Battery thermal runaway early warning method, device and equipment based on multi-dimensional feature fusion

Abstract

The invention relates to the technical field of battery safety, in particular to a battery thermal runaway early warning method, device and equipment based on multi-dimensional feature fusion, and the method comprises the steps: obtaining a multi-dimensional feature signal of a current battery; preprocessing the multi-dimensional feature signals, extracting key features of each feature signal based on a preprocessing result, and obtaining current thermal runaway features based on the key features of each feature signal; based on a preset deep learning model, the current thermal runaway features are analyzed, a feature analysis output result is obtained, and the preset deep learning model is obtained by training a hybrid architecture of a long-short-term memory network and a convolutional neural network based on a preset learning rate; and obtaining a thermal runaway early warning result of the current battery according to a feature analysis output result. Therefore, the problems of detection delay, high false alarm rate, poor universality and the like caused by dependence on single battery characteristics in related technologies are solved, and the real-time performance, the accuracy and the universality of battery thermal runaway early warning are improved.

Description

Technical Field

[0001] The present invention relates to the field of battery safety technology, and in particular to a battery thermal runaway early warning method, device and equipment based on multi-dimensional feature fusion. Background Art

[0002] Lithium-ion batteries have been widely used in electric vehicles, energy storage power stations and other fields due to their high energy density and long cycle life. Lithium-ion batteries may experience thermal runaway under certain conditions, leading to serious fire and explosion accidents. Therefore, it is very necessary to provide thermal runaway warning for batteries.

[0003] In the related art, a warning threshold is usually formulated based on the change of a single characteristic (such as temperature or voltage) and the specific operating conditions of the battery to achieve a thermal runaway warning for the battery.

[0004] However, the single feature change adopted by the related technology has problems of detection delay and high false alarm rate, and is not applicable to various battery abuse conditions. Its versatility is poor and needs to be solved urgently. Summary of the invention

[0005] The present invention provides a battery thermal runaway warning method, device and equipment based on multi-dimensional feature fusion to solve the problems of detection delay, high false alarm rate and poor versatility caused by reliance on a single battery feature in the related technology, and to improve the real-time, accuracy and versatility of battery thermal runaway warning.

[0006] The first aspect of the present invention provides a battery thermal runaway warning method based on multi-dimensional feature fusion, wherein the method includes: obtaining a multi-dimensional feature signal of the current battery; preprocessing the multi-dimensional feature signal, and based on the preprocessing result, extracting the key features of each feature signal, and obtaining the current thermal runaway feature based on the key features of each feature signal; based on a preset deep learning model, analyzing the current thermal runaway feature to obtain a feature analysis output result, wherein the preset deep learning model is trained based on a hybrid architecture of a long short-term memory network and a convolutional neural network; and obtaining the thermal runaway warning result of the current battery according to the feature analysis output result.

[0007] Further, in some embodiments, obtaining the thermal runaway warning result of the current battery based on the feature analysis output result includes: determining whether the feature analysis output result is greater than a preset battery thermal runaway threshold; if the feature analysis output result is greater than the preset battery thermal runaway threshold, determining that the current battery is in an early thermal runaway state; generating a warning signal based on the early thermal runaway state, and issuing a warning based on the warning signal.

[0008] Further, in some embodiments, before determining whether the feature analysis output result is greater than the preset battery thermal runaway threshold, it further includes: obtaining the health state data, state of charge data, and environmental data of the current battery; and obtaining the preset battery thermal runaway threshold according to the health state data, the state of charge data, and the environmental data.

[0009] Further, in some embodiments, before analyzing the current thermal runaway features based on the preset deep learning model, it further includes: obtaining a battery operating state parameter database, where the battery operating state parameter database includes multiple battery features and the feature analysis output result corresponding to each battery feature; dividing the battery operating state parameter database into a training set and a test set based on a preset division ratio; training a hybrid architecture based on a long short-term memory network and a convolutional neural network using the training set based on a preset learning rate to obtain an initial deep learning model, and testing the initial deep learning model using the test set, and when the test result meets the preset test conditions, using the initial deep learning model as the preset deep learning model.

[0010] Further, in some embodiments, after testing the initial deep learning model using the test set, it further includes: if the test result does not meet the preset test conditions, adjusting the preset division ratio, adjusting the preset learning rate based on a preset adjustment strategy, and based on the new learning rate, performing the step of training the hybrid architecture based on the long short-term memory network and the convolutional neural network using the training set until the new test result meets the preset test conditions; using the deep learning model corresponding to the new test result as the preset deep learning model.

