Energy storage battery thermal runaway early warning method and device, electronic equipment and storage medium

Abstract

The application provides a thermal runaway early warning method and device for an energy storage battery, electronic equipment and a storage medium. The thermal runaway early warning method comprises the following steps: obtaining impedance spectrum data of a target energy storage battery at different temperatures and SOC; preprocessing the impedance spectrum data to obtain first data; determining a Nyquist diagram of the target energy storage battery according to the first data, fitting element parameter values of the target energy storage battery, and determining an equivalent impedance model; standardizing the element parameter values; analyzing the standardized element parameter values and operation data through a temperature relationship analysis model to predict the relationship between the standardized element parameter values and the internal temperature of the target energy storage battery, and then determining the predicted internal temperature; comparing the internal temperature with a preset thermal runaway condition, and early warning the thermal runaway of the target energy storage battery according to the comparison result. The application has the beneficial effect of improving the accuracy of battery thermal runaway early warning.

Description

Technical Field

[0001] This invention relates to the field of energy storage and battery management technology, and in particular to a method, device, electronic device and storage medium for early warning of thermal runaway in energy storage batteries. Background Technology

[0002] In modern energy storage and management systems, energy storage batteries play a crucial role. However, with the expansion of battery applications and the increasing complexity of their operating conditions, battery safety and reliability issues have gradually attracted widespread attention. Batteries in these systems generate heat during charging and discharging. If this heat cannot be effectively dissipated, it may lead to an abnormal increase in the internal temperature of the battery, thereby triggering thermal runaway.

[0003] The existing technical solutions have the following problems:

[0004] (1) The battery voltage and surface temperature obtained by BMS monitoring cannot actually reflect the actual temperature inside the battery;

[0005] (2) There is a complex nonlinear relationship between the AC impedance parameter of the battery and its internal temperature;

[0006] (3) In the prediction of battery internal temperature, the short-term trend (instantaneous changes in battery) and the long-term trend (cumulative changes in temperature) are related to the early warning of thermal runaway, and the temperature cannot be accurately predicted by a single trend. Summary of the Invention

[0007] The main objective of this invention is to provide a method, device, electronic device, and storage medium for early warning of thermal runaway in energy storage batteries, thereby improving the accuracy of predicting thermal runaway risk when battery internal resistance is inconsistent.

[0008] One aspect of the present invention provides a method for early warning of thermal runaway in energy storage batteries, comprising:

[0009] Acquire impedance spectrum data of the target energy storage battery at different temperatures and different SOCs;

[0010] The impedance spectrum data is preprocessed to obtain first data, which includes the frequency, real part, and imaginary part of the impedance spectrum data.

[0011] The Nyquist plot of the target energy storage battery is determined based on the first data, and the component parameter values ​​of the target energy storage battery are fitted to obtain the fitting result. The equivalent impedance model is determined based on the fitting result and the Nyquist plot.

[0012] The component parameter values ​​are standardized to obtain standardized component parameters;

[0013] The operating data of the target energy storage battery is obtained, and the temperature relationship analysis model is used to analyze the standardized component parameter values ​​and the operating data to predict the relationship between the standardized component parameter values ​​and the internal temperature of the target energy storage battery.

[0014] The predicted internal temperature is determined based on the internal temperature relationship. The internal temperature is compared with the preset thermal runaway conditions. Based on the comparison results, an early warning is issued for the thermal runaway of the target energy storage battery.

[0015] According to the aforementioned method for early warning of thermal runaway in energy storage batteries, the Nyquist plot of the target energy storage battery is determined based on the first data, and the component parameter values ​​of the target energy storage battery are fitted to obtain a fitting result. Based on the fitting result and the Nyquist plot, an equivalent impedance model is determined, including:

[0016] The Nyquist plot of the target energy storage battery is obtained by using the real part of the impedance spectrum data as the horizontal axis and the negative value of the imaginary part as the vertical axis.

[0017] Combine different component parameter values ​​and draw the Nyquist plot for each combination.

[0018] The Nyquist plot of the target energy storage battery is compared with the Nyquist plot of the combination of component parameter values. Based on the most similar comparison result, the equivalent impedance model is determined.

[0019] According to the aforementioned energy storage battery thermal runaway early warning method, the component parameter values ​​are standardized to obtain standardized component parameter values, including:

[0020] The component parameter values ​​are processed using the z-score normalization method, which is as follows:

[0021]

[0022] Where, x * Here, x represents the normalized sample value, and x represents the sample value to be normalized. μ x is the sample mean. σ This represents the sample standard deviation.

