Artificial intelligence battery early warning system and method based on adaptive deep learning

Abstract

The invention discloses an artificial intelligence battery early warning system and method based on adaptive deep learning, and the method comprises the steps: carrying out the collection and preprocessing of battery data: carrying out the collection of the data of a battery, and carrying out the adaptive feature extraction: extracting features facilitating the description of the state of an energy storage system from the preprocessed data, deep learning model training: selecting a proper deep learning model architecture, and training the model by using a historical battery data set; battery health state evaluation: analyzing the real-time data of the battery by using the trained deep learning model, and evaluating the health state of the battery; and fault early warning signal generation: according to the battery health state evaluation result, the artificial intelligence battery early warning system and method based on adaptive deep learning perform real-time analysis on the operation data of the battery through the deep learning technology, and realize battery health state monitoring and potential fault early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence technology in battery monitoring, and particularly to an artificial intelligence battery warning system and method based on adaptive deep learning. Background Art

[0002] With the rapid development of technology, batteries have become an indispensable energy storage unit in modern society. In multiple fields such as mobile communication and smart grid energy storage systems, the performance and safety of batteries are crucial. The performance degradation and sudden failures of batteries not only affect the user experience of devices, but also may trigger safety accidents such as fires or explosions, causing casualties and property losses.

[0003] To ensure the reliability and safety of battery systems, it is particularly important to monitor the battery state in real time and give early warnings of faults. Currently, the methods for battery monitoring and fault diagnosis mainly include the following: ① Monitoring of voltage and current thresholds; ② Battery model analysis; ③ State monitoring and health state analysis; ④ Machine learning technology.

[0004] Although the above methods have achieved certain development in the field of battery monitoring, there are still the following limitations: ① Insufficient accuracy; ② Limited generalization ability; ③ Poor real-time performance; ④ High maintenance cost. Summary of the Invention

[0005] In view of the limitations of the prior art, the present invention proposes an artificial intelligence battery warning system and method based on adaptive deep learning. The system adapts the parameters of the deep learning model to batteries of different types and usage conditions, improving the accuracy of early warnings and the generalization ability of the system. At the same time, the system can analyze the operation data of the battery and quickly generate fault warning signals, effectively enhancing the safety and reliability of the battery system.

[0006] To solve the above technical problems, one technical solution adopted by the present invention is to provide an artificial intelligence battery warning system based on adaptive deep learning, including:

[0007] A data acquisition module for collecting multivariate parameter data of the system operation in real time;

[0008] A data preprocessing module for cleaning and normalizing the collected data;

[0009] An adaptive feature extraction module that uses an adaptive algorithm to extract features related to the battery health state from the preprocessed data;

[0010] A deep learning model training module for constructing a deep learning model for battery health state evaluation and fault prediction;

[0011] A health status assessment module, which is used to analyze the real-time data of the battery by using a trained model, evaluate the health status of the battery, and provide real-time monitoring and evaluation of the battery status;

[0012] A fault warning signal generation module, which is used to generate fault warning signals and maintenance measures according to the health status assessment results;

[0013] A communication module, which is responsible for data exchange with other system components to ensure the real-time and reliability of data transmission;

[0014] A power management module, which provides a stable power supply for the entire system to ensure that the system can continue to operate during signal interruption.

[0015] The present invention also provides another technical solution: a warning method for an artificial intelligence battery system based on adaptive deep learning, including the following steps:

[0016] S1. Collect the data of the battery;

[0017] S2. Preprocess the data of the battery;

[0018] S3. Adaptive feature extraction: Extract features that help describe the state of the energy storage system from the preprocessed data, and use statistical analysis methods to extract the most optimal representative features according to the actual operating conditions of the battery;

[0019] S4. Deep learning model training: Select a suitable deep learning model architecture, train the model using a historical pool dataset, adjust the model parameters through cross-validation methods, and optimize the model performance;

[0020] S5. Battery health status assessment: Analyze the real-time data of the battery by using the trained deep learning model, evaluate the health status of the battery, and the model outputs the health status score or fault probability of the battery;

[0021] S6. Fault warning signal generation: According to the battery health status assessment results, set a warning threshold, and generate a fault warning signal when the health status of the battery is lower than the threshold.