[0011] Further, in some embodiments, the multi-dimensional feature signal includes at least two of battery temperature, gas concentration, impedance, and voltage.

[0012] Further, in some embodiments, the preprocessing of the multi-dimensional feature signal includes: performing noise reduction processing, and / or filtering processing, and / or normalization processing on the multi-dimensional feature signal to obtain the preprocessing result.

[0013] According to the battery thermal runaway warning method based on multi-dimensional feature fusion provided by an embodiment of the present invention, by acquiring the multi-dimensional feature signals of the current battery, preprocessing them, extracting the key features of each feature signal, and then obtaining the current thermal runaway features. Then, based on a preset deep learning model trained by a hybrid architecture of a long short-term memory network and a convolutional neural network, the current thermal runaway features are analyzed to obtain a feature analysis output result, and based on this, a thermal runaway warning result of the current battery is obtained, which solves the problems of detection delay, high false alarm rate, and poor versatility caused by relying on a single battery feature in the related art, and improves the real-time performance, accuracy, and versatility of battery thermal runaway warning.

[0014] An embodiment of the second aspect of the present invention provides a battery thermal runaway warning device based on multi-dimensional feature fusion. Wherein, the device includes: an acquisition module, configured to acquire multi-dimensional feature signals of the current battery; a data processing module, configured to preprocess the multi-dimensional feature signals, and based on the preprocessing result, extract the key features of each feature signal, and obtain the current thermal runaway features based on the key features of each feature signal; an artificial intelligence module, configured to analyze the current thermal runaway features based on a preset deep learning model to obtain a feature analysis output result, wherein the preset deep learning model is trained based on a hybrid architecture of a long short-term memory network and a convolutional neural network; a warning module, configured to obtain a thermal runaway warning result of the current battery according to the feature analysis output result.

[0015] Further, in some embodiments, the warning module is specifically configured to: determine whether the feature analysis output result is greater than a preset battery thermal runaway threshold; if the feature analysis output result is greater than the preset battery thermal runaway threshold, determine that the current battery is in an early stage of thermal runaway; generate a warning signal based on the early stage of thermal runaway, and issue a warning based on the warning signal.

[0016] Further, in some embodiments, before determining whether the feature analysis output result is greater than the preset battery thermal runaway threshold, the artificial intelligence module is further configured to: acquire the health state data, state of charge data, and environmental data of the current battery; obtain the preset battery thermal runaway threshold according to the health state data, the state of charge data, and the environmental data.

[0017] Further, in some embodiments, before analyzing the current thermal runaway feature based on the preset deep learning model, the artificial intelligence module is further configured to: obtain a battery operating state parameter database, where the battery operating state parameter database includes multiple battery features and the corresponding feature analysis output results of each battery feature; divide the battery operating state parameter database into a training set and a test set based on a preset division ratio; train a hybrid architecture based on a long short-term memory network and a convolutional neural network using the training set based on a preset learning rate to obtain an initial deep learning model, and test the initial deep learning model using the test set, and when the test result meets the preset test conditions, use the initial deep learning model as the preset deep learning model.

[0018] Further, in some embodiments, after testing the initial deep learning model using the test set, the artificial intelligence module is further configured to: when the test result does not meet the preset test conditions, adjust the preset division ratio, adjust the preset learning rate based on a preset adjustment strategy, and based on the new learning rate, perform the step of training the hybrid architecture based on the long short-term memory network and the convolutional neural network using the training set until the new test result meets the preset test conditions; use the deep learning model corresponding to the new test result as the preset deep learning model.

[0019] Further, in some embodiments, the multi-dimensional feature signal includes at least two of battery temperature, gas concentration, impedance, and voltage.

[0020] Further, in some embodiments, the data processing module is specifically configured to: perform noise reduction processing, and / or filtering processing, and / or normalization processing on the multi-dimensional feature signal to obtain the preprocessing result.

[0021] According to the battery thermal runaway warning device based on multi-dimensional feature fusion provided by the embodiments of the present invention, by obtaining the multi-dimensional feature signal of the current battery, preprocessing it and then extracting the key features of each feature signal, the current thermal runaway feature is obtained. Then, based on a preset deep learning model trained by a hybrid architecture of a long short-term memory network and a convolutional neural network, the current thermal runaway feature is analyzed to obtain a feature analysis output result, and based on this, the thermal runaway warning result of the current battery is obtained, solving the problems such as detection delay, high false alarm rate, and poor versatility caused by relying on a single battery feature in the related art, and improving the real-time performance, accuracy, and versatility of the battery thermal runaway warning.

[0022] In a third aspect embodiment of the present invention, an electronic device is provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the program to implement the battery thermal runaway warning method based on multi-dimensional feature fusion as described in the above embodiments.