[0023] According to the aforementioned thermal runaway early warning method for energy storage batteries, the temperature relationship analysis model is trained through the following steps:

[0024] The operating data and the standardized component parameter values ​​are used as input features, wherein the operating data includes the collected voltage value, state of charge value, current value and ambient temperature;

[0025] The input features are convolutionally processed using a convolutional block, which includes a one-dimensional convolutional neural network, two max pooling layers, and two activation layers, with the activation layers using the ReLU activation function.

[0026] Temperature features are obtained by adaptively extracting the input features after convolution operation through sliding windows and convolutional blocks. Short-term temperature trend features are then determined based on these temperature features. The convolutional blocks are connected in parallel and do not share training parameters.

[0027] Short-term temperature trend features are mapped to a multi-dimensional space, where the multi-dimensional space is used to embed location information;

[0028] The Transformer model is used to determine the position encoding using sine and cosine functions, and the position encoding vector is embedded into a multi-dimensional space;

[0029] Long-term temperature trend features are extracted by using the Transformer model based on the embedded location encoding vector.

[0030] The short-term temperature trend features and the long-term temperature trend features are integrated to obtain integrated features, wherein the integration is accomplished by neurons connected in multiple fully connected layers of the Transformer model.

[0031] Based on the integrated characteristics, a temperature relationship analysis model is obtained by using a convolutional neural network and a Transformer model to minimize the reconstruction error of the input and output data as the training objective.

[0032] According to the aforementioned method for early warning of thermal runaway in energy storage batteries, a Transformer model is used to determine the location code using sine and cosine functions, and the location code vector is embedded into a multi-dimensional space, including:

[0033] The position coding vector is determined based on the sine and cosine functions, including:

[0034] P pos,2i =sin(pos / 10000) 2i / d )

[0035] P pos,2i+1 =cos(pos / 10000) 2i / d )

[0036] Where pos = 0, 1, ..., N represents the index position of the input sample in the input sequence, N is the length of the input sequence, d represents the dimension of the position encoding vector (dimension is 256), i = 0, 1, ..., d / 2 represents the index position of the position encoding vector dimension, P pos,2i P represents the value at an even index in the positional encoding vector.pos,2i+1 The value at the odd-indexed position is then used to generate a position code for each position using sine and cosine functions:

[0037]

[0038] A certain dimension 2i or 2i+1 of the position vector at position pos+k can be represented as a linear combination of the dimensions 2i and 2i+1 of the position vectors at positions pos and k. The position vector includes relative information about the positions before and after.

[0039] According to the aforementioned method for early warning of thermal runaway in energy storage batteries, the long-term temperature trend features are extracted using a Transformer model based on the embedded location encoding vector, including:

[0040] The Transformer model includes multiple basic units, each of which includes a multi-head attention mechanism module, a residual connection module, a layer normalization module, and a fully connected network module.

[0041] The multi-head attention mechanism module performs attention calculations on the input data along different sub-feature spaces using attention heads. The attention outputs of the attention heads are then concatenated with the vector matrix to obtain the final attention output. The weight parameters of each attention head are independent. The concatenation process includes:

[0042]

[0043] Where R represents the attention output, Indicates the concatenation operation, R head The output matrix calculated by the self-attention mechanism;

[0044] The residual connection module transmits the information from the lower layer to the higher layer of the network, and reduces the situation where the gradient is zero during error backpropagation by identity mapping. The calculation process of the residual connection includes adding the output f(x) of the calculation module to the input x of the calculation module as the final result, to obtain f(x)+x. The calculation module includes a multi-head attention mechanism module and a layer normalization module.

[0045] The layer normalization module uses layer normalization to pull the distribution of the input data back to a standard normal distribution and performs normalization operations on the output of the hidden layer, including:

[0046]

[0047] Where g and b represent trainable weights, This indicates element-wise multiplication between vectors.

[0048] The data dimensionality transformation and complex feature representation learning are performed through fully connected network modules, and the same operation is performed on the vector x at each position of the input sequence:

[0049] F(x) = W2ReLU(W1x + b1) + b2

[0050] Where W1 and W2 represent parameter matrices, and b1 and b2 represent bias vectors;

[0051] The input sequence H is processed by the multi-head attention mechanism module, the layer normalization module, and the residual connection module to obtain the hidden layer state matrix. The fully connected network module, the layer normalization module, and the residual connection module obtain the long-term temperature trend characteristics of the current Transformer basic unit.

[0052] According to the aforementioned thermal runaway early warning method for energy storage batteries, a convolutional neural network and a Transformer model are used to minimize the reconstruction error of the input and output data as the training objective. Training is then performed to obtain a temperature relationship analysis model, including:

[0053] The training objective is:

[0054]

[0055] in, For input X in With output The reconstruction error, m is the input X in With output The dimension, N is the total duration, x in,i,t For input X in The element value at time t in the i-th dimension, x out,i,t For output The element value at time t in the i-th dimension.

