Lead-carbon battery rapid recession diagnosis method based on feature fusion

Abstract

The invention relates to a lead-carbon battery rapid recession diagnosis method and system based on feature fusion, electronic equipment and a storage medium, multiple features related to battery aging are extracted, an effective feature fusion method is adopted, multiple change factors of a battery in the aging process are comprehensively considered, and the diagnosis accuracy is improved. Compared with a mode of only depending on a single parameter or simply and independently analyzing a plurality of parameters in the prior art, the method can more accurately diagnose whether the accelerated aging of the battery occurs, and greatly improves the diagnosis accuracy. Besides, a diagnosis model constructed based on technologies of deep separable one-dimensional convolution, residual connection, LSTM and the like can quickly process a large amount of battery data, timely discover signs of accelerated aging of the battery, avoid sharp reduction of the performance of the whole module due to untimely diagnosis, guarantee stable operation of the lead-carbon energy storage power station, and improve the reliability of the lead-carbon energy storage power station. And the safety and the reliability of the system are improved.

Description

Technical Field

[0001] This application relates to the field of battery diagnosis, and particularly to a method, system, computer device, and computer-readable storage medium for rapid battery degradation diagnosis based on feature fusion. Background Technique

[0002] The operation architecture of a lead-carbon energy storage power station usually depends on a large number of single lead-carbon batteries closely combined in series and parallel. While this combination method enables high-energy storage and output, it also poses extremely high requirements for the performance stability of single batteries.

[0003] In related technologies, there are limitations in the diagnostic technologies for the accelerated aging of single batteries. Some diagnostic methods rely only on a single parameter. For example, only the battery capacity is monitored to judge the battery state. Since the battery degradation process is complex, a single parameter cannot comprehensively reflect the internal changes of the battery. For example, when the internal resistance of the battery changes, it is difficult to detect the signs of accelerated battery aging in a timely manner only by capacity monitoring.

[0004] In addition, although some methods consider multiple parameters, they simply analyze each parameter independently and do not fully explore the internal relationships between parameters. Facing complex battery degradation situations, the accuracy and timeliness of the diagnostic results are poor.

[0005] Currently, no effective solution has been proposed for the problem of poor accuracy in battery degradation diagnosis in related technologies. Summary of the Invention

[0006] Embodiments of this application provide a method, system, computer device, and computer-readable storage medium for rapid degradation diagnosis of lead-carbon batteries based on feature fusion, so as to at least solve the problem of poor accuracy in battery degradation diagnosis in related technologies.

[0007] In a first aspect, embodiments of this application provide a method for rapid degradation diagnosis of lead-carbon batteries based on feature fusion, which is implemented based on a diagnostic model. Any round of operation steps of the model includes: Obtain a time series vector and scalar features related to the battery. Among them, the time series vector is used to describe the electrical state and thermal state of the battery during operation from multiple dimensions, and the scalar features are used to describe the characteristic changes during the battery aging process from multiple dimensions; Perform depthwise separable one-dimensional convolution processing on the time series vector, and through a residual connection method, add the output of the convolution processing to the original time series vector to obtain a deep feature representation; Through an LSTM network, obtain the long-term dependencies in the time series vector based on the deep feature representation, and perform average pooling on the LSTM output corresponding to each time series vector to obtain optimized time series features; Fuse the optimized timing features and the scalar features to obtain combined features, and through the self-attention mechanism, dynamically weight and fuse the combined features to obtain the battery diagnosis result.

[0008] In some embodiments, the method further includes: Collect the original battery data related to battery aging, and assign labels to the original battery data according to whether accelerated aging occurs; Extract multi-dimensional key features reflecting the performance changes during the battery aging process from the original battery data after label assignment, and respectively construct a timing vector and scalar features according to the multi-dimensional key features; Construct a data set based on the timing vector and the scalar features, and divide the data set into a training set and a test set; Based on the training set, perform model training by iteratively executing the running steps of the diagnostic model. Among them, use the cross-entropy loss function to measure the difference between the probability distribution predicted by the model and the true label distribution, and based on this difference, use the backpropagation algorithm and the optimizer to gradually optimize the model parameters. When the model converges, obtain the trained diagnostic model; Use the trained diagnostic model to perform rapid degradation diagnosis based on the target battery data in the actual scenario.

[0009] In some embodiments, the timing vector corresponds to the battery voltage, current, and temperature data of a specific time segment; the scalar features correspond to the valley mean of the voltage difference curve and the main peak area and main peak height in the capacity increment curve, which are used to describe the characteristic changes during the battery aging process from multiple dimensions.