[0023] In a fourth aspect embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and the program is executed by a processor to implement the battery thermal runaway warning method based on multi-dimensional feature fusion as described in the above embodiments.

[0024] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0026] Figure 1 is a flowchart of the battery thermal runaway warning method based on multi-dimensional feature fusion according to an embodiment of the present invention;

[0027] Figure 2 is a block diagram of the battery thermal runaway warning device based on multi-dimensional feature fusion according to an embodiment of the present invention;

[0028] Figure 3 is a schematic structural diagram of the electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0030] The following describes a battery thermal runaway warning method, device, and equipment based on multi-dimensional feature fusion according to embodiments of the present invention. Aiming at the problems of detection delay, high false alarm rate, and poor versatility caused by relying on a single battery feature in the above-mentioned background technology, the present invention provides a battery thermal runaway warning method based on multi-dimensional feature fusion. By obtaining multi-dimensional feature signals of the current battery, key features of each feature signal are extracted after preprocessing, and then the current thermal runaway feature is obtained. Then, based on a preset deep learning model trained by a hybrid architecture of a long short-term memory network and a convolutional neural network, the current thermal runaway feature is analyzed to obtain a feature analysis output result, and based on this, a thermal runaway warning result of the current battery is obtained, solving the problems of detection delay, high false alarm rate, and poor versatility caused by relying on a single battery feature in the related technology, and improving the real-time performance, accuracy, and versatility of battery thermal runaway warning.

[0031] Specifically, Figure 1 FIG. is a flowchart of a battery thermal runaway warning method based on multi-dimensional feature fusion according to an embodiment of the present invention.

[0032] As Figure 1 shown, the battery thermal runaway warning method based on multi-dimensional feature fusion includes the following steps:

[0033] In step S101, multi-dimensional feature signals of the current battery are obtained.

[0034] Among them, the multi-dimensional feature signals of the current battery are some measurable and perceivable signals or parameters that can reflect the specific state or characteristics of the battery. Multiple dimensions of these parameters can involve the electrical characteristics, physical characteristics, and chemical characteristics of the battery, etc.

[0035] Among them, in some embodiments, the multi-dimensional feature signals include at least two of battery temperature, gas concentration, impedance, and voltage.

[0036] For example, the acquisition of the battery temperature signal can be achieved by arranging high-precision temperature sensors at key positions on the surface and inside of the battery to monitor the temperature inside and outside the battery and its change rate in real time. The acquisition of the gas concentration signal can be realized by installing a high-sensitivity gas sensor array inside the battery module to detect the gas concentration and its change in real time. The acquisition of the impedance signal can utilize on-line AC impedance measurement technology to monitor the change of the internal resistance of the battery in real time. The acquisition of the voltage signal can be achieved by collecting the real-time voltage of the battery cell through a high-precision voltage sensor and analyzing the voltage change trend in combination with the charge and discharge current data.

[0037] It should be noted that the above multi-dimensional feature signal acquisition methods are only exemplary and do not limit the present invention. Those skilled in the art can obtain them according to the actual situation. To avoid redundancy, no detailed description is given here.

[0038] In step S102, the multi-dimensional feature signals are preprocessed. Based on the preprocessing results, the key features of each feature signal are extracted, and the current thermal runaway feature is obtained based on the key features of each feature signal.

[0039] Among them, preprocessing the multi-dimensional feature signals is a process of cleaning and transforming the collected raw data before feature extraction to improve data quality; the key feature of each feature signal is the feature that reflects the change characteristics of each feature signal, and the current thermal runaway feature is the feature vector used for subsequent prediction obtained according to the key features of each feature signal.

[0040] Among them, in some embodiments, preprocessing the multi-dimensional feature signals includes: performing noise reduction processing, and / or filtering processing, and / or normalization processing on the multi-dimensional feature signals to obtain the preprocessing results.

[0041] As a possible implementation manner, the key features of each feature signal may include the temperature rise rate of the battery, the gas concentration change rate, the inflection point of the impedance change curve, and the voltage mutation point, etc., which can reflect the trend of battery change. The (principal components analysis, PCA) principal component analysis method is used to reduce the dimension of the key features of each feature signal, and the most representative feature vector is extracted, that is, the current thermal runaway feature.

[0042] For example, the temperature rise rate of the battery can be obtained by calculating the first derivative of the temperature time series through a sliding window The gas concentration change rate can be calculated by using the difference method to calculate the gas concentration change at adjacent time points The inflection point of the impedance change curve is the local extreme point of the impedance curve detected based on the second derivative method The voltage mutation point is detected by the threshold method for the instantaneous change amplitude of the voltage (|V t+1 -V t |>α*V, where α is a preset coefficient).