[0056] Another aspect of the present invention provides a thermal runaway early warning device for energy storage batteries, comprising:

[0057] The first module is used to acquire impedance spectrum data of the target energy storage battery at different temperatures and different SOCs.

[0058] The second module is used to preprocess the impedance spectrum data to obtain first data, which includes the frequency, real part and imaginary part of the impedance spectrum data.

[0059] The third module is used to determine the Nyquist plot of the target energy storage battery based on the first data, and to fit the component parameter values ​​of the target energy storage battery to obtain the fitting result, and to determine the equivalent impedance model based on the fitting result and the Nyquist plot.

[0060] The fourth module is used to standardize the component parameter values ​​to obtain standardized component parameter values;

[0061] The fifth module is used to acquire the operating data of the target energy storage battery, and to analyze the standardized component parameter values ​​and the operating data through a temperature relationship analysis model to predict the relationship between the standardized component parameters and the internal temperature of the target energy storage battery.

[0062] The sixth module is used to determine the predicted internal temperature based on the internal temperature relationship, compare the internal temperature with the preset thermal runaway conditions, and provide early warning of thermal runaway of the target energy storage battery based on the comparison results.

[0063] Another aspect of the present invention provides an electronic device, including a processor and a memory;

[0064] The memory is used to store programs;

[0065] The processor executes the program to implement the method as described above.

[0066] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the methods described above.

[0067] The beneficial effects of this invention are as follows: The impedance model of the energy storage battery utilizes the battery's AC impedance parameters as partial features, combining the battery's electrochemical characteristics, temperature changes, and environmental parameters from multiple dimensions, enabling a more comprehensive and accurate prediction of the energy storage battery's state; it employs a deep learning method combining CNN and Transformer, including using CNN to extract local features and quickly capturing the battery's electrochemical changes and temperature fluctuations over a short period; by processing long-term series data using Transformer, it can extract global information about the battery's state, analyze the battery's long-term operating trends and potential thermal runaway risks, achieving comprehensive feature extraction from local to global and from short-term to long-term, thus improving the accuracy of early warning for battery thermal runaway. Attached Figure Description

[0068] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0069] Figure 1 This is a schematic diagram of the thermal runaway early warning process for energy storage batteries according to an embodiment of the present invention.

[0070] Figure 2This is a schematic diagram of the process for determining the equivalent impedance model according to an embodiment of the present invention.

[0071] Figure 3 This is a flowchart illustrating the establishment of an impedance model according to an embodiment of the present invention.

[0072] Figure 4 This is a schematic diagram of the temperature relationship analysis model generation process according to an embodiment of the present invention.

[0073] Figure 5 This is a schematic diagram illustrating the generation of the input sample set according to an embodiment of the present invention.

[0074] Figure 6 This is a schematic diagram of the CNN-Block structure according to an embodiment of the present invention.

[0075] Figure 7 This is a schematic diagram of the basic Transformer model unit in an embodiment of the present invention.

[0076] Figure 8 This is a schematic diagram of the CNN-Transformer model according to an embodiment of the present invention.

[0077] Figure 9 This is a flowchart of an energy storage battery early warning method according to an embodiment of the present invention.

[0078] Figure 10 This is a schematic diagram of an energy storage battery thermal runaway early warning device according to an embodiment of the present invention. Detailed Implementation

[0079] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings. Throughout the description, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" can be used interchangeably. Terms such as "first," "second," etc., are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features. In the following description, the consecutive reference numerals for method steps are for ease of review and understanding. Adjusting the implementation order of steps, in conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0080] refer to Figure 1 ,in Figure 1This is a schematic diagram of the thermal runaway early warning process for an energy storage battery according to an embodiment of the present invention. It includes, but is not limited to, steps S100 to S600:

[0081] S100 acquires impedance spectrum data of the target energy storage battery at different temperatures and SOCs.

[0082] S200 preprocesses the impedance spectrum data to obtain the first data, which includes the frequency, real part, and imaginary part of the impedance spectrum data.

[0083] In some embodiments, the real part represents resistance data, while the imaginary part represents reactance data.

[0084] S300, determine the Nyquist plot of the target energy storage battery based on the first data, and fit the component parameter values ​​of the target energy storage battery to obtain the fitting result, and determine the equivalent impedance model based on the fitting result and the Nyquist plot.

[0085] In some embodiments, reference Figure 2 The flowchart shown below illustrates the process for determining the equivalent impedance model, including but not limited to steps S310 to S330:

[0086] S310, using the real part of the impedance spectrum data as the horizontal axis and the negative value of the imaginary part as the vertical axis, the Nyquist plot of the target energy storage battery is obtained.