[0010] In some embodiments, perform separable one-dimensional convolution processing on the timing vector, and add the output of the operation convolution to the original timing vector through a residual connection method. The obtained deep feature representation includes: Perform separable one-dimensional convolution processing on each timing vector respectively, extract the local features in each time series and reduce the computational complexity, and output a local feature set; Adopt a residual connection method to add the local feature set to the original time series to obtain the deep feature representation.

[0011] In some embodiments, through the LSTM network, obtain the long-term dependencies in the timing vector based on the deep feature representation, including: The LSTM network takes the deep deep feature representation as the input, and initializes the cell state and hidden state in the network; Through the forget gate, determine the long-term features to be retained, and through the input gate, evaluate the importance of the current input, and generate short-term key information according to the importance; Merge the retained long-term features and the short-term key information to update the memory unit, and generate the current hidden state through the output gate according to the updated memory unit; By executing the above process step by step in time, dynamically balance the long-term features and short-term anomalies to obtain the LSTM output.

[0012] In some embodiments, the average pooling of the LSTM output corresponding to each time series vector includes: Obtain the output result of the LSTM network, and by taking the average of all elements in the LSTM output corresponding to each time series vector, compress the LSTM output into a fixed-length vector representation to obtain the optimized time series features.

[0013] In some embodiments, dynamically weighted fusion of the combined features through the self-attention mechanism to obtain the battery diagnosis result includes: Obtain the combined features, and calculate the query vector by combining the combined features with a weight matrix and a bias vector; Through the Softmax function, convert the elements in the query vector into probability values to obtain the attention weights, where the sum of the probabilities corresponding to all elements is 1, and the attention weights represent the relative importance of different features under the current attention mechanism; Combine the combined features with the attention weights to obtain the attention output, and process the attention output through the Softnax function again to obtain the probability distribution, where the probability distribution reflects the possibility that the original battery data belongs to different labels.

[0014] In a third aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method described in the first aspect above is implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method described in the first aspect above is implemented.

[0016] Compared with the related technologies, the method for rapid decline diagnosis of lead-carbon batteries based on feature fusion provided by the embodiments of the present application extracts a variety of features related to battery aging and adopts an effective feature fusion method, comprehensively considering various changing factors during the battery aging process. Compared with the prior art that only relies on a single parameter or simply analyzes multiple parameters independently, it can more accurately diagnose whether the battery has accelerated aging, greatly improving the accuracy of diagnosis. In addition, the diagnostic model constructed based on technologies such as depthwise separable one-dimensional convolution, residual connection, and LSTM can quickly process a large amount of battery data, timely detect signs of battery accelerated aging, avoid a sharp decline in the performance of the entire module due to untimely diagnosis, ensure the stable operation of the lead-carbon energy storage power station, and improve the safety and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments and descriptions thereof of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings: Figure 1 is a flowchart of a method for rapid decline diagnosis of lead-carbon batteries based on feature fusion according to an embodiment of the present application; Figure 2 is a structural block diagram of a system for rapid decline diagnosis of lead-carbon batteries based on feature fusion according to an embodiment of the present application; Figure 3 is an internal structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be described and explained below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. Based on the embodiments provided by the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0019] Obviously, the drawings in the following description are only some examples or embodiments of the present application. For those of ordinary skill in the art, without creative efforts, the present application can also be applied to other similar scenarios based on these drawings. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, for those of ordinary skill in the art related to the content disclosed in the present application, some design, manufacturing or production changes based on the technical content disclosed in the present application are only conventional technical means and should not be understood that the content disclosed in the present application is insufficient.

[0020] References to "embodiments" in this application mean that specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application can be combined with other embodiments without conflict.

[0021] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the ordinary meaning understood by those of ordinary skill in the technical field to which this application belongs. The words such as "a", "an", "one kind", "the" and the like involved in this application do not indicate a limitation in quantity and can represent a singular or plural number. The terms "include", "comprise", "have" and any variations thereof involved in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may further include steps or units not listed, or may further include other steps or units inherent to these processes, methods, products or devices. The words such as "connect", "be connected", "couple" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships may exist. For example, "A and / or B" may represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0022] In the field of lead-carbon energy storage power stations, they are usually composed of a large number of single lead-carbon batteries connected in series and parallel. At present, there are certain limitations in the diagnosis of the accelerated aging of single batteries in the prior art. Some diagnostic methods rely only on a single parameter, such as only monitoring the battery capacity to judge the battery state. This approach is too one-sided because the decline of the battery is a complex process, and a single parameter cannot comprehensively reflect the various changes inside the battery. For example, when the internal resistance of the battery changes, it may not be possible to detect the signs of accelerated aging of the battery in time only through capacity monitoring. In addition, although some methods consider multiple parameters, they only simply analyze each parameter independently and do not fully explore the internal relationship between different parameters. This results in poor accuracy and timeliness of the diagnostic results when facing complex battery decline situations.