[0043] Further, the basis for the most representative feature vector of the PCA method is the cumulative variance contribution rate. For example, the feature vector with the first k principal component contribution rates ≥ 95% can be regarded as the most representative feature vector. The specific formula is:

[0044]

[0045] Among them, Vi is the eigenvalue of the covariance matrix, and n is the feature dimension of each feature signal.

[0046] In step S103, based on a preset deep learning model, the current thermal runaway characteristics are analyzed to obtain a feature analysis output result. The preset deep learning model is obtained by training a hybrid architecture of a long short-term memory network and a convolutional neural network based on a preset learning rate.

[0047] Among them, the preset deep learning model is a deep learning model obtained by training with a pre-acquired battery operating state parameter database, and the feature analysis output result is a scoring result output by the deep learning model.

[0048] For example, the feature analysis output result is a normalized thermal runaway risk score, and the scoring result is a probability value between 0 and 1, which is used for thermal runaway risk judgment. The scoring calculation is based on the Softmax function (multi-classification) or the Sigmoid function (binary classification).

[0049] Furthermore, in some embodiments, before analyzing the current thermal runaway characteristics based on the preset deep learning model, it further includes: obtaining a battery operating state parameter database, where the battery operating state parameter database includes multiple battery characteristics and the feature analysis output result corresponding to each battery characteristic; dividing the battery operating state parameter database into a training set and a test set based on a preset division ratio; training a hybrid architecture based on a long short-term memory network and a convolutional neural network using the training set based on a preset learning rate to obtain an initial deep learning model, and testing the initial deep learning model using the test set. When the test result meets the preset test conditions, the initial deep learning model is used as the preset deep learning model.

[0050] Among them, the battery operating state parameter database is a pre-prepared data set with labels. The multiple battery characteristics and the feature analysis output result corresponding to each battery characteristic are experimental results of the battery under different abuse conditions obtained through a large number of experiments, including the characteristics of the battery under normal conditions and the thermal runaway characteristics of the battery under abuse conditions. The preset division ratio is the ratio for dividing the battery operating state parameter database into a training set and a test set, and the preset learning rate is the initial learning rate when using the training set to train the hybrid architecture based on the long short-term memory network and the convolutional neural network.

[0051] Furthermore, in some embodiments, after testing the initial deep learning model using the test set, it further includes: if the test result does not meet the preset test conditions, adjusting the preset division ratio, adjusting the preset learning rate based on a preset adjustment strategy, and based on the new learning rate, performing the step of training the hybrid architecture based on the long short-term memory network and the convolutional neural network using the training set until the new test result meets the preset test conditions; using the deep learning model corresponding to the new test result as the preset deep learning model.

[0052] Specifically, the Long Short-Term Memory (LSTM) network is a type of recurrent neural network suitable for processing and predicting important events in time series. The Convolutional Neural Networks (CNN) is a type of neural network with a deep structure that includes convolutional computations and is suitable for learning and processing features in data. Therefore, a deep learning model based on the LSTM network and the CNN can fully extract and process features and obtain the output results of feature analysis.

[0053] For example, the LSTM contains 1 layer of LSTM units (with 128 hidden units and a time step of 10) for capturing long-term dependencies in the time series, and the CNN contains 2 convolutional layers (with a convolutional kernel size of 3×3, a stride of 1, and the ReLU activation function) for extracting local spatio-temporal features.

[0054] Furthermore, to enable the deep learning model to have better generalization, the embodiments of the present invention also adopt transfer learning technology to transfer the preset deep learning model parameters to the deep learning model corresponding to a specific scenario, enabling the battery thermal runaway warning method based on multi-dimensional feature fusion in the embodiments of the present invention to be applicable to different types of lithium-ion batteries and their application scenarios.

[0055] For example, a pre-trained LSTM-CNN model is used. By freezing the parameters of the first few layers and fine-tuning the last two layers to adapt to the data distribution of the new battery type, the parameters are fine-tuned using the target data set. Among them, the learning rate is reduced to 1 / 10 of the original value for model training and parameter adjustment.

[0056] In step S104, the thermal runaway warning result of the current battery is obtained according to the output result of feature analysis.

[0057] Among them, the thermal runaway warning result of the current battery is the result of judging whether the battery will undergo thermal runaway according to the output result of feature analysis based on a preset warning threshold and giving corresponding warning information.