[0022] In a preferred embodiment of the present invention, in S1, the data of the battery includes the following parameters: battery voltage, battery current, battery temperature, and battery charge and discharge cycle times.

[0023] In a preferred embodiment of the present invention, the battery data preprocessing includes the following steps:

[0024] S201. Clean the collected original battery data to remove outliers and noise interference;

[0025] S202. Normalize the data so that the battery data reaches the input range of the deep learning model.

[0026] In a preferred embodiment of the present invention, the features helpful for describing the state of the energy storage system include the temperature change rate, voltage fluctuation, and current peak.

[0027] In a preferred embodiment of the present invention, the PCA calculation formula involved in the statistical analysis method in S3 is as follows: ① Calculate the mean and covariance matrix of the original data:

[0028]

[0029] where x i is the data point, is the mean vector, ∑ is the covariance matrix, n represents the number of data points, and T represents the transpose of the data point;

[0030] ② Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors:

[0031] ∑ = λv

[0032] where v is the eigenvector and λ is the corresponding eigenvalue;

[0033] ③ Select the eigenvectors corresponding to the eigenvalues to form a projection matrix:

[0034] P = [v1, v2,..., v k

[0035] where P is the projection matrix composed of eigenvectors.

[0036] ④ The data after dimensionality reduction:

[0037] x′ = P T x

[0038] where x′ is the data after dimensionality reduction, P T is the data of the projection matrix, and x is the original data matrix.

[0039] In a preferred embodiment of the present invention, in S4, the historical pool dataset includes normal data and fault data.

[0040] In a preferred embodiment of the present invention, in S4, the calculation formula involved in the deep learning model architecture is as follows:

[0041] Loss function formula:

[0042]

[0043] where y i is the true value, ​is the predicted value, L is the loss function, and n is the number of samples in the dataset; weight formula:

[0044]

[0045] where ω is the weight and η is the learning rate, is the gradient of the loss function with respect to the weight.

[0046] In a preferred embodiment of the present invention, in S5, the calculation formula of the long short-term memory network involved in the health status assessment is:

[0047] Forget gate:

[0048] f t = σ(W f · [h t-1 , x t + b f )

[0049] where f t are the activation vectors of the forget gate respectively, σ is the activation function, W f is the weight matrix of the forget gate, b f is the bias vector of the forget gate, h i-1 is the hidden state at the previous moment, and x t is the input at the current moment;

[0050] Input gate:

[0051] i t = σ(W i · [h t-1 , x t + b i )

[0052]

[0053] where i t is the activation vector of the input gate, is the candidate memory cell state, σ is the activation function, W i and W c are the weight matrices of the input gate and the candidate memory cell state respectively, h i-1 is the hidden state at the previous moment, x t is the input at the current moment, tanh is the hyperbolic tangent activation function, b i and b i are the corresponding bias vectors. Memory cell:

[0054]

[0055] where C tis the memory cell state at the current memory moment, f t is the activation vector of the forget gate, C t-1 is the memory cell state at the previous moment, is the candidate memory cell state, i t is the activation vector of the input gate.

[0056] Output gate:

[0057] o t = σ(W o · [h t-1 , x t + b0)

[0058] h t = o t * tanh(C t )

[0059] where, o t is the activation vector of the output gate, C t is the state value of the memory cell, h t is the output value, W o and b0 are the weight and bias parameters, σ is the activation function, h t-1 is the hidden state at the previous moment.

[0060] In a preferred embodiment of the present invention, in S6, the calculation formula involved in the generation of the fault warning signal is:

[0061] Dynamic Time Warping (DTW) distance calculation formula:

[0062] DTW(Q, T) = min(min(DTW(Q[1:m - 1], T[1:n - 1]), DTW(Q[2:m], T[1:n - 1])), min(DTW(Q[1:m - 1], T[2:n]), DTW(Q[1:m], T[2:n])))

[0063] where, Q and T are two time series, m and n are the sequence lengths, the DTW distance measures the similarity of the two sequences, considering the stretching of the time axis.