[0087] S320 combines different component parameter values ​​and plots the Nyquist plot for each combination.

[0088] S330 compares the Nyquist plot of the target energy storage battery with the Nyquist plot of the combination of component parameter values ​​to determine the equivalent impedance model based on the most similar comparison result.

[0089] In some embodiments, the time interval T is set to fit the component parameter values ​​in the model.

[0090] In some embodiments, reference Figure 3 The flowchart shown illustrates the establishment of an impedance model. The modeling rule for the equivalent impedance model is based on the degree of agreement between the Nyquist plot of the impedance spectrum plotted using calculated parameter values ​​and the Nyquist plot of the measured impedance spectrum. Generally, at different temperatures, an equivalent circuit with smaller overall error and parameter fitting errors not exceeding 100% is selected as the AC impedance model for the battery. The main steps include:

[0091] (1) Based on the obtained impedance data (first data), plot the Nyquist plot of the impedance spectrum with the real part Z as the horizontal axis and the negative imaginary part -Z' as the vertical axis.

[0092] (2) Different combinations of components result in different Nyquist plots. The Nyquist plot of the impedance spectrum of a composite component consisting of resistor R, capacitor C, and inductor L connected in series is a straight line, but the quadrants are different. The Nyquist plot of the impedance spectrum of a composite component consisting of resistor R, capacitor C, and inductor L connected in parallel is a semicircle. The number of composite components (CR) determines the number of arcs and inflection points in the impedance spectrum. The inductor determines whether there is a part less than 0 in the impedance data. The constant phase angle component Q determines whether there are non-standard arcs in the Nyquist plot of the impedance spectrum. Based on this principle, and combined with the measured impedance spectrum of the energy storage battery, the equivalent impedance model of the energy storage battery can be preliminarily obtained.

[0093] S400 standardizes the component parameter values ​​to obtain standardized component parameter values.

[0094] In some embodiments, the standardized component parameter values ​​are obtained using the z-score normalization method, including:

[0095]

[0096] Where, x * x represents the normalized sample value; x represents the sample value to be normalized; x μ x is the sample mean; σ This represents the sample standard deviation.

[0097] The S500 acquires the operating data of the target energy storage battery, and uses a temperature relationship analysis model to analyze the standardized component parameter values ​​and operating data to predict the relationship between the standardized component parameter values ​​and the internal temperature of the target energy storage battery.

[0098] In some embodiments, the temperature change of an energy storage battery during charging and discharging is affected by multiple factors, including current, ambient temperature, and battery health status. During the initial stages of rapid charging or discharging, the electrochemical reactions inside the battery intensify, and the temperature may rise rapidly. As the charging process nears completion, the current decreases, and the rate of temperature rise may slow down or even stabilize. This necessitates a predictive model that can accurately integrate short-term and long-term trends for judgment. CNNs, due to their computational characteristics such as shared weights, local connections, and pooling sampling, possess powerful mapping function learning capabilities and strong adaptive feature extraction abilities. Transformers address the long-term dependency problem of RNNs through their attention mechanism; therefore, a temperature prediction model combining CNNs and Transformers is proposed.

[0099] In some embodiments, reference Figure 4 The schematic diagram shown illustrates the process of generating the temperature relationship analysis model, which includes, but is not limited to, steps S510 to S580:

[0100] The S510 uses operating data and standardized component parameter values ​​as input features, including the acquired voltage, state of charge, current and ambient temperature.

[0101] In some embodiments, the input feature x is determined. i x i This includes, but is not limited to, component parameter values ​​from the equivalent impedance model, voltage (V), state of charge (SOC), current (I), and ambient temperature (T). The component parameter values ​​have been analyzed in detail and standardized as described above. Variables such as voltage (V) can be monitored in real time by the battery management system (BMS). These variables are preprocessed using the standardization method described in step S300, and sampling is performed according to the time interval T described in the above embodiment.

[0102] S520 uses convolutional blocks to perform convolution operations on the input features. Each convolutional block includes a one-dimensional convolutional neural network, two max pooling layers, and two activation layers. The activation layers use the ReLU activation function.

[0103] In some embodiments, a CNN-Block specifically includes three one-dimensional convolutional neural network layers, two max-pooling layers, and two activation layers. The activation layers use the ReLU activation function. The final convolutional layer converts the high-dimensional data into a one-dimensional representation, which is then directly added to the position encoding vector in the subsequent Transformer model to assign positional information to the temperature features. The formula for calculating the convolution operation is:

[0104]

[0105] in Refers to the i-th input feature in the l-th layer; Refers to the i-th output feature of the l-th layer; b refers to the nth parameter trained in the l-th convolutional kernel, and b refers to the offset.