[0058] Among them, in some embodiments, obtaining the thermal runaway warning result of the current battery according to the output result of feature analysis includes: judging whether the output result of feature analysis is greater than the preset battery thermal runaway threshold; if the output result of feature analysis is greater than the preset battery thermal runaway threshold, it is determined that the current battery is in the early stage of thermal runaway; a warning signal is generated based on the early stage of thermal runaway, and a warning is issued based on the warning signal.

[0059] For example, the range of the feature analysis output result is 0 to 1. For example, the feature analysis output result can be 0.85 or can be 0.3. The preset battery thermal runaway threshold is 0.85. When the score exceeds the threshold, a warning is triggered. If the feature analysis output result is 0.9, at this time the feature analysis output result is greater than the preset battery thermal runaway threshold, it is determined that the current battery is in the early stage of thermal runaway, a warning signal is generated based on the early stage of thermal runaway, and a warning is carried out based on the warning signal; if the feature analysis output result is 0.3, at this time the feature analysis output result is less than the preset battery thermal runaway threshold, it is determined that the current battery is not in the early stage of thermal runaway, and no warning is given.

[0060] The buzzer built in the battery system can be used for buzzer warning, or the lighting device built in the battery system can be used for lighting warning, or the upper computer display corresponding to the battery system can be used for warning, which will not be elaborated here.

[0061] It should be noted that after the warning is given, the risk of battery accidents can also be reduced by linking with the battery management system, such as taking corresponding measures such as cutting off the charging circuit, starting the cooling system, and isolating the faulty battery.

[0062] Thus, by obtaining the multi-dimensional feature signals of the current battery, extracting the key features of each feature signal after preprocessing, and then obtaining the current thermal runaway feature. Then, based on the preset deep learning model, the current thermal runaway feature is analyzed to obtain the feature analysis output result, and based on this, the thermal runaway warning result of the current battery is obtained, realizing the warning of battery thermal runaway faults.

[0063] Furthermore, in some embodiments, before determining whether the feature analysis output result is greater than the preset battery thermal runaway threshold, it further includes: obtaining the health state data, state of charge data, and environmental data of the current battery; obtaining the preset battery thermal runaway threshold according to the health state data, state of charge data, and environmental data.

[0064] Among them, the health state data of the current battery is the State of Health (SOH) of the battery, which is an important indicator for evaluating the battery performance and reflects the performance degradation of the battery relative to its initial state. The state of charge data is the State of Charge (SOC) of the battery, which is a proportional value used to describe the current remaining power of the battery relative to its total capacity. The environmental data is the relevant parameters of the environment where the current battery is located, including the temperature and humidity of the environment where the battery is located, etc. The preset battery thermal runaway threshold is an adaptive dynamic warning threshold calculated according to the health state data of the current battery through an adaptive dynamic warning threshold.

[0065] For example, the calculation formula of the adaptive dynamic warning threshold is:

[0066]

[0067] Among them, T adjusted is the result of the adaptive dynamic warning threshold, and T base is the reference threshold, which is determined by experimental data under standard conditions (i.e., SOH = 100%, SOC = 50%, ΔT = T - T1, H = H1). α is the SOH decrease adjustment coefficient, β is the SOC increase adjustment coefficient, γ is the temperature deviation from the reference value adjustment coefficient, and δ is the humidity change impact intensity adjustment coefficient for the threshold. Through experimental calibration, T1 is the reference ambient temperature, such as 25°C, and H1 is the reference ambient humidity, such as 50%RH.

[0068] Specifically, the calibration of the SOH decrease adjustment coefficient α is achieved through an accelerated aging experiment, such as a 1C charge-discharge cycle. First, the battery SOH is reduced to 80%, 60%, and 40%. Under each stage of SOH, overcharge and high-temperature abuse conditions are applied, and the thermal runaway occurrence time is recorded. It is calibrated that α = -0.2. Among them, for every 10% decrease in SOH, the threshold decreases by 2%; the calibration of the SOC increase adjustment coefficient β is achieved by conducting high-temperature (60°C) overcharge tests at different SOCs, such as 20%, 50%, 80%, and 100%. When SOC > 80%, the thermal runaway risk increases significantly. It is calibrated that β = 0.15. Among them, for every 10% increase in SOC, the threshold decreases by 1.5%; the calibration of the temperature deviation from the reference value adjustment coefficient γ is achieved by conducting thermal abuse tests at different temperatures, such as 0°C, 25°C, 40°C, and 60°C. For every 10°C increase in temperature, the thermal runaway trigger time is shortened by 50%. It is calibrated that γ = -0.05. Among them, for every 1°C increase in temperature, the exponential term weight is -0.05; the calibration of the humidity change impact intensity adjustment coefficient δ for the threshold is achieved by conducting short-circuit tests at different humidities, such as 30%RH, 50%RH, and 70%RH. At high humidity (>60%RH), the electrolyte decomposition accelerates. At this time, it is calibrated that δ = 0.1. Among them, for every 10%RH increase in humidity, the threshold decreases by 1%; the determination of the reference threshold T base is obtained by statistically analyzing 100 thermal runaway experiments under standard working conditions. Set T base = 0.85. Among them, when the output probability of the deep learning model ≥ 0.85, a warning is triggered.