[0064] The beneficial effects of the present invention are: by adaptively adjusting the parameters of the deep learning model to adapt to batteries under different types and usage conditions, the accuracy of the warning and the generalization ability of the system are improved. At the same time, the system can analyze the operation data of the battery and quickly generate a fault warning signal, effectively improving the safety and reliability of the battery system. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, where:

[0066] Figure 1 It is a schematic diagram of the early warning method for the artificial intelligence battery system based on adaptive deep learning provided by the embodiment of the present invention;

[0067] Figure 2 It is a flowchart of the early warning method for the artificial intelligence battery system based on adaptive deep learning provided by the embodiment of the present invention:

[0068] Figure 3 It is a flowchart of data preprocessing provided by the embodiment of the present invention;

[0069] Figure 4 It is a flowchart of the learning of the deep learning model provided by the embodiment of the present invention;

[0070] Figure 5 It is a flowchart of the generation of the fault early warning signal provided by the embodiment of the present invention;

[0071] Figure 6 It is a schematic diagram of the structure of the artificial intelligence battery early warning system based on adaptive deep learning provided by the embodiment of the present invention. Specific embodiments

[0072] The following will clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0073] Please refer to Figure 1-2 , the embodiments of the present invention include: an early warning method for an artificial intelligence battery system based on adaptive deep learning, including the following steps:

[0074] S1. Battery data collection;

[0075] The collection of the battery data includes parameters such as voltage, current, temperature, and number of charge and discharge cycles.

[0076] S2. Battery data preprocessing: The data preprocessing includes: S201. Cleaning the collected original battery data to remove outliers and noise interference; S202. Normalizing the data to make the battery data reach the input range of the deep learning model;

[0077] S3. Adaptive Feature Extraction: For adaptive feature extraction, it includes: S301. Extract features (such as temperature change rate, voltage fluctuation, current peak, etc.) that are helpful for describing the state of the energy storage system from the preprocessed data; S302. Use statistical analysis methods (Principal Component Analysis PCA) to extract the most representative features according to the actual operation of the battery.

[0078] Among them, the PCA calculation formula involved in the statistical analysis method is:

[0079] ① Calculate the mean and covariance matrix of the original data:

[0080]

[0081] Among them, x i is the data point, is the mean vector, ∑ is the covariance matrix, n represents the number of data points, and T represents the transpose of the data point.

[0082] ② Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors:

[0083] ∑ = λv

[0084] Among them, v is the eigenvector and λ is the corresponding eigenvalue.

[0085] ③ Select the eigenvectors corresponding to the eigenvalues to form a projection matrix:

[0086] P = [v1, v2,..., v k

[0087] Among them, P is the projection matrix composed of eigenvectors.

[0088] ④ The data after dimensionality reduction:

[0089] x′ = P T x

[0090] Among them, x′ is the data after dimensionality reduction, P T is the data of the projection matrix, and x is the original data matrix.

[0091] S4. Deep Learning Model Training;

[0092] For the training of the deep learning model, it includes: S401. Select a suitable deep learning model architecture;

[0093] S402. Use the historical pool dataset (including normal data and fault data) to train the model;

[0094] S403. Adjust the model parameters through the cross-validation method to optimize the model performance. ​

[0095] Among them, the calculation formulas involved in the deep learning model architecture are as follows:

[0096] Loss function formula:

[0097]

[0098] Among them, y i is the true value, is the predicted value, L is the loss function, and n is the number of samples in the dataset; Weight formula:

[0099]

[0100] Among them, ω is the weight, η is the learning rate, is the gradient of the loss function with respect to the weight.

[0101] S5, Health status assessment;

[0102] For the above-mentioned health status assessment, it includes: S501, Analyze the real-time data of the battery using the trained deep learning model to evaluate the health status of the battery; S502, The model outputs the health status score or failure probability of the battery.

[0103] Among them, the calculation formulas of the long short-term memory network (LSTM) involved in the health status assessment are as follows:

[0104] Forget gate:

[0105] f t = σ(W f · [h t-1 , x t + b f )

[0106] Among them, f t are the activation vectors of the forget gate respectively, σ is the activation function, W f is the weight matrix of the forget gate, b f is the bias vector of the forget gate, h i-1 is the hidden state at the previous moment, and x t is the input at the current moment.