[0106] S530 adaptively extracts temperature features from the input features processed by the convolution operation through a sliding window and convolutional blocks. Based on the temperature features, short-term temperature trend features are determined. The convolutional blocks are connected in parallel and do not share training parameters.

[0107] In some embodiments, reference Figure 5The diagram shown illustrates the generation of the input sample set. It uses a sliding window to sample the input feature X, with each sliding window forming one input sample. After obtaining all input sequence samples through multiple sliding window operations, multiple parallel convolutional blocks (CNN-Block) are set up to extract temperature features from the sliding window, achieving adaptive extraction of temperature features and learning the short-term temperature variation trend inside the energy storage battery.

[0108] S540 maps short-term temperature trend features to a multi-dimensional space, where the multi-dimensional space is used to embed location information.

[0109] In some embodiments, the CNN-Block operation obtains short-term trend features, which are then mapped to a d-dimensional space through a fully connected network as a mapping layer, facilitating the subsequent embedding of location information. The CNN-Blocks are parallel structures, meaning they do not share training parameters.

[0110] The S550 uses the Transformer model to determine the position encoding using sine and cosine functions, and embeds the position encoding vector into a multidimensional space.

[0111] In some embodiments, determining the position coding vector based on a sine function and a cosine function includes:

[0112] P pos,2i =sin(pos / 10000) 2i / d )

[0113] P pos,2i+1 =cos(pos / 10000) 2i / d )

[0114] Where pos = 0, 1, ..., N represents the index position of the input sample in the input sequence, N is the length of the input sequence, d represents the dimension of the position encoding vector (dimension is 256), i = 0, 1, ..., d / 2 represents the index position of the position encoding vector dimension, P pos,2i P represents the value at an even index in the positional encoding vector. pos,2i+1 The value at the odd-indexed position is then used to generate a position code for each position using sine and cosine functions:

[0115]

[0116] A certain dimension 2i or 2i+1 of the position vector at position pos+k can be represented as a linear combination of the dimensions 2i and 2i+1 of the position vectors at positions pos and k. The position vector includes relative information about the positions before and after.

[0117] S560 extracts long-term temperature trend features using the Transformer model based on the embedded position encoding vector.

[0118] In some embodiments, reference Figure 6 The diagram shown is of the CNN-Block structure. Figure 7 The diagram shows a basic Transformer model unit, and Figure 8 The diagram shows a CNN-Transformer model. In this embodiment of the invention, long-term temperature trend features are extracted by combining CNN and Transformer models.

[0119] In some embodiments, the short-term trend features extracted by CNN-Block are embedded with location information and then passed to the long-term trend feature extraction module for the extraction of long-term trend features. This module consists of 6 basic Transformer units, which mainly consist of multi-head attention, residual connection, layer normalization, and feed-forward network.

[0120] In some embodiments, the multi-head attention mechanism in the Transformer basic unit can capture the feature information of the sequence from different perspectives, improving the ability to mine hidden state information at various positions. The multi-head attention mechanism uses multiple sets of parameters to perform attention calculations on the input data along different sub-feature spaces. Each attention head has independent weight parameters. Finally, the attention output vector matrices of each attention head are concatenated to form the final attention output, defined as follows:

[0121]

[0122] in: Indicates the concatenation operation, R head The output matrix is ​​calculated by the self-attention mechanism.

[0123] In some embodiments, residual connections in the basic Transformer unit can reduce the attenuation of low-level information, maximizing the transmission of low-level information to higher layers of the network. Simultaneously, constructing identity mappings reduces the likelihood of zero gradients during error backpropagation, mitigating the gradient vanishing problem. The computation process of residual connections involves adding the output f(x) of the computation module to the input x of the computation module to obtain the final result f(x) + x.

[0124] Layer normalization in the basic Transformer unit pulls the distribution of input data back to a standard normal distribution, speeding up training efficiency and network convergence, and normalizing the output of the hidden layer.

[0125]

[0126] Where g and b represent trainable weights, This represents element-wise multiplication between vectors. Furthermore, the overall model employs layer normalization.

[0127] The fully connected network in the basic Transformer unit transforms the data dimension, learns complex feature representations, and improves the model's representational ability. The multi-head attention mechanism module and the fully connected network connecting two layers after layer normalization ensure that the fully connected layer performs the same operation on the vector x at each position in the input sequence.

[0128] F(x) = W2ReLU(W1x + b1) + b2

[0129] In the formula, W1 and W2 represent parameter matrices, and b1 and b2 refer to bias vectors.

[0130] The input sequence H is processed through a multi-head attention mechanism, layer normalization, and residual connections to obtain the hidden layer state matrix. Subsequently, a fully connected network, layer normalization, and residual connections are used to obtain the final output of the current Transformer basic unit.

