voltage data set input is recorded as a high-value input.

[0023] Further: the SOH decoder and the SOC decoder both include dense layers. The dense layers of the SOH decoder layer are used to extract long-term degradation features related to battery aging, capture low-frequency, slowly changing global trends, and obtain SOH decoded data after nonlinear transformation of charging encoded data; the dense layers of the SOC decoder are used to extract dynamic instantaneous features of the battery state, process high-frequency, rapidly changing local signals, and obtain SOC decoded data after nonlinear transformation of the input.

[0024] Furthermore, the SOH encoder includes a Gaussian noise layer, a filling layer and a dense layer, wherein the dense layer is used to process fine-grained feature relationships, and the SOH encoded data is nonlinearly transformed and then input as part of the SOC decoder.

[0025] The beneficial effects of the present invention are as follows: bidirectional jump TCN-BiGRU is used to effectively reduce redundant data input, so that one model can be used to solve the SOH-SOC joint estimation problem of quasi-solid-state battery data; the SOH-SOC joint estimation model has high accuracy, high stability and generalization ability, and its parallel processing speed is fast and the computational efficiency is increased. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the present invention;

[0027] Figure 2 It is a structural block diagram of the SOH-SOC joint estimation model in the present invention;

[0028] Figure 3 It is a structural block diagram of the bidirectional jump TCN-BiGRU network module in the present invention. DETAILED DESCRIPTION

[0029] The present invention is described in detail below in conjunction with the accompanying drawings. The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention. The directional terms such as left, middle, right, top, and bottom in the embodiments of the present invention are only relative concepts or are based on the normal use state of the product, and should not be considered as restrictive.

[0030] The SOH-SOC joint estimation method of quasi-solid-state batteries for energy storage scenarios, such as Figure 1 As shown, the following steps are included:

[0031] Step 1: Obtain a quasi-solid-state battery dataset, which includes a full voltage dataset and a specific voltage dataset;

[0032] Specifically, the self-built quasi-solid-state battery data set includes a total voltage data set and a specific voltage data set. The self-built data set of the present invention includes the capacity and charge changes during the charging process of the entire life cycle. The data set used in the present invention is the data collected by the ArbinBT2000 battery experimental system during the continuous charging experiment.

[0033] The experiment obtained two quasi-solid-state battery data sets under two charging conditions. The experiment was carried out at a controlled ambient temperature of 25°C, and the battery was charged using a constant current and constant voltage charging protocol. During the constant current charging process, data was collected every 60s. Four batteries were selected from the first condition of the self-built quasi-solid-state battery data set, named L11, L12, L13, and L14. Eight batteries were selected from the second condition, named PH01-PH08. The nominal capacity of the batteries in both conditions is 2AH, and the nominal voltage is 3.6V. The data set contains comprehensive information about the aging process, including voltage, current, charging time, and other charging time and other voltage-related parameters, which provides a reference for the attenuation of the battery throughout its life. For subsequent experiments, L11, L12, and L13 were selected as training data sets, while L14 was selected as the test data set, PH01-PH06 as the training data set, and PH07 and PH08 as the test data set to evaluate the performance of the training results. Analysis of the attenuation curve of the quasi-solid-state battery can analyze three different characteristics. First, the capacity degradation process shows significant differences between different batteries. Second, the trend of capacity degradation is nonlinear, and the degradation rate increases as the life cycle progresses. Third, the capacity degradation of the battery follows an oscillating decline pattern rather than a monotonic decrease. In this invention, the state of health (State of Health, SOH) is defined as the ratio of the current capacity of the battery to the initial capacity, and the state of charge (State of Charge, SOC) is positioned as the ratio of the current charge capacity of the battery to the fully charged charge capacity. The definitions of SOH and SOC are shown in Formula 1 and Formula 2.

[0034] (1) (2)

[0035] in, Indicates the current maximum remaining discharge capacity of the battery. Indicates the initial capacity of the battery. When the health status drops below 80%, it means that the battery is close to the upper limit of its life and should be replaced in time. Indicates the current charge capacity of the battery. Indicates fully charged charge capacity.