[0107] Input gate:

[0108] i t = σ(W i · [h t-1 , x t + b i )

[0109]

[0110] Among them, it is the activation vector of the input gate, is the candidate memory cell state, σ is the activation function, W i and W c are the weight matrices of the input gate and the candidate memory cell state respectively, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, tanh is the hyperbolic tangent activation function, b i and b i are the corresponding bias vectors.

[0111] Memory cell:

[0112]

[0113] where C t is the memory cell state at the current memory moment, f t is the activation vector of the forget gate, C t-1 is the memory cell state at the previous moment, is the candidate memory cell state, i i is the activation vector of the input gate.

[0114] Output gate:

[0115] o t = σ(W o · [h t-1 , x t + b0)

[0116] h t = o t * tanh(C t )

[0117] where o t is the activation vector of the output gate, C t is the state value of the memory cell, h t is the output value, W0 and b0 are the weight and bias parameters, σ is the activation function, h i-1 is the hidden state at the previous moment.

[0118] S6, Fault warning signal generation;

[0119] For the generation of the fault warning signal, it includes: S601. Set the warning threshold according to the battery health status evaluation result; S602. Generate a fault warning signal (fault type, possible impacts of the fault, and recommended maintenance measures) when the health status of the battery is lower than the threshold.

[0120] Among them, the calculation formula involved in the generation of the fault warning signal is:

[0121] Dynamic Time Warping (DTW) distance calculation formula:

[0122] DTW(Q, T) = min(min(DTW(Q[1:m - 1], T[1:n - 1]), DTW(Q[2:m], T[1:n - 1])), min(DTW(Q[1:m - 1], T[2:n]), DTW(Q[1:m], T[2:n])))

[0123] Where Q and T are two time series, m and n are the lengths of the series, and the DTW distance measures the similarity between the two series, taking into account the stretching of the time axis.

[0124] As an alternative implementation, Figure 3 The flowchart of data preprocessing provided by the embodiments of the present invention is as Figure 3 shown. Preprocessing the collected data includes:

[0125] S201. Cleaning the collected original battery data to remove outliers and noise interference; S202. Normalizing the data to make the battery data reach the input range of the deep learning model.

[0126] As an alternative implementation, Figure 4 The flowchart of deep learning model training provided by the embodiments of the present invention is as Figure 4 shown. Training the deep learning model includes:

[0127] S401. Selecting a suitable deep learning model architecture; S402. Training the model using a historical pool dataset (including normal data and fault data); S403. Adjusting the model parameters through a cross - validation method to optimize the model performance.

[0128] Among them, the calculation formulas involved in the deep learning model architecture are:

[0129] Loss function formula:

[0130]

[0131] Where y i is the true value, is the predicted value, L is the loss function, and n is the number of samples in the dataset;

[0132] Weight formula:

[0133]

[0134] Where ω is the weight, η is the learning rate, is the gradient of the loss function with respect to the weight.

[0135] As an alternative embodiment, Figure 5 is a flowchart for generating a fault warning signal provided by an embodiment of the present invention. As shown in Figure 5 shown, generating a signal for battery faults includes:

[0136] S601. Set a warning threshold according to the battery health status evaluation result; S602. Generate a fault warning signal (fault type, possible impacts of the fault, and recommended maintenance measures) when the battery health status is lower than the threshold.

[0137] Among them, the calculation formula involved in generating the fault warning signal is:

[0138] Dynamic Time Warping (DTW) distance calculation formula:

[0139] DTW(Q, T) = min(min(DTW(Q[1:m - 1], T[1:n - 1]), DTW(Q[2:m], T[1:n

[0140] - 1])), min(DTW(Q[1:m - 1], T[2:n]), DTW(Q[1:m], T[2:n])))

[0141] Among them, Q and T are two time series, m and n are the sequence lengths, the DTW distance measures the similarity between the two sequences, considering the stretching of the time axis.

[0142] An embodiment of the present invention also provides an artificial intelligence battery warning system based on adaptive deep learning, including:

[0143] A data acquisition module 100 for collecting multivariate parameter data of the system operation in real time. The multivariate parameter data of the system are key parameters such as voltage, current, temperature, charge and discharge cycle times, etc. The multivariate parameter data of the system are crucial for subsequent battery state evaluation and fault diagnosis. Therefore, in order to improve the accuracy of subsequent calculations, as many multivariate parameter data in the system historical data as possible can be obtained.