[0023] During the actual operation of a lead-carbon energy storage power station, once a single battery shows accelerated aging, it will seriously affect the performance of the entire module. Due to the deficiencies of existing diagnostic technologies, accelerated aging single batteries cannot be detected and processed in a timely manner, resulting in a sharp decline in the overall performance of the entire energy storage power station, and thus posing great safety hazards, such as risks of overheating, fire, etc. The main reasons for this inaccurate and untimely diagnosis are the insufficient analysis of the complex relationships among various influencing factors during the battery degradation process and the lack of effective multi-parameter fusion analysis methods. When solving these problems, difficulties are faced in accurately extracting features related to battery accelerated aging and effectively fusing different types of features to improve the diagnosis accuracy.

[0024] In view of this, an embodiment of the present application provides a rapid degradation diagnosis method for lead-carbon batteries based on feature fusion. Figure 1 It is a flowchart of a rapid degradation diagnosis method for lead-carbon batteries based on feature fusion according to an embodiment of the present application, as Figure 1 shown. This method is implemented based on a diagnostic model, and the operation process of the model in any round includes the following steps: S101, collect the original battery data related to battery aging, and assign labels to the original battery data according to whether accelerated aging occurs; The original battery data is data related to battery aging, and it can be obtained through various ways. In this embodiment, the specific acquisition channels and methods are not specifically limited.

[0025] Furthermore, screen the collected battery aging data, and clearly divide it into the accelerated aging stage and the non-accelerated aging stage according to the actual aging state of the battery. For the convenience of subsequent data analysis and model training, set the accelerated aging stage as label 1 and the non-accelerated aging stage as label 0. In this way, the originally complex problem of determining the battery aging inflection point is cleverly transformed into a simple binary classification problem. During model training, the labels provide clear learning goals for the model, and the model tries to make the output results as close as possible to the true categories represented by these labels by continuously adjusting its own parameters.

[0026] S102, extract multi-dimensional key features reflecting the performance changes during the battery aging process from the original battery data after label assignment, and construct a time series vector and scalar features respectively according to the multi-dimensional key features; Among them, multi-dimensional key features are extracted from the original battery aging data for subsequent model analysis, which can reflect the performance changes of the battery during the aging process from different perspectives. Among them, the multi-dimensional key features include but are not limited to: basic features such as current, voltage, and temperature, as well as derivative features under specific conditions, that is, the average valley value of the voltage differential curve trough within the battery power range of 20% - 70%, and the main peak area and peak height in the capacity increment curve.

[0027] In this embodiment, the extracted multi-dimensional characteristic parameters can reflect the battery aging performance changes from multiple dimensions. Current, voltage, and temperature provide information from the basic working state level of the battery; the average valley value of the voltage differential curve trough reveals the aging characteristics from the perspective of the relationship between power and voltage changes; the area and height of the main peak of the capacity increment curve give aging clues from the perspective of battery capacity changes. Combining these characteristic parameters provides rich and key inputs for subsequent model analysis, helps the model accurately capture the characteristic patterns of battery accelerated aging, and improves the accuracy and reliability of battery accelerated aging diagnosis.

[0028] Furthermore, when the lead-carbon energy storage power station is operating, the time series vector composed of battery voltage (V), current (I), and temperature (T) data within a specific time segment can reflect the electrical and thermal states of the battery in real time.

[0029] Among them, the voltage data can reflect the battery's ability to store and release electrical energy, and its fluctuations over time can reflect the internal chemical reaction process of the battery; the magnitude of the current is closely related to the battery's charge and discharge efficiency, and the change in the battery's internal resistance during the aging process will change the current response; the temperature is affected by the internal reaction heat effect of the battery, and abnormal temperature changes indicate that there may be problems with the battery.

[0030] Furthermore, the variable S1 is the average value of the voltage differential curve trough. Within the battery power range of 20% - 70%, its change reflects the voltage change characteristics of the battery in this common power range and can reveal the battery structure and reaction changes. S2 and S3 are the main peak area and peak height of the capacity increment curve respectively. The capacity increment curve reflects the charge and discharge capacity changes of the battery, and the related parameters of its main peak can reflect the battery capacity decline rate and total amount changes, quantifying the battery aging characteristics from different dimensions, complementing the vectors V, I, and T, and jointly providing a comprehensive basis for accurately diagnosing battery accelerated aging, helping to deeply understand the battery aging process.