[0131] S570 integrates short-term and long-term temperature trend features to obtain integrated features, which are accomplished by neurons connected in multiple fully connected layers of the Transformer model.

[0132] S580, based on integrated features, uses a convolutional neural network and a Transformer model to minimize the reconstruction error of the input and output data as the training objective, and performs training to obtain a temperature relationship analysis model.

[0133] In some embodiments, a trend feature fusion module is constructed to integrate short-term and long-term trend features. This module consists of a fully connected network, in which each neuron is connected to all neurons in the previous layer.

[0134] x i = f[x0,x1,...,x i-1 ]

[0135] Where f represents a non-linear mapping relationship, the trend fusion module contains three fully connected layers with 64, 32 and 1 neurons respectively.

[0136] The S600 determines the predicted internal temperature based on the internal temperature relationship, compares the internal temperature with the preset thermal runaway conditions, and provides an early warning of thermal runaway of the target energy storage battery based on the comparison results.

[0137] In some embodiments, the preset thermal runaway condition is a temperature rise rate > 1°C / s for more than 3 seconds and the occurrence of a pressure drop.

[0138] In some embodiments, the objective of the feature mining model training process based on CNN-Transformer is to minimize the reconstruction error of the input and output data, which is expressed as:

[0139]

[0140] in, For input X in With output The reconstruction error; m is the input X in With output The dimension, N is the total duration; x in,i,t For input X in The element value at time t in the i-th dimension; x out,i,t For output The element value at time t in the i-th dimension.

[0141] In some embodiments, reference Figure 9 The flowchart shown illustrates an early warning method for energy storage batteries. By establishing an impedance model of the energy storage battery and utilizing the battery's AC impedance parameters as partial features, it effectively combines the battery's electrochemical characteristics with temperature change factors, proposing a battery internal temperature prediction method that integrates CNN and Transformer models. CNN is used to extract local features from battery data, accurately capturing short-term trends such as battery electrochemical characteristics and temperature changes; while Transformer is responsible for extracting global features, considering the long-term trend of battery state and complex time-series information. This multi-dimensional feature extraction method considers both short-term temperature fluctuations and predicts long-term state evolution, improving the accuracy of early warning for thermal runaway in energy storage batteries and enabling timely identification of abnormal battery temperatures and potential runaway risks.

[0142] Figure 10 This is a diagram of an energy storage battery thermal runaway early warning and analysis device according to an embodiment of the present invention. The device includes a first module 1010, a second module 1020, a third module 1030, a fourth module 1040, a fifth module 1050, and a sixth module 1060.

[0143] The system comprises the following modules: First, it acquires impedance spectrum data of the target energy storage battery at different temperatures and SOCs; second, it preprocesses the impedance spectrum data to obtain first data, which includes the frequency, real part, and imaginary part of the impedance spectrum data; third, it determines the Nyquist plot of the target energy storage battery based on the first data, and fits the component parameter values ​​of the target energy storage battery to obtain the fitting result, and determines the equivalent impedance model based on the fitting result and the Nyquist plot; fourth, it standardizes the component parameter values ​​to obtain standardized component parameter values; fifth, it acquires the operating data of the target energy storage battery, analyzes the standardized component parameter values ​​and operating data using a temperature relationship analysis model, and predicts the relationship between the standardized component parameter values ​​and the internal temperature of the target energy storage battery; sixth, it determines the predicted internal temperature based on the internal temperature relationship, compares the internal temperature with preset thermal runaway conditions, and provides an early warning of thermal runaway of the target energy storage battery based on the comparison results.

[0144] For example, with the cooperation of the first, second, third, fourth, fifth, and sixth modules in the device, the embodiment device can implement any of the aforementioned methods for early warning of thermal runaway of energy storage batteries, namely: acquiring impedance spectrum data of the target energy storage battery at different temperatures and different SOCs; preprocessing the impedance spectrum data to obtain first data, the first data including the frequency, real part, and imaginary part of the impedance spectrum data; determining the Nyquist plot of the target energy storage battery based on the first data; fitting the component parameter values ​​of the target energy storage battery to obtain a fitting result; determining the equivalent impedance model based on the fitting result and the Nyquist plot; standardizing the component parameter values ​​to obtain standardized component parameter values; acquiring the operating data of the target energy storage battery; analyzing the standardized component parameter values ​​and operating data using a temperature relationship analysis model to predict the relationship between the standardized component parameter values ​​and the internal temperature of the target energy storage battery; determining the predicted internal temperature based on the internal temperature relationship; comparing the internal temperature with preset thermal runaway conditions; and issuing an early warning of thermal runaway of the target energy storage battery based on the comparison results. The beneficial effects of this invention are as follows: The impedance model of the energy storage battery utilizes the battery's AC impedance parameters as partial features, combining the battery's electrochemical characteristics, temperature changes, and environmental parameters from multiple dimensions, enabling a more comprehensive and accurate prediction of the energy storage battery's state; it employs a deep learning method combining CNN and Transformer, including using CNN to extract local features and quickly capturing the battery's electrochemical changes and temperature fluctuations over a short period; by processing long-term series data using Transformer, it can extract global information about the battery's state, analyze the battery's long-term operating trends and potential thermal runaway risks, achieving comprehensive feature extraction from local to global and from short-term to long-term, thus improving the accuracy of early warning for battery thermal runaway.