[0069] Furthermore, in some embodiments, the temperature, SOC (estimated by Coulomb counting method), SOH (estimated by the capacity attenuation model), and humidity are collected every 5 seconds, and the current threshold is calculated in real time according to the formula:

[0070]

[0071] If SOH = 80%, SOC = 90%, T = 40°C, H = 60%RH, then:

[0072] T adjusted = 0.85*(1 - 0.2*0.2)*(1 + 0.15*0.9)*e -0.05*15 *(1 + 0.1*1.2)

[0073] = 0.85*0.96*1.135*0.472*1.12 ≈ 0.45

[0074] Furthermore, if the output result of the feature analysis of the deep learning model at this time is 0.6, which is greater than the current threshold of 0.45, it is determined that the current battery is in the early stage of thermal runaway. An early warning signal is generated based on the early stage of thermal runaway, and an early warning is carried out based on the early warning signal.

[0075] It should be noted that the experimental environment of the parameter calibration experiment is as follows: an NMC ternary lithium-ion battery (rated capacity 100Ah, nominal voltage 3.7V) is used. Its standard working condition is: SOH = 100% (i.e., a new battery), SOC = 50%, T1 = 25°C, H1 = 50%RH. The test equipment includes: a high-precision temperature control box, a gas sensor (detecting thermal runaway characteristic gases such as CO and H2), an impedance analyzer, and a data acquisition system. If the on-site data shows a high false alarm rate in a high-temperature environment, the absolute value of γ can be increased, for example, adjusted from -0.05 to -0.07, to enhance the temperature sensitivity.

[0076] To enable relevant technical personnel in the field to better understand the battery thermal runaway early warning method based on multi-dimensional feature fusion in the embodiments of the present invention, the following will be described in conjunction with specific embodiments.

[0077] Specifically, if the current ambient temperature is 35°C and the humidity is 70%RH, and the SOH of the battery module is 75% and the SOC is 85%, then according to the adaptive dynamic threshold calculation formula:

[0078]

[0079] The calculated early warning threshold is:

[0080] T adjusted = 0.85*(1 - 0.2*0.25)*(1 + 0.15*0.85)*e -0.05*10 *(1 + 0.1*1.4) ≈ 0.85*0.95*1.127*0.607*1.14 ≈ 0.57;

[0081] At this time, due to the high state of charge (SOC) and high temperature, the model determines that the risk is relatively high. For example, the output of the deep learning model is 0.65. Since the output value 0.65 of the deep learning model is greater than the warning threshold of 0.57, it is determined that the current battery is in the early stage of thermal runaway. An early warning signal is generated based on the early stage of thermal runaway, and an early warning is carried out based on the early warning signal. At the same time, the battery management system immediately isolates the battery module and starts the liquid cooling system to avoid accidents caused by battery thermal runaway.

[0082] According to the battery thermal runaway early warning method based on multi-dimensional feature fusion provided by the embodiment of the present invention, by obtaining the multi-dimensional feature signals of the current battery, preprocessing them, and extracting the key features of each feature signal, the current thermal runaway features are obtained. Then, based on a preset deep learning model trained by a hybrid architecture of a long short-term memory network and a convolutional neural network, the current thermal runaway features are analyzed to obtain a feature analysis output result, and based on this, the thermal runaway early warning result of the current battery is obtained, solving the problems such as detection delay, high false alarm rate, and poor versatility caused by relying on a single battery feature in the related art, and improving the real-time performance, accuracy, and versatility of battery thermal runaway early warning.

[0083] Next, a battery thermal runaway early warning device based on multi-dimensional feature fusion proposed according to the embodiment of the present invention will be described with reference to the accompanying drawings.

[0084] Figure 2 It is a block diagram of a battery thermal runaway early warning device based on multi-dimensional feature fusion provided according to the embodiment of the present invention.

[0085] As Figure 2 shown, the battery thermal runaway early warning device based on multi-dimensional feature fusion includes: an acquisition module 100, a data processing module 200, an artificial intelligence module 300, and an early warning module 400.