[0031] S103, construct a data set based on the time series vector and scalar features, and divide the data set into a training set and a test set; Among them, the training set is used to train the diagnostic model. By enabling the model to learn the relationship between the features and labels of the data in the training set, its parameters are continuously adjusted. For example, based on the characteristic parameters such as voltage, current, temperature, etc. of different batteries during the aging process and the corresponding accelerated aging labels in the training set (1 indicates accelerated aging, 0 indicates non-accelerated aging) in the training set, the model optimizes the weight matrices and bias vectors of the internal convolutional layers, fully connected layers, etc. Using the backpropagation algorithm and an optimizer (such as the Adam optimizer), the model gradually improves the classification accuracy of the training data during the training process, enabling accurate judgments when facing battery data with a similar characteristic distribution to the training set.

[0032] The test set is used to evaluate the performance of the model. Since the test set data is not used during the model training process, when the model shows good diagnostic performance on the test set, it indicates that the model has good generalization ability. Reasonably dividing the data set can effectively avoid this situation, ensuring that the model can accurately diagnose whether the battery has accelerated aging on different battery data samples, whether it is data from different time periods of the same energy storage power station or battery data from different energy storage power stations, thereby ensuring the stable operation of the lead-carbon energy storage power station and improving the reliability and safety of the entire system in practical applications.

[0033] S104, based on the training set, the model is trained by iteratively executing the running steps of the diagnostic model. Among them, the cross-entropy loss function is used to measure the difference between the probability distribution predicted by the model and the true label distribution, and based on the difference, the backpropagation algorithm and the optimizer are used to gradually optimize the model parameters. When the model converges, the trained diagnostic model is obtained; Specifically, an exemplary round of the running process of the diagnostic model includes the following steps: Step1, obtain the time series vector and scalar features related to the battery. Among them, the time series vector is used to describe the electrical state and thermal state of the battery during operation from multiple dimensions, and the scalar features are used to describe the characteristic changes during the battery aging process from multiple dimensions; Through the above steps, the time series vector and scalar features are combined as the input for the subsequent model training process respectively. This type of input feature covers the immediate and long-term, macroscopic and microscopic state changes of the battery, comprehensively capturing the battery aging characteristics. There are defects in diagnosing battery aging relying only on a single parameter in the prior art. For example, only considering the capacity cannot detect aging caused by changes in internal resistance in a timely manner. These two types of inputs combine multiple factors, avoid one-sidedness, provide sufficient basis for accurate diagnosis, and improve the reliability and accuracy of diagnosis.

[0034] Step2, perform depthwise separable one-dimensional convolution processing on the time series vector, and through the residual connection method, add the output of the operation convolution to the original time series vector to obtain a deep feature representation; Specifically, for the time series vectors (V, I, T), the operations of Yv = 1D_CNN(V), YI = 1D_CNN(I), and YT = 1D_CNN(T) are respectively performed. In this process, through a specific convolutional kernel sliding along the time dimension, operations such as weighted summation are performed on each time point and its adjacent data, so as to extract local features. For example, for the voltage time series vector V, the convolutional kernel will focus on the voltage value changes within a certain time period, capturing features such as the amplitude and frequency of voltage fluctuations. These features reflect the changes in the electrical performance of the battery during this period. The same applies to the current vector I and the temperature vector T.

[0035] Compared with the traditional convolution method, depthwise separable one-dimensional convolution decomposes the convolution operation into depthwise convolution and pointwise convolution, greatly reducing the computational amount. When dealing with large-scale battery data, this reduction in computational complexity is particularly important, which can improve the training and inference efficiency of the model, while not losing key feature information, ensuring that the model can process battery aging-related data quickly and accurately.

[0036] Furthermore, adopting the residual connection method, the output after depthwise separable one-dimensional convolution is added to the original input to obtain a deep feature representation containing more information and enhanced learning ability, which is specifically realized through the following formula: Yresl_V = Yv + V, Yresl_I = Yi + I, Yresl_T = YT + T; It can be understood that the added features realize the fusion of the original input features and the local features extracted by convolution. The original input contains the most basic information of the data, while depthwise separable one-dimensional convolution extracts the local feature patterns in the data. The addition of the two enables the new deep feature representation to retain both the original information of the data and incorporate the more representative local features extracted through the convolution operation, allowing the model to understand and learn the data from multiple perspectives.

[0037] Furthermore, in a deep neural network, as the number of network layers increases, the gradient may become very small during the backpropagation process, resulting in the model being difficult to learn deep features, that is, the problem of gradient vanishing occurs. In this embodiment, the residual connection (adding the convolution output to the original input) provides a direct path for the propagation of the gradient. When performing backpropagation, the gradient can directly backpropagate from the added result to the original input, avoiding the continuous attenuation of the gradient in the multi-layer network, enabling the model to more efficiently learn deep features and improving the training effect and performance of the model.