[0144] A data preprocessing module 200 for cleaning and normalizing the collected data;

[0145] Clean the collected original data, remove outliers and noise interference, improve the data quality, and normalize the data to make it adapt to the input range of the deep learning model.

[0146] An adaptive feature extraction module 300 uses an adaptive algorithm to extract features related to the battery health status from the preprocessed data; Using the adaptive algorithm can dynamically adjust the feature extraction strategy according to the actual operation of the battery to improve the representativeness and discrimination of the features.

[0147] Among them, the PCA calculation formula involved in the statistical analysis method is as follows:

[0148] ① Calculate the mean and covariance matrix of the original data:

[0149]

[0150] Among them, x i is a data point, is the mean vector, ∑ is the covariance matrix, n represents the number of data points, and T represents the transpose of the data point.

[0151] ② Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors:

[0152] ∑ = λv

[0153] Among them, v is the eigenvector and λ is the corresponding eigenvalue.

[0154] ③ Select the eigenvectors corresponding to the eigenvalues to form a projection matrix:

[0155] P = [v1, v2,..., v k

[0156] Among them, P is the projection matrix composed of eigenvectors.

[0157] ④ The data after dimensionality reduction:

[0158] x' = P T x

[0159] Among them, x' is the data after dimensionality reduction, P T is the data of the projection matrix, and x is the original data matrix.

[0160] The deep learning model training module 400 is used to build a deep learning model for battery health state assessment and fault prediction. It is responsible for selecting a suitable deep learning model architecture, training the deep learning model using the historical pool dataset (including normal data and fault data), adjusting the model parameters through the cross-validation method, optimizing the model performance, forming a suitable deep learning model architecture, and evaluating and warning the battery health state and faults;

[0161] The health state assessment module 500 is used to analyze the real-time battery data using the trained model, evaluate the battery health state, and provide real-time monitoring and evaluation of the battery state;

[0162] The fault warning signal generation module is responsible for generating fault warning signals and maintenance measures according to the health state assessment results;

[0163] ​The fault warning signal generation module will set corresponding warning thresholds according to the health status assessment results. If the evaluated value is not within the safe warning threshold range, a fault warning signal will be generated; if the evaluated value is within the safe warning threshold range, no fault warning signal will be generated.

[0164] In addition, the generated fault warning signal includes not only simple numerical values, but also fault types, possible impacts, and recommended maintenance measures, so that the staff can carry out maintenance measures as early as possible to reduce the losses caused by faults.

[0165] The communication module 600 is responsible for data exchange with other system components to ensure the real-time and reliability of data transmission.

[0166] The power management module 700 provides a stable power supply for the entire system to ensure that the system can continue to operate in case of signal interruption.

[0167] The beneficial effects of the artificial intelligence battery warning system and method based on adaptive deep learning of the present invention are as follows: By adaptively adjusting the parameters of the deep learning model to adapt to batteries of different types and usage conditions, the accuracy of warning and the generalization ability of the system are improved. At the same time, the system can analyze the operation data of the battery and quickly generate fault warning signals, effectively improving the safety and reliability of the battery system.

[0168] The above are only the embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification of the present invention, or directly or indirectly applied to other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. An artificial intelligence battery early warning system based on adaptive deep learning, characterized in that: include: Data acquisition module, used to collect multivariate parameter data of system operation in real time; Data preprocessing module, used to clean and normalize the collected data; Adaptive feature extraction module, which uses an adaptive algorithm to extract features related to the battery health status from the preprocessed data; Deep learning model training module, used to build deep learning models for battery health status assessment and fault prediction; The health status assessment module is used to analyze the real-time battery data using the trained model, evaluate the battery health status, and provide real-time monitoring and assessment of the battery status; Fault warning signal generation module, used to generate fault warning signals and maintenance measures according to health status assessment results; The communication module is responsible for data exchange with other system components to ensure the real-time and reliability of data transmission; The power management module provides a stable power supply for the entire system, ensuring that the system can continue to operate when the signal is interrupted.