[0036] Subsequent data analysis focuses on the life within a specified time range. The first condition constant current constant voltage charging process has an initial charging current of 0.25C (0.5A) and transitions to a constant voltage charge of 4.2V to the cut-off current. The second condition constant current constant voltage charging process has an initial charging current of 2C (4A) and transitions to a constant voltage charge of 4.2V to the cut-off current. The duration of the constant current constant voltage charging cycle of the battery data set to reach the rated voltage decreases as the battery life cycle progresses during the constant current charging process, thus supporting the relationship between the remaining capacity and the constant current charging. The voltage range of the constant current charging mainly spans [3.5V, 4.2V], the voltage initially rises rapidly, then transitions to a gradual increase, and the rate of rise becomes more and more obvious. Previous studies have used parameter changes during the charging process to estimate SOH. In this invention, the data of voltage changes within a specific range of the constant current charging stage are used for SOH estimation, and the voltage range of [3.5V, 4.2V] with a larger change program is selected as the specific voltage data set. Many studies have analyzed and verified that the voltage, current and time changes of battery data are the most suitable input parameters for estimating SOC and SOH. Therefore, these aspects will not be further investigated in this paper. In this dataset, the data collection interval during CC charging is fixed. Here, the time parameter has limited impact on the model, so we choose to focus only on voltage and current as input parameters of the model.

[0037] Step 2: Establish a SOH-SOC joint estimation model, such as Figure 2 As shown, the SOH-SOC joint estimation model includes a charging encoder, a SOH encoder, a SOH decoder, and a SOC decoder;

[0038] In order to better estimate the SOC throughout the life cycle of the battery, the current SOH of the battery must be obtained. Moreover, accurate estimation of the SOH requires data collected within a specific voltage range during the charging process. Therefore, the present invention creates a SOH-SOC joint estimation model based on a charging encoder. The model uses multiple data points of a specific voltage data set in each charging cycle to jointly estimate SOC and SOH, and obtains a SOH decoder capable of estimating SOH. As the battery life decays, multiple observed variables appear at each time point. By arranging these variables in chronological order, a series of current-voltage features are formed, so that the current-voltage features are used as input to the charging encoder.

[0039] Step 3: Send the specific voltage data set to the charge encoder and obtain the charge encoding data; wherein the charge encoder includes a one-dimensional convolution module (1D-CNN), a bidirectional jump TCN-BiGRU network module, a fully connected layer and a dense layer, and specifically includes the following steps:

[0040] Step 3.1: The specific voltage dataset is converted by the 1D-CNN layer to obtain local feature data;

[0041] The first layer of the charging encoder consists of 1D-CNN, which has powerful feature extraction capabilities and is particularly suitable for processing one-dimensional sequence data. In the joint estimation of SOH and SOC, the battery parameters such as current and voltage usually exist in the form of a one-dimensional sequence. 1D-CNN can effectively extract local features from these parameters by sliding windows in the convolutional layer, thereby capturing subtle information about battery state changes. It has two main purposes: dimensionality conversion and local feature extraction. It is used to capture important features and local patterns of current and voltage features.

[0042] (3)

[0043] In the above equation, Indicates that at time step The output, represents the time step The input on represents the weight, and Represents the kernel size. One-dimensional convolution enables parameter sharing and efficient feature extraction from sequential data, which makes it particularly suitable for tasks such as time series analysis and natural language processing;

[0044] Step 3.2: The local feature data is passed through the bidirectional jump TCN-BiGRU network module to obtain the model data; Figure 3 As shown in the figure, the bidirectional jump TCN-BiGRU network module includes a forward temporal convolutional network, a reverse temporal convolutional network, a fully connected layer and a bidirectional gated recurrent unit. The forward temporal convolutional network uses dilated causal convolution to process local feature data in chronological order to ensure that the temporal dependency conforms to the causal relationship; the reverse temporal convolutional network flips the local feature data along the time axis and then performs dilated causal convolution to capture reverse time information; one output branch of the forward temporal convolutional network and the reverse temporal convolutional network are both sent to the bidirectional gated recurrent unit through the fully connected layer, and the other output branches of the two are respectively sent to the bidirectional gated recurrent unit through two jump connections, and when the jump connection conditions are met, the outputs of the forward temporal convolutional network and the reverse temporal convolutional network are directly sent to the bidirectional gated recurrent unit through the jump connection; the conditions for enabling jump connections are as follows:

[0045] (1) The output of the forward temporal convolutional network or the output of the reverse temporal convolutional network is inconsistent with the input dimension of the bidirectional gated recurrent unit. The input dimension is the number of channels or the time step. Jump connections are enabled to perform dimension transformation to unify the input.