[0038] Step3, through the LSTM network, based on the deep feature representation, obtain the long-term dependencies in the time series vectors, and perform average pooling on the LSTM output corresponding to each time series vector to obtain optimized time series features; It should be noted that the above steps have performed operations such as depthwise separable one-dimensional convolution and residual network on the time series vectors of the battery's voltage, current, temperature, etc., and obtained an output that contains local features and has been fused. However, these operations have limited ability to mine long-term dependencies in the data. The battery aging process is a process that gradually changes over time and has obvious long-term dependency features. For example, the charge and discharge cycles of the battery over a long period of time affect its aging degree.

[0039] Long Short-Term Memory (LSTM) is a special type of Recurrent Neural Network (RNN) that can effectively capture long-term dependencies in time series. LSTM controls the flow and memory of information through its unique gating mechanism (input gate, forget gate, and output gate), and can selectively remember or forget information in the time series. When processing battery aging data in the application scenario of this application, LSTM can learn the associations between the state changes of the battery at different time points. For example, the impact of the battery temperature change in the past period on the current battery aging degree, so as to extract more valuable time series features.

[0040] Specifically, the operation process of an LSTM network includes the following steps: S1, The LSTM network takes the deep deep feature representation as the input and initializes the cell state and hidden state in the network; S2, Through the forget gate, judge the long-term features that need to be retained, and through the input gate, evaluate the importance of the current input, and generate short-term key information according to the importance; S3, Combine the retained long-term features and short-term key information to update the memory unit, and generate the current hidden state through the output gate according to the updated memory unit; S4, By executing the above process step by step in time, dynamically balance the long-term features and short-term anomalies to obtain the LSTM output.

[0041] In the above steps, through the forget gate, LSTM can selectively forget the irrelevant information in the past, avoiding the problems of gradient vanishing or gradient explosion in traditional RNNs, enabling the model to process longer time series. Further, the existence of the input gate and output gate enables LSTM to selectively update the cell state and output the hidden state according to the current input and the past hidden state, thus effectively capturing the long-term dependencies in the time series.

[0042] It can be understood that after the LSTM processing, each time series vector has a corresponding LSTM output. Further, the global average pooling operation processes these LSTM outputs. Optionally, the average value of all elements in the LSTM output corresponding to each time series vector is calculated, and the features of the entire time series are compressed into a vector representation of a fixed length to obtain optimized time series features, which can be specifically implemented through the following formula:

[0043] It can be understood that through the global average pooling operation, it enables the features of different time series to be processed in a unified dimension during subsequent feature fusion, facilitating effective fusion with other scalar features (such as the average value of the trough of the voltage difference curve and the area and peak height of the main peak in the capacity increment curve).

[0044] Step4, fuse the optimized time series features and scalar features to obtain combined features, and through the self-attention mechanism, dynamically weighted fuse the transformed scalar features and time series deep feature representations to obtain the battery diagnosis result.

[0045] Specifically, for the scalar features S1, S2, and S3, their representation forms are transformed through a fully connected layer (FCL) with the aim of enabling them to be effectively fused with the previously processed time series features. The fully connected layer can learn appropriate weights and biases to map S1, S2, and S3 to a feature space that matches the time series features, obtaining the fused feature S.

[0046] Further, fuse the optimized time series features and scalar features to obtain combined features, where the combined features can be expressed by the following expression:

[0047] In the feature fusion stage, it is necessary to integrate and further process these features from different sources and types, and subsequently perform operations such as query vector calculation and attention weight calculation based on it to achieve more effective feature fusion and battery accelerated aging diagnosis.

[0048] Further, through the self-attention mechanism, dynamically weighted fuse the combined features to obtain the battery diagnosis result, which specifically includes the following steps: S1, obtain the combined features, and calculate the query vector by combining the combined features with the weight matrix and bias vector; Among them, the query vector is expressed by the following formula:

[0049] Among them is the weight matrix, which determines the pair of The weighting degrees of different features is the bias vector, which plays a role in adjusting the calculation result. The calculated query vector will participate in the subsequent calculation of attention weights, and its role is to measure the importance of different positions in the data, so as to help the model focus on key information and improve the effect of data processing and analysis. For example, in tasks such as classification, it can better extract and utilize effective features.

[0050] S2, through the Softmax function, converts the elements in the query vector into probability values to obtain attention weights. Among them, the sum of the probabilities corresponding to all elements is 1, and the attention weights represent the relative importance of different features under the current attention mechanism; The specific process can be expressed by the following formula:

[0051] where A is the attention weight.