[0086] Among them, the acquisition module 100 is used to acquire the multi-dimensional feature signals of the current battery; the data processing module 200 is used to preprocess the multi-dimensional feature signals, and based on the preprocessing result, extract the key features of each feature signal, and obtain the current thermal runaway features based on the key features of each feature signal; the artificial intelligence module 300 is used to analyze the current thermal runaway features based on a preset deep learning model to obtain a feature analysis output result, where the preset deep learning model is trained based on a hybrid architecture of a long short-term memory network and a convolutional neural network; the early warning module 400 is used to obtain the thermal runaway early warning result of the current battery according to the feature analysis output result.

[0087] Further, in some embodiments, the warning module 400 is specifically configured to: determine whether the output result of the feature analysis is greater than a preset battery thermal runaway threshold; if the output result of the feature analysis is greater than the preset battery thermal runaway threshold, determine that the current battery is in the early stage of thermal runaway; generate a warning signal based on the early stage of thermal runaway, and issue a warning based on the warning signal.

[0088] Further, in some embodiments, before determining whether the output result of the feature analysis is greater than the preset battery thermal runaway threshold, the artificial intelligence module 300 is further configured to: obtain the health status data, state of charge data, and environmental data of the current battery; obtain the preset battery thermal runaway threshold according to the health status data, state of charge data, and environmental data.

[0089] Further, in some embodiments, before analyzing the current thermal runaway features based on a preset deep learning model, the artificial intelligence module 300 is further configured to: obtain a battery operating state parameter database, where the battery operating state parameter database includes multiple battery features and the feature analysis output results corresponding to each battery feature; divide the battery operating state parameter database into a training set and a test set based on a preset division ratio; train a hybrid architecture based on a long short-term memory network and a convolutional neural network based on the training set to obtain an initial deep learning model, and test the initial deep learning model using the test set, and when the test result meets the preset test conditions, use the initial deep learning model as the preset deep learning model.

[0090] Further, in some embodiments, after testing the initial deep learning model using the test set, the artificial intelligence module 300 is further configured to: if the test result does not meet the preset test conditions, adjust the preset division ratio, adjust the preset learning rate based on a preset adjustment strategy, and based on the new learning rate, execute the step of training the hybrid architecture based on the long short-term memory network and the convolutional neural network using the training set until the new test result meets the preset test conditions; use the deep learning model corresponding to the new test result as the preset deep learning model.

[0091] Further, in some embodiments, the multi-dimensional feature signal includes at least two of battery temperature, gas concentration, impedance, and voltage.

[0092] Further, in some embodiments, the data processing module 200 is specifically configured to: perform noise reduction processing, and / or filtering processing, and / or normalization processing on the multi-dimensional feature signal to obtain a preprocessing result.

[0093] It should be noted that the foregoing explanation of the embodiments of the battery thermal runaway warning method based on multi-dimensional feature fusion also applies to the battery thermal runaway warning device based on multi-dimensional feature fusion in this embodiment, and will not be elaborated here.

[0094] According to the battery thermal runaway warning device based on multi-dimensional feature fusion provided by an embodiment of the present invention, by acquiring the multi-dimensional feature signals of the current battery, preprocessing them, extracting the key features of each feature signal, and then obtaining the current thermal runaway features. Then, based on a preset deep learning model trained by a hybrid architecture of a long short-term memory network and a convolutional neural network, the current thermal runaway features are analyzed to obtain a feature analysis output result, and based on this, a thermal runaway warning result of the current battery is obtained, which solves the problems such as detection delay, high false alarm rate, and poor versatility caused by relying on a single battery feature in the related art, and improves the real-time performance, accuracy, and versatility of battery thermal runaway warning.

[0095] Figure 3 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. The electronic device may include:

[0096] A memory 301, a processor 302, and a computer program stored on the memory 301 and executable on the processor 302.

[0097] When the processor 302 executes the program, it implements the battery thermal runaway warning method based on multi-dimensional feature fusion provided in the above embodiment.

[0098] Further, the electronic device further includes:

[0099] A communication interface 303 for communication between the memory 301 and the processor 302.

[0100] The memory 301 is used to store a computer program executable on the processor 302.

[0101] The memory 301 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0102] If the memory 301, the processor 302, and the communication interface 303 are implemented independently, the communication interface 303, the memory 301, and the processor 302 may be connected to each other through a bus and complete communication with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 3It is represented by only one thick line, but it does not mean that there is only one bus or one type of bus.

[0103] Optionally, in a specific implementation, if the memory 301, the processor 302, and the communication interface 303 are integrated on a single chip, the memory 301, the processor 302, and the communication interface 303 can communicate with each other through an internal interface.