[0046] (2) When the input value of a specific voltage data set is very large or very small, the gradient will approach 0, causing the gradient to vanish. When the depth of the bidirectional gated recurrent unit exceeds 10 layers, it may cause a gradient explosion. Enabling skip connections to enhance the gradient back propagation capability solves the gradient vanishing and explosion problems.

[0047] (3) The weight operation in the forward temporal convolutional network and the reverse temporal convolutional network determines whether a specific voltage dataset input is a high-value input. When the weight of a specific voltage dataset input is greater than 0.5, the specific voltage dataset input is recorded as a high-value input. When it is a high-value input, the skip connection is enabled to avoid feature loss.

[0048] Although the forward temporal convolutional network and the reverse temporal convolutional network can capture recent trends and short-term dependencies, they are limited by the fixed length of historical information and cannot flexibly adapt to battery data of different lengths, which limits the network in dealing with historical dependency problems. Secondly, in the deep structure, the distance of information transmission will be limited, making it difficult to capture the correlation between distant time steps. In order to solve the problem of historical dependency of battery data and the problem of information transmission loss, the present invention adopts a bidirectional gated recurrent unit to deal with this problem. The bidirectional gated recurrent unit can simultaneously obtain the forward and reverse information of the input data through the forward and backward recurrent neural network structure, so as to better capture the long-term dependencies in the data. During the degradation process of the battery, the changes in SOH and SOC are often affected by historical data. This ability of the bidirectional gated recurrent unit enables it to more accurately predict the future state of the battery. The bidirectional gated recurrent unit contains an update gate and a reset gate to control the flow of information. The function of the update gate is to determine how much information of the previous time step needs to be retained for the hidden state of the current time step. The output value of the update gate is between 0 and 1. The larger the value, the more past information is retained, and the smaller the value, the more dependent on the current input information. In addition, the receptive field determines the range of dependencies that can be captured by the forward temporal convolutional network and the reverse temporal convolutional network. By utilizing dilated convolutions, the forward temporal convolutional network and the reverse temporal convolutional network can effectively model distant dependencies without significantly increasing the number of parameters. The key advantage of the forward temporal convolutional network and the reverse temporal convolutional network is its ability to capture long-term dependencies while maintaining computational efficiency. The forward temporal convolutional network and the reverse temporal convolutional network provide a powerful tool for analyzing sequential data. 1D-CNN allows local feature extraction, while the forward temporal convolutional network and the reverse temporal convolutional network extend this concept to capture long-term dependencies. .

[0049] Step 3.3: The model data is sequentially passed through the fully connected layer and the dense layer to obtain the charging encoding data. The dense layer contains the weight matrix, the bias vector and the activation function. The model data is nonlinearly transformed through the ReLU activation function to extract high-order features, and the output dimension of the previous module is matched with the input dimension of the next module, which can effectively fit the complex dynamic relationship in the battery aging process. It is responsible for accepting more abstract global features in the model data and obtaining the charging encoding data after nonlinear transformation of the model data. The fully connected layer can receive the model data processed by the previous bidirectional jump TCN-BiGRU network module and integrate these model data. In the SOH-SOC joint estimation task, various state information of the battery may be scattered in different model data. The fully connected layer can integrate these scattered features together to form a comprehensive feature representation, which is convenient for subsequent estimation tasks. Moreover, in the battery SOH-SOC joint estimation model, overfitting is a common problem. Overfitting can be further reduced by adding a fully connected layer after the bidirectional jump TCN-BiGRU network module.

[0050] Step 4: Send the charging coded data to the SOH encoder and obtain specific SOH decoded data; the SOH decoder includes a dense layer, and the dense layer of the SOH decoding layer is used to extract the long-term degradation characteristics related to battery aging, capture the low-frequency and slowly changing global trend, and obtain the SOH decoded data after performing a nonlinear transformation on the charging coded data;

[0051] Step 5: Send the specific SOH decoded data and the entire voltage data set to the SOH encoder to obtain the entire SOH encoded data; the SOH encoder includes a Gaussian noise layer, a filling layer, and a dense layer. The dense layer is used to process fine-grained feature relationships and perform nonlinear transformation on the SOH encoded data as part of the SOC decoder input. In the SOH-SOC joint estimation task, the various state information of the battery can be efficiently integrated and mapped through the dense layer and converted into SOH decoded data, providing strong support for subsequent estimation tasks.