2. An early warning method for an artificial intelligence battery system based on adaptive deep learning, characterized in that: The following steps are involved: S1. Collect battery data; S2, preprocessing the battery data; S3, Adaptive feature extraction: Extract features that are helpful in describing the state of the energy storage system from the preprocessed data, and use statistical analysis methods to extract the most representative features based on the actual operation of the battery; S4. Deep learning model training: Select a suitable deep learning model architecture, use the historical pool dataset to train the model, adjust the model parameters through cross-validation method, and optimize the model performance; S5. Battery health status assessment: Use the trained deep learning model to analyze the real-time data of the battery and assess the battery health status. The model outputs the battery health status score or failure probability. S6. Fault warning signal generation: according to the battery health status assessment result, a warning threshold is set, and a fault warning signal is generated when the battery health status is lower than the threshold.

3. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: In S1, the battery data includes the following parameters: battery voltage, battery current, battery temperature, and battery charge and discharge cycle times.

4. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: Battery data preprocessing includes the following steps: S201, cleaning the collected raw battery data to remove abnormal values ​​and noise interference; S202: Normalize the data so that the battery data falls within the input range of the deep learning model.

5. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: The characteristics that are helpful in describing the state of the energy storage system include temperature change rate, voltage fluctuation, and current peak value.

6. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: The PCA calculation formula involved in the statistical analysis method in S3 is: ① Calculate the mean and covariance matrix of the original data: Among them, x i is a data point, is the mean vector, ∑ is the covariance matrix, n represents the number of data points, and T represents the transpose of the data points; ② Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors: ∑=λv Among them, v is the eigenvector and λ is the corresponding eigenvalue; ③Select the eigenvector corresponding to the eigenvalue to form a projection matrix: P=[v1,v2,...,v k ] Among them, P is the projection matrix composed of eigenvectors. ④Data after dimensionality reduction: x′=P T x Among them, x′ is the data after dimension reduction, P T is the data of the projection matrix, and x is the original data matrix.

7. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: In S4, the historical pool data set includes normal data and fault data.

8. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: In S4, the calculation formula involved in the deep learning model architecture is: Loss function formula: Among them, y i is the true value, is the predicted value, L is the loss function, and n is the number of samples in the data set; weight formula: Among them, ω is the weight, η is the learning rate, is the gradient of the loss function with respect to the weights.

9. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2 is characterized in that: In S5, the calculation formula of the long short-term memory network involved in health status assessment is: Forget Gate: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t are the activation vectors of the forget gate, σ is the activation function, and W f is the weight matrix of the forget gate, b f is the bias vector of the forget gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment; Input Gate: i t =σ(W i ·[h t-1 ,x t ]+b i ) Among them, i t is the activation vector of the input gate, is the candidate memory cell state, σ is the activation function, W i and W c are the weight matrices of the input gate and candidate memory cell states, h i-1 is the hidden state at the previous moment, x t is the input at the current moment, tanh is the hyperbolic tangent activation function, b i and b c is the corresponding bias vector; Memory Cells: Among them, C t is the state of the memory cell at the current memory moment, f t is the activation vector of the forget gate, C t-1 is the state of the memory cell at the previous moment, is the candidate memory cell state, i t is the activation vector of the input gate; Output Gate: o t =σ(W o ·[h t-1 ,x t ]+b0) h t =o t *tanh(C t ) Among them, t is the activation vector of the output gate, C t is the state value of the memory cell, h t is the output value, W o and b0 are weight and bias parameters, σ is the activation function, h t-1 is the hidden state at the previous moment.

10. The early warning method of an artificial intelligence battery system based on adaptive deep learning according to claim 2, characterized in that: In S6, the calculation formula involved in generating the fault warning signal is: Dynamic Time Warping (DTW) distance calculation formula: DTW(Q,T)=min(min(DTW(Q[1:m-1], T[1:n-1]), DTW(Q[2:m], T[1:n-1])), min(DTW(Q[1:m-1], T[2:n]), DTW(Q[1:m], T[2:n]))) Among them, Q and T are two time series, m and n are the sequence lengths, and the DTW distance measures the similarity between two sequences, taking into account the expansion and contraction of the time axis.

Images