[0104] The processor 302 may be a central processing unit (CPU for short), or an application specific integrated circuit (ASIC for short), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0105] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the battery thermal runaway warning method based on multi-dimensional feature fusion as described above.

[0106] In the description of this specification, the descriptions with reference to the terms "an embodiment", "some embodiments", "example", "specific example", or "some examples", etc., mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without conflict, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0107] In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0108] Any process or method description, whether in a flowchart or otherwise described herein, can be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a customized logic function or process. The scope of the preferred embodiments of the present invention includes additional implementations where functions may be executed in a substantially simultaneous manner or in an order opposite to that shown or discussed, according to the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0109] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), and the like.

[0110] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried out in implementing the methods of the above embodiments can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

Claims

1. A battery thermal runaway early warning method based on multi-dimensional feature fusion, characterized in that: include: Obtain multi-dimensional characteristic signals of the current battery; Preprocessing the multidimensional characteristic signal, and based on the preprocessing result, extracting the key feature of each characteristic signal, and obtaining the current thermal runaway feature based on the key feature of each characteristic signal; Based on a preset deep learning model, the current thermal runaway feature is analyzed to obtain a feature analysis output result, wherein the preset deep learning model is obtained by training a hybrid architecture of a long short-term memory network and a convolutional neural network based on a preset learning rate; A thermal runaway warning result of the current battery is obtained according to the characteristic analysis output result.

2. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 1 is characterized in that: The step of obtaining the thermal runaway warning result of the current battery according to the characteristic analysis output result includes: Determining whether the characteristic analysis output result is greater than a preset battery thermal runaway threshold; If the characteristic analysis output result is greater than the preset battery thermal runaway threshold, it is determined that the current battery is in an early stage of thermal runaway; A warning signal is generated based on the early state of thermal runaway, and a warning is issued based on the warning signal.

3. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 2 is characterized in that: Before determining whether the characteristic analysis output result is greater than the preset battery thermal runaway threshold, the method further includes: Acquire health status data, charge status data and environmental data of the current battery; The preset battery thermal runaway threshold is obtained according to the health status data, the charge status data and the environmental data.

4. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 1 is characterized in that: Before analyzing the current thermal runaway feature based on the preset deep learning model, the method further includes: Acquire a battery operating state parameter database, wherein the battery operating state parameter database includes a plurality of battery features and a feature analysis output result corresponding to each battery feature; Based on a preset division ratio, the battery operating status parameter database is divided into a training set and a test set; Based on the preset learning rate, the training set is used to train a hybrid architecture based on a long short-term memory network and a convolutional neural network to obtain an initial deep learning model, and the test set is used to test the initial deep learning model. When the test result meets the preset test condition, the initial deep learning model is used as the preset deep learning model.

5. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 4 is characterized in that: After testing the initial deep learning model using the test set, the method further includes: If the test result does not meet the preset test condition, the preset division ratio is adjusted, and the preset learning rate is adjusted based on a preset adjustment strategy, and based on the new learning rate, the step of training the hybrid architecture based on the long short-term memory network and the convolutional neural network using the training set is executed until the new test result meets the preset test condition; The deep learning model corresponding to the new test result is used as the preset deep learning model.

6. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 1 is characterized in that: The multi-dimensional characteristic signal includes at least two of battery temperature, gas concentration, impedance and voltage.

7. The battery thermal runaway early warning method based on multi-dimensional feature fusion according to claim 1 is characterized in that: The preprocessing of the multidimensional feature signal comprises: The multi-dimensional feature signal is subjected to noise reduction processing, filtering processing, and / or normalization processing to obtain the preprocessing result.

8. A battery thermal runaway warning device based on multi-dimensional feature fusion, characterized in that: The device comprises: An acquisition module, used to acquire multi-dimensional characteristic signals of the current battery; A data processing module, used for preprocessing the multidimensional characteristic signal, extracting key features of each characteristic signal based on the preprocessing result, and obtaining the current thermal runaway feature based on the key features of each characteristic signal; An artificial intelligence module, configured to analyze the current thermal runaway characteristics based on a preset deep learning model to obtain a characteristic analysis output result, wherein the preset deep learning model is obtained by training a hybrid architecture of a long short-term memory network and a convolutional neural network based on a preset learning rate; The early warning module is used to obtain the thermal runaway early warning result of the current battery according to the characteristic analysis output result.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the battery thermal runaway warning method based on multi-dimensional feature fusion as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement a battery thermal runaway warning method based on multi-dimensional feature fusion as described in any one of claims 1 to 6.

Images