[0145] This invention also provides an electronic device, which includes a processor and a memory;

[0146] The memory stores the program;

[0147] The processor executes a program to perform the aforementioned thermal runaway early warning method for energy storage batteries; the electronic device has the function of carrying and running the software system for thermal runaway early warning of energy storage batteries provided in the embodiments of the present invention, such as a personal computer, minicomputer, mainframe, workstation, network or distributed computing environment, standalone or integrated computer platform, or communicating with charged particle tools or other imaging devices, etc.

[0148] This invention also provides a computer-readable storage medium storing a program that is executed by a processor to implement the energy storage battery thermal runaway early warning method described above.

[0149] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0150] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned energy storage battery thermal runaway early warning method.

[0151] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0152] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0153] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0154] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0155] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0156] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0157] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0158] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for early warning of thermal runaway in energy storage batteries, characterized in that, include: Acquire impedance spectrum data of the target energy storage battery at different temperatures and different SOCs; The impedance spectrum data is preprocessed to obtain first data, which includes the frequency, real part, and imaginary part of the impedance spectrum data. The Nyquist plot of the target energy storage battery is determined based on the first data, and the component parameter values ​​of the target energy storage battery are fitted to obtain the fitting result. The equivalent impedance model is determined based on the fitting result and the Nyquist plot. The component parameter values ​​are standardized to obtain standardized component parameter values; The operating data of the target energy storage battery is obtained, and the temperature relationship analysis model is used to analyze the standardized component parameter values ​​and the operating data to predict the relationship between the standardized component parameter values ​​and the internal temperature of the target energy storage battery. The predicted internal temperature is determined based on the internal temperature relationship. The internal temperature is compared with the preset thermal runaway conditions. Based on the comparison results, an early warning is issued for the thermal runaway of the target energy storage battery.

2. The method for early warning of thermal runaway in energy storage batteries according to claim 1, characterized in that, The steps of determining the Nyquist plot of the target energy storage battery based on the first data, and fitting the component parameter values ​​of the target energy storage battery to obtain a fitting result, and determining the equivalent impedance model based on the fitting result and the Nyquist plot, include: The Nyquist plot of the target energy storage battery is obtained by using the real part of the impedance spectrum data as the horizontal axis and the negative value of the imaginary part as the vertical axis. Combine different component parameter values ​​and draw the Nyquist plot for each combination. The Nyquist plot of the target energy storage battery is compared with the Nyquist plot of the combination of component parameter values. Based on the most similar comparison result, the equivalent impedance model is determined.

3. The method for early warning of thermal runaway in energy storage batteries according to claim 1, characterized in that, The standardization process for the component parameter values ​​to obtain standardized component parameter values ​​includes: The component parameter values ​​are processed using the z-score normalization method, which is as follows: Where, x * Here, x represents the normalized sample value, and x represents the sample value to be normalized. μ x is the sample mean. σ This represents the sample standard deviation.

4. The method for early warning of thermal runaway in energy storage batteries according to claim 1, characterized in that, The temperature relationship analysis model is trained through the following steps: The operating data and the standardized component parameter values ​​are used as input features, wherein the operating data includes the collected voltage value, state of charge value, current value and ambient temperature; The input features are convolutionally processed using a convolutional block, which includes a one-dimensional convolutional neural network, two max pooling layers, and two activation layers, with the activation layers using the ReLU activation function. Temperature features are obtained by adaptively extracting the input features after convolution operation through sliding windows and convolutional blocks. Short-term temperature trend features are then determined based on these temperature features. The convolutional blocks are connected in parallel and do not share training parameters. Short-term temperature trend features are mapped to a multi-dimensional space, where the multi-dimensional space is used to embed location information; The Transformer model is used to determine the position encoding using sine and cosine functions, and the position encoding vector is embedded into a multi-dimensional space; Long-term temperature trend features are extracted by using the Transformer model based on the embedded location encoding vector. The short-term temperature trend features and the long-term temperature trend features are integrated to obtain integrated features, wherein the integration is accomplished by neurons connected in multiple fully connected layers of the Transformer model. Based on the integrated characteristics, a temperature relationship analysis model is obtained by using a convolutional neural network and a Transformer model to minimize the reconstruction error of the input and output data as the training objective.