[0052] Among them, the essence of the Softmax function is a kind of normalized exponential function. For each element in the query vector, the attention weight after being converted by the Softmax function. The sum of the attention weights corresponding to all these elements is 1. In the context of battery aging diagnosis, these attention weights represent the relative importance of different position features under the current attention mechanism.

[0053] If the attention weight corresponding to a certain feature is large, it means that in the current model's process of judging whether the battery is accelerating aging, this feature is more critical. The model will pay more attention to these important features in subsequent processing, thereby improving the accuracy and reliability of the diagnosis, helping to more accurately identify the key information of the battery aging state, and avoiding misjudgment caused by interference from secondary information.

[0054] S3, combines the combined features with the attention weights to obtain the attention output, and processes the attention output through the Softmax function again to obtain the probability distribution. Among them, the probability distribution reflects the possibility that the original battery data belongs to different labels.

[0055] Among them, for the combined features , calculate the attention output to re-weight the combined features.

[0056] For example, if the attention weight corresponding to the battery voltage feature at a certain time point is high, then when calculating the attention output , the voltage feature at this time point will be strengthened in the new deep feature representation, and its value will relatively increase; while for the feature with a low weight, its influence in will be weakened.

[0057] The advantage of doing this is that it can highlight the feature information that is more crucial for diagnosing the accelerated aging of the battery, enabling the model to focus more on these important features during subsequent operations such as classification output, reducing the interference of irrelevant or secondary features, and thus improving the accuracy and reliability of the entire diagnostic model.

[0058] Further, after calculating the attention output, the classification output is calculated, where the output is converted into a probability distribution through the softmax() function, and the final output result is obtained; This process can be represented by the following formula:

[0059] After a series of previous operations of feature extraction, fusion, and attention mechanism, the obtained result needs to be further transformed into an intuitive form with clear judgment basis. The softmax function can map the output values to between 0 and 1, and the sum of all output values is 1, forming a probability distribution.

[0060] When judging whether the battery is accelerating aging, different probability values in this probability distribution correspond to the possibility of the battery being in different states (accelerated aging or non-accelerated aging). For example, if the probability value corresponding to accelerated aging in the output probability distribution is close to 1, and the probability value of non-accelerated aging is close to 0, then it can be more deterministically judged that the battery has undergone accelerated aging.

[0061] During the above model operation process, the LSTM is used to capture the long-term dependence relationship of the battery aging data, mine the hidden information of the data changing over time, and further combine the global average pooling to compress the features into a fixed-length vector, unifying the feature dimensions; in the feature fusion link, the scalar features are converted through the fully connected layer and fused with the processed time series features to integrate multi-type features and enrich the diagnostic information dimensions; finally, the self-attention mechanism further optimizes the fusion effect, calculates the query vector, attention weights, and output, re-weights the features to highlight the key information, and finally converts it into a probability distribution output result through the softmax function, providing an intuitive and accurate basis for judging whether the battery is accelerating aging, enhancing the diagnostic ability of the model for complex battery aging situations, and ensuring the stable operation of the lead-carbon energy storage power station.

[0062] Combined with the above embodiments, it can be understood that the overall model training process in the solution of the present application is carried out based on the training set by iteratively executing the running steps of the diagnostic model. Specifically, the cross-entropy loss function is selected for the model training process to accurately measure the degree of difference between the model prediction result and the true label.

[0063] Among them, the loss function can be represented by the following formula:

[0064] Furthermore, during training, the model parameters are updated by combining the backpropagation algorithm with the Adam optimizer. The backpropagation algorithm calculates the gradient information based on the loss function and propagates the error from the output layer back to each hidden layer and the input layer, thereby determining the contribution of each parameter to the error. The Adam optimizer uses an adaptive learning rate adjustment strategy to dynamically adjust the learning step size for different parameters, enabling the model to converge faster and more stably during training.

[0065] Through continuous iterative training and repeated adjustment of the weight matrices and bias vectors involved in each convolutional layer, fully connected layer, etc., the model continuously learns the characteristic patterns and rules in the battery aging data, gradually reduces the loss value, improves the classification performance, and finally achieves accurate classification of the accelerated aging state of the battery, effectively solving the problems of inaccurate and untimely diagnosis in the existing technology and ensuring the safe and stable operation of the lead-carbon energy storage power station.

[0066] S105, using the trained diagnostic model, perform a rapid degradation diagnosis on the target battery in the actual scenario.

[0067] It can be understood that after the training of the diagnostic model is completed, it enters the actual application stage. In the actual scenario, for the target battery, collect and process its related time series data such as voltage, current, temperature, etc., as well as characteristic data such as voltage difference curves and capacity increment curves to make them meet the model input requirements.