[0052] Step 6: Send all SOH encoded data and charging encoded data to the SOC decoder to obtain SOC decoded data; the SOC decoder includes a dense layer, and the dense layer of the SOC decoder is used to extract the dynamic instantaneous characteristics of the battery state, process high-frequency and rapidly changing local signals, and obtain the SOC decoded data after performing nonlinear transformation on the input;

[0053] In the previous steps, a pre-trained charge encoder was used to encode the local feature data of the charging process, and a dedicated SOH encoder specifically for encoding SOH was combined. Subsequently, all SOH encoded data and charge encoded data were simultaneously input to the SOC decoder, thereby achieving accurate decoding and estimation of SOC. It is worth noting that the SOH decoded data in the actual estimation represents an estimated value, which may have some inherent errors, which is different from the measured SOH value used in the training process. In order to alleviate overfitting, Gaussian noise is introduced to the SOH decoded data during the training process. According to the actual usage requirements, it is necessary to estimate the SOC throughout the charging process of the entire life cycle. However, the initial stage of constant current charging cannot provide a long enough sequence to be effectively input into the model. In order to overcome this limitation. A padding operation is introduced at the beginning of a specific SOH decoded data, that is, padding the entire voltage data set, which enables the model to accurately estimate the SOC throughout the charging process. In addition, in the early stages of constant current charging, the SOH decoded data is unclear because the model has no estimate yet. Therefore, padding is also applicable to the early stages of SOH input. It is worth emphasizing that the SOH input in the early stage of charging has little effect on the accuracy of SOC estimation. Therefore, padding the entire voltage dataset at this stage does not significantly affect the accuracy of SOC estimation. The charge sequence and current SOH of multiple features are encoded using a charge encoder and a SOH encoder, respectively. Subsequently, the SOC and SOH are estimated using a SOC decoder and a SOH decoder, respectively. In the early stages of each charging cycle, when the length of the entire voltage dataset is less than the input sequence length of the model, a padding operation is applied to the sequence. In addition, it is not feasible to accurately estimate SOH when no specific voltage fluctuations are observed. In this case, only SOC is estimated, and the entire voltage dataset is padded. As charging progresses and the voltage enters a specific range, the model can estimate SOC and SOH simultaneously. In order to better represent the estimated value of SOH, the final result of SOH is the average of the SOH estimates of the known sequence.

[0054] Step 7: Using the specific SOH decoded data and the SOC decoded data as estimated values ​​of the battery state of health and the battery state of charge, respectively.

[0055] In this study, we used multiple evaluation metrics to comprehensively evaluate the performance of the proposed encoder-decoder. These metrics help to objectively measure the estimation ability of the model and reveal its accuracy and robustness from different perspectives. The following are the five evaluation metrics we used:

[0056] 1. Mean Squared Error (MSE): Mean squared error is a commonly used metric to evaluate the difference between the model’s estimated results and the true values; it calculates the average of the squared differences between the estimated values ​​and the true values ​​to measure the average deviation of the estimate; (4)

[0057] Where n represents the number of samples, t represents the index of the sample, represents the estimated value of sample t, Represents the true target value of sample t.

[0058] 2. Root Mean Square Error (RMSE): The root mean square error is the square root of the mean square error, which represents the average difference between the estimated value and the true value. RMSE is more sensitive to outliers and can be used to measure the accuracy of the model. (5)

[0059] 3. Mean absolute error (MAE): The mean absolute error is the average of the absolute differences between the estimated value and the true value, which measures the average error of the estimate. Unlike MSE, MAE does not amplify the impact of large errors, so it better reflects the overall accuracy of the estimate. (6)

[0060] 4. Mean Percent Absolute Error (MAPE): The mean percentage absolute error is the average of the relative differences between the estimated value and the true value, expressed as a percentage; it measures the relative error of the model in different data ranges and can reflect the relative accuracy of the estimate; (7)

[0061] 5. Maximum Absolute Error (MAXE): The maximum absolute error is the maximum absolute difference between the estimated value and the true value. It identifies the estimation error of the model in the worst case. MAXE is particularly sensitive to outliers and helps to understand the maximum risk of the model in estimation. (8)

[0062] in, Indicates the maximum value operation.