5. The method for early warning of thermal runaway in energy storage batteries according to claim 4, characterized in that, The method of using the Transformer model to determine the position encoding using sine and cosine functions and embedding the position encoding vector into a multi-dimensional space includes: The position coding vector is determined based on the sine and cosine functions, including: P pos,2i =sin(pos / 10000 2i / d ) P pos,2i+1 =cos(pos / 10000 2i / d ) Where pos = 0, 1, ..., N represents the index position of the input sample in the input sequence, N is the length of the input sequence, d represents the dimension of the position encoding vector (dimension is 256), i = 0, 1, ..., d / 2 represents the index position of the position encoding vector dimension, P pos,2i P represents the value at an even index in the positional encoding vector. pos,2i+1 The value at the odd-indexed position is then used to generate a position code for each position using sine and cosine functions: A certain dimension 2i or 2i+1 of the position vector at position pos+k can be represented as a linear combination of the dimensions 2i and 2i+1 of the position vectors at positions pos and k. The position vector includes relative information about the positions before and after.

6. The method for early warning of thermal runaway in energy storage batteries according to claim 4, characterized in that, The step of extracting long-term temperature trend features based on the embedded positional encoding vector using a Transformer model includes: The Transformer model includes multiple basic units, each of which includes a multi-head attention mechanism module, a residual connection module, a layer normalization module, and a fully connected network module. The multi-head attention mechanism module performs attention calculations on the input data along different sub-feature spaces using attention heads. The attention outputs of the attention heads are then concatenated with the vector matrix to obtain the final attention output. The weight parameters of each attention head are independent. The concatenation process includes: Where R represents the attention output, Indicates the concatenation operation, R head The output matrix calculated by the self-attention mechanism; The residual connection module transmits the information from the lower layer to the higher layer of the network, and reduces the situation where the gradient is zero during error backpropagation by identity mapping. The calculation process of the residual connection includes adding the output f(x) of the calculation module to the input x of the calculation module as the final result, to obtain f(x)+x. The calculation module includes a multi-head attention mechanism module and a layer normalization module. The layer normalization module uses layer normalization to pull the distribution of the input data back to a standard normal distribution and performs normalization operations on the output of the hidden layer, including: Where g and b represent trainable weights, This indicates element-wise multiplication between vectors. The data dimensionality transformation and complex feature representation learning are performed through fully connected network modules, and the same operation is performed on the vector x at each position of the input sequence: F(x) = W2 ReLU(W1x + b1) + b2 Where W1 and W2 represent parameter matrices, and b1 and b2 represent bias vectors; The input sequence H is processed by the multi-head attention mechanism module, the layer normalization module, and the residual connection module to obtain the hidden layer state matrix. The fully connected network module, the layer normalization module, and the residual connection module obtain the long-term temperature trend characteristics of the current Transformer basic unit.

7. The method for early warning of thermal runaway in energy storage batteries according to claim 4, characterized in that, The method employs convolutional neural networks and Transformer models, with the goal of minimizing the reconstruction error of the input and output data, to train and obtain a temperature relationship analysis model, including: The training objective is: in, For input X in With output The reconstruction error, m is the input X in With output The dimension, N is the total duration, x in,i,t For input X in The element value at time t in the i-th dimension, x out,i,t For output The element value at time t in the i-th dimension.

8. A thermal runaway early warning device for energy storage batteries, characterized in that, include: The first module is used to acquire impedance spectrum data of the target energy storage battery at different temperatures and different SOCs. The second module is used to preprocess the impedance spectrum data to obtain first data, which includes the frequency, real part and imaginary part of the impedance spectrum data. The third module is used to determine the Nyquist plot of the target energy storage battery based on the first data, and to fit the component parameter values ​​of the target energy storage battery to obtain the fitting result, and to determine the equivalent impedance model based on the fitting result and the Nyquist plot. The fourth module is used to standardize the component parameter values ​​to obtain standardized component parameter values; The fifth module is used to acquire the operating data of the target energy storage battery, and to analyze the standardized component parameter values ​​and the operating data through a temperature relationship analysis model to predict the relationship between the standardized component parameters and the internal temperature of the target energy storage battery. The sixth module is used to determine the predicted internal temperature based on the internal temperature relationship, compare the internal temperature with the preset thermal runaway conditions, and provide early warning of thermal runaway of the target energy storage battery based on the comparison results.

9. An electronic device, characterized in that, Including the processor and memory; The memory is used to store programs; The processor executes the program to implement the energy storage battery thermal runaway early warning method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a program, which is executed by a processor to implement the thermal runaway early warning method for energy storage batteries as described in any one of claims 1-7.

Images