[0068] Then input these data into the trained diagnostic model. The model will quickly determine whether the target battery is in an accelerated aging and degradation state based on the characteristic patterns and rules of battery aging learned during training. This step fully utilizes the value of the previous work such as data preprocessing, feature extraction, model construction, and training, can timely and accurately detect the aging problems of the target battery, provide a strong guarantee for the safe and stable operation of the lead-carbon energy storage power station, and effectively avoid problems such as the performance decline of the module and safety hazards caused by the failure to timely diagnose the accelerated aging of the battery.

[0069] Through the above steps S101 to S105, a variety of features related to battery aging are extracted, and an effective feature fusion method is adopted, comprehensively considering various changing factors during the battery aging process. Compared with the existing technology that only relies on a single parameter or simply analyzes multiple parameters independently, it can more accurately diagnose whether the battery is undergoing accelerated aging, greatly improving the accuracy of diagnosis. In addition, the diagnostic model constructed based on technologies such as depthwise separable one-dimensional convolution, residual connection, and LSTM can quickly process a large amount of battery data, timely detect signs of accelerated battery aging, avoid a sharp decline in the performance of the entire module due to untimely diagnosis, ensure the stable operation of the lead-carbon energy storage power station, and improve the safety and reliability of the system.

[0070] In a second aspect, the embodiments of the present application further provide a rapid degradation diagnosis system for lead-carbon batteries based on feature fusion. Figure 2 It is a structural block diagram of a rapid degradation diagnosis system for lead-carbon batteries based on feature fusion according to an embodiment of the present application, as Figure 2 shown. The system includes: a preprocessing module 20, a feature optimization module 21, and an output module 22, where: The preprocessing module 20 is configured to obtain time-series vectors and scalar features related to the battery. Among them, the time-series vectors are used to describe the electrical state and thermal state of the battery during operation from multiple dimensions, and the scalar features are used to describe the feature changes during the battery aging process from multiple dimensions; and perform depthwise separable one-dimensional convolution processing on the time-series vectors, and through a residual connection method, add the output of the convolution processing to the original time-series vectors to obtain a deep feature representation. The feature optimization module 21 is configured to obtain the long-term dependencies in the time-series vectors based on the deep feature representation through an LSTM network, and perform average pooling on the LSTM output corresponding to each time-series vector to obtain optimized time-series features. The output module 22 is configured to fuse the optimized time-series features and scalar features to obtain combined features, and dynamically weight and fuse the combined features through a self-attention mechanism to obtain a battery diagnosis result.

[0071] Through the above system, various features related to battery aging are extracted, and an effective feature fusion method is adopted, comprehensively considering various change factors during the battery aging process. Compared with the prior art that only relies on a single parameter or simply analyzes multiple parameters independently, it can more accurately diagnose whether the battery has accelerated aging, greatly improving the accuracy of diagnosis. In addition, the diagnostic model constructed based on technologies such as depthwise separable one-dimensional convolution, residual connection, and LSTM can quickly process a large amount of battery data, timely detect signs of battery accelerated aging, avoid a sharp decline in the performance of the entire module due to untimely diagnosis, ensure the stable operation of the lead-carbon energy storage power station, and improve the safety and reliability of the system.

[0072] In one embodiment, Figure 3 It is an internal structure schematic diagram of an electronic device according to an embodiment of the present application, as Figure 3 shown. An electronic device is provided. The electronic device can be a server, and its internal structure diagram can be as Figure 3As shown. The electronic device includes a processor, a network interface, an internal memory, and a non-volatile memory connected through an internal bus. Among them, the non-volatile memory stores an operating system, a computer program, and a database. The processor is used to provide computing and control capabilities. The network interface is used to communicate with external terminals through a network connection. The internal memory is used to provide an environment for the operation of the operating system. The computer program, when executed by the processor, implements a method for rapid degradation diagnosis of lead-carbon batteries based on feature fusion. The database is used to store data.

[0073] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the electronic device to which the solution of this application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0074] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above embodiments of the methods. Among them, any reference to memory, storage, database, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0075] The above embodiments only represent several implementation manners of this application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several deformations and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of the patent of this application should be subject to the appended claims.