[0063] In this paper, quasi-solid-state batteries L11, L12 and L13 are selected as training data sets, L14 is selected as test data sets, PH01-PH06 are selected as training data sets, and PH07 and PH08 are selected as test data sets. The mean absolute error (MAE) is used as the optimization target to balance outlier sensitivity and convergence stability. When the MAE of the validation set decreases by <0.1% for 10 consecutive epochs, the early stopping mechanism is triggered, and the minimum learning rate is set to 1% of the initial value. Parameter sensitivity experiments show that when Batch size=64, the gradient update direction stability is the best, and the MAE variance is reduced by 28.7% and 45.2% compared with Batch size of 32 and 16 respectively; the combination of Epoch=100 and learning rate=0.0025 reaches the optimal balance point on the validation set, and further increasing the training rounds leads to overfitting. The AdamW optimizer is used, and its core improvement is to decouple weight decay and adaptive learning rate.

[0064] Table 1 Equipment configuration and model parameters

[0065] In order to verify the effectiveness of the proposed estimation model, a series of experiments were conducted, including:

[0066] To ensure the effectiveness of the experiment, we uniformly use the parameters set in Table 1 and conduct experiments in the same environment. It can be clearly seen that the encoder-decoder achieves the best results in all five evaluation indicators. In the robustness experiment, we evaluated the impact of noise to explore the performance of the encoder-decoder under different noise levels. In the experiment, we introduced noise of different amplitudes (50mV, 100mV, and 150mV) to simulate the uncertainty of real battery data. The experimental results are shown in Table 2: Table 2 Results of introducing noise

[0067] It can be clearly observed from the results in the table that under all evaluation indicators, as the noise level increases, the performance of each model decreases. However, the encoder-decoder still maintains a high estimation accuracy under different noise levels. From the test set L14, as the noise level increases from 50mV to 150mV, the MSE of the encoder-decoder only increases from 0.53% to 0.88%. This shows that the encoder-decoder is robust to noise and can resist the influence of data uncertainty to a certain extent.

[0068] In the comparative experiments, we further verified the accuracy of the proposed encoder-decoder by comparing it with multiple open source models (GRU, LSTM and KAN) and evaluating it using the test set. In this series of comparative experiments, we maintained a unified experimental parameter setting and the same environmental conditions to ensure the comparability of the experimental results. The results are shown in Table 3: Table 3 Comparative test results

[0069] It can be clearly observed from the experimental results that the encoder-decoder shows significant advantages under all evaluation indicators. From the test set L14, it can be seen that the encoder-decoder reduces the MSE, RMSE, MAE, MAPE and MAXE of the GRU, LSTM and KAN models by about 56%, 42%, 44%, 36% and 51% respectively. It can be concluded that the encoder-decoder achieved more accurate and stable estimation results on the test set, and achieved obvious advantages under multiple evaluation indicators compared with the open source models GRU, LSTM and KAN. This further proves the excellent performance of the encoder-decoder in the joint estimation of SOH-SOC of quasi-solid-state batteries, and provides a powerful tool and guidance for the optimization of battery management and maintenance strategies.

[0070] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A quasi-solid-state battery SOH-SOC joint estimation method for energy storage scenarios, characterized by: The following steps are involved: Step 1: Obtain a quasi-solid-state battery dataset, which includes a full voltage dataset and a specific voltage dataset; Step 2: Establish a SOH-SOC joint estimation model, which includes a charging encoder, a SOH encoder, a SOH decoder, and a SOC decoder; Step 3: Send the specific voltage data set to the charging encoder and obtain charging encoding data; Step 4: Send the charge encoding data to the SOH encoder and obtain specific SOH decoding data; Step 5: Send the specific SOH decoded data and the entire voltage data set to the SOH encoder to obtain the entire SOH encoded data; Step 6: Send all SOH encoding data and charging encoding data to the SOC decoder to obtain SOC decoding data; Step 7: Using the specific SOH decoded data and the SOC decoded data as estimated values ​​of the battery state of health and the battery state of charge, respectively.

2. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 1 is characterized by: The quasi-solid-state battery dataset is a self-built dataset and includes the capacity and charge changes of the quasi-solid-state battery during the charging process of the entire life cycle; the dataset is the data collected by the ArbinBT2000 battery experimental system during the continuous charging experiment, and all experimental quasi-solid-state batteries follow the same charging curve; the experiment was carried out at a controlled ambient temperature of 25°C, and the battery was charged using a constant current and constant voltage charging protocol; during the constant current charging process, data was collected every 60 seconds; four batteries were selected from the self-built quasi-solid-state battery dataset, named L11, L12, L13 and L14, all of which use lithium iron phosphate as the positive electrode material, with a nominal capacity of 1.1AH and a nominal voltage of 4.2V; The dataset contains comprehensive information about the aging process, including voltage, current, and charging time; L11, L12, and L13 are selected as training datasets, while L14 is selected as the test dataset. A complete constant current and constant voltage charging process is performed with an initial charging current of 0.55A, transitioning to a constant voltage of 4.2V to charge to the cut-off current.

3. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 2 is characterized in that: The specific voltage data set includes data on voltage variation within the voltage range of [3.7V, 3.95V] during the constant current charging phase.

4. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 1 is characterized in that: The charging encoder includes a one-dimensional convolution module, a bidirectional jump TCN-BiGRU network module, a fully connected layer and a dense layer. Step 3 specifically includes the following steps: Step 3.1: The specific voltage dataset is converted by the 1D-CNN layer to obtain local feature data; Step 3.2: The local feature data is passed through the bidirectional jump TCN-BiGRU network module to obtain the model data; Step 3.3: The model data is passed through the fully connected layer and the dense layer in sequence to obtain the charge encoding data.

5. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 4 is characterized in that: The bidirectional jump TCN-BiGRU network module includes a forward temporal convolutional network, a reverse temporal convolutional network, a fully connected layer and a bidirectional gated recurrent unit. The forward temporal convolutional network uses dilated causal convolution to process local feature data in chronological order to ensure that the temporal dependency conforms to the causal relationship; the reverse temporal convolutional network flips the local feature data along the time axis and then performs dilated causal convolution to capture reverse time information; one output branch of both the forward temporal convolutional network and the reverse temporal convolutional network are sent to the bidirectional gated recurrent unit through the fully connected layer, and the other output branch of the two is sent to the bidirectional gated recurrent unit through two skip connections respectively, and when the skip connection conditions are met, the outputs of both the forward temporal convolutional network and the reverse temporal convolutional network are directly sent to the bidirectional gated recurrent unit through the skip connection.

6. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 5 is characterized in that: The jump connection condition is that the output of the forward temporal convolutional network or the output of the reverse temporal convolutional network is inconsistent with the input dimension of the bidirectional gated recurrent unit, and the input dimension is the number of channels or the time step.

7. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 5 is characterized in that: The skip connection conditions are: the input value of a specific voltage dataset is very large or very small, and the depth of the bidirectional gated recurrent unit exceeds 10 layers.

8. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 5 is characterized by: The jump connection condition is: the weight operation in the forward temporal convolutional network and the reverse temporal convolutional network determines whether a specific voltage data set input is a high-value input, wherein when the weight of a specific voltage data set input is greater than 0.5, the specific voltage data set input is recorded as a high-value input.

9. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 1 is characterized by: Both the SOH decoder and the SOC decoder include dense layers. The dense layers of the SOH decoder layer are used to extract long-term degradation features related to battery aging, capture low-frequency, slowly changing global trends, and obtain SOH decoded data after nonlinear transformation of the charging encoded data; the dense layers of the SOC decoder are used to extract dynamic instantaneous features of the battery state, process high-frequency, rapidly changing local signals, and obtain SOC decoded data after nonlinear transformation of the input.

10. The SOH-SOC joint estimation method for quasi-solid-state batteries for energy storage scenarios according to claim 1 is characterized by: The SOH encoder includes a Gaussian noise layer, a padding layer, and a dense layer. The dense layer is used to process fine-grained feature relationships and input the SOH encoded data as part of the SOC decoder after nonlinear transformation.

Images