Claims

1. A lead-carbon battery rapid degradation diagnosis method based on feature fusion, characterized in that: Based on the diagnosis model implementation, any round of operation of the model includes: Obtaining battery-related timing vectors and scalar features, wherein the timing vectors are used to describe the electrical and thermal states of the battery during operation from multiple dimensions, and the scalar features are used to describe the characteristic changes during battery aging from multiple dimensions; Performing a depth-separable one-dimensional convolution process on the time series vector, and adding the output of the convolution process to the original time series vector through a residual connection method to obtain a deep feature representation; Through the LSTM network, the long-term dependency in the time series vector is obtained based on the deep feature representation, and the LSTM output corresponding to each time series vector is averaged and pooled to obtain optimized time series features; The optimized timing feature and the scalar feature are fused to obtain a combined feature, and the combined feature is dynamically weighted and fused through a self-attention mechanism to obtain a battery diagnosis result.

2. The method according to claim 1, characterized in that: The method further comprises: Collecting raw battery data related to battery aging, and assigning labels to the raw battery data according to whether accelerated aging occurs; Extracting multi-dimensional key features reflecting performance changes during battery aging from raw battery data after label assignment, and constructing time series vectors and scalar features respectively based on the multi-dimensional key features; Constructing a data set based on the time series vector and the scalar feature, and dividing the data set into a training set and a test set; Based on the training set, model training is performed by iteratively executing the operation steps of the diagnostic model, wherein a cross entropy loss function is used to measure the difference between the probability distribution predicted by the model and the true label distribution, and based on the difference, a back propagation algorithm and an optimizer are used to gradually optimize the model parameters, and when the model converges, a trained diagnostic model is obtained; The trained diagnostic model is used to perform rapid degradation diagnosis based on target battery data in actual scenarios.

3. The method according to claim 1, characterized in that The time series vector corresponds to the battery voltage, current and temperature data of a specific time segment; the scalar feature corresponds to the valley mean of the voltage difference curve and the main peak area and main peak height in the capacity increment curve, which is used to describe the characteristic changes of the battery aging process from multiple dimensions.

4. The method according to claim 1, characterized in that: The time series vector is subjected to a separable one-dimensional convolution process, and the output of the operation convolution is added to the original time series vector through a residual connection method to obtain a deep feature representation including: Perform separable one-dimensional convolution processing on each time series vector, extract local features in each time series and reduce computational complexity, and output a local feature set; The local feature set is added to the original time series by using a residual connection method to obtain the deep feature representation.

5. The method according to claim 4, characterized in that Through the LSTM network, the long-term dependencies in the time series vector are obtained based on the deep feature representation, including: The LSTM network takes the deep feature representation as input and initializes the cell state and hidden state in the network; Through the forget gate, determine the long-term features that need to be retained, and through the input gate, evaluate the importance of the current input and generate short-term key information based on the importance; Merging the retained long-term features with the short-term key information to update the memory unit, and generating a current hidden state according to the updated memory unit through an output gate; By executing the above process time by time, the long-term features and short-term anomalies are dynamically balanced to obtain the LSTM output.

6. The method according to claim 5, characterized in that Average pooling of the LSTM output corresponding to each time series vector includes: The output result of the LSTM network is obtained, and the optimized time series feature is obtained by averaging all elements in the LSTM output corresponding to each time series vector and compressing the LSTM output into a vector representation of a fixed length.

7. The method according to claim 5, characterized in that The combined features are dynamically weighted and fused through the self-attention mechanism to obtain the battery diagnosis results including: Obtaining the combined feature, and calculating a query vector by combining the combined feature with a weight matrix and a bias vector; The elements in the query vector are converted into probability values ​​through the Softmax function to obtain attention weights, where the sum of the probabilities corresponding to all elements is 1, and the attention weights represent the relative importance of different features under the current attention mechanism; The combined feature is combined with the attention weight to obtain an attention output, and the attention output is processed again by the Softnax function to obtain a probability distribution, wherein the probability distribution reflects the possibility that the original battery data belongs to different labels.

8. A lead-carbon battery rapid degradation diagnosis system based on feature fusion, characterized in that: The system comprises: a preprocessing module, a feature optimization module and an output module, wherein: The preprocessing module is used to obtain battery-related time series vectors and scalar features, wherein the time series vectors are used to describe the electrical state and thermal state of the battery during operation from multiple dimensions, and the scalar features are used to describe the characteristic changes during the battery aging process from multiple dimensions; And, performing a depth-separable one-dimensional convolution process on the time series vector, and adding the output of the convolution process to the original time series vector through a residual connection method to obtain a deep feature representation; The feature optimization module is used to obtain the long-term dependency in the time series vector based on the deep feature representation through the LSTM network, and average pool the LSTM output corresponding to each time series vector to obtain the optimized time series feature; The output module is used to fuse the optimized timing features and the scalar features to obtain a combined feature, and dynamically weighted fuse the combined features through a self-attention mechanism to obtain a battery diagnosis result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images