Electric vehicle battery state studying and judging method and device based on AI model

Abstract

The invention relates to the technical field of electric vehicle battery intelligent monitoring and management, and provides an electric vehicle battery state studying and judging method and device based on an AI model, and the method comprises the steps: importing a trained battery charging algorithm model; monitoring the curve change of the charging pulse in real time, carrying out analysis and feature extraction on the curve change of the charging pulse, judging whether charging is abnormal or not, and obtaining charging pulse curve data; acquiring sensor data in real time based on a multi-dimensional sensor, wherein the sensor data comprises environment and vehicle state data; charging pulse curve data and sensor data are input into a battery charging algorithm model for battery state research and judgment; and in response to the study and judgment result that the battery is abnormal, executing an emergency process, including automatic power off, sending out an alarm signal and notifying related personnel to carry out manual recovery. According to the invention, the real-time monitoring and abnormity diagnosis of the charging state of the battery can be realized, the safety and reliability of the battery use are improved, and the potential risk caused by the abnormity of the battery is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and management of electric vehicle batteries, and more specifically, to a method and device for analyzing the status of electric vehicle batteries based on an AI model. Background Art

[0002] In the field of electric vehicle battery management, real-time monitoring and accurate judgment of battery status are crucial to ensuring battery safety and extending battery life. Existing technologies mainly rely on single sensor data and simple threshold judgments, which have obvious deficiencies in the accuracy and real-time performance of battery status monitoring. For example, traditional monitoring systems may not be able to effectively identify abnormal battery conditions under complex environmental conditions, such as overcurrent, overheating, and environmental abnormalities. In addition, existing technologies often lack efficient data processing and analysis capabilities when processing large amounts of sensor data, resulting in delays and misjudgments of monitoring results.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: First, the prior art lacks in-depth analysis and feature extraction of charging pulse curve changes in battery status monitoring, resulting in the inability to accurately identify abnormal conditions in the charging process; second, when processing multidimensional sensor data, the prior art fails to effectively utilize the multivariate linear regression algorithm and the least squares method to optimize the regression coefficients, which affects the accuracy of the prediction results; third, in the construction process of the existing battery charging algorithm model, it fails to make full use of deep learning technology, especially the combination of convolutional neural networks and recurrent neural networks, resulting in the performance of the model being limited when processing complex data; finally, in the process of battery status analysis, the prior art lacks an effective emergency response mechanism and is unable to take corresponding protective measures in time when battery abnormalities occur. Summary of the invention

[0004] The present invention provides an electric vehicle battery status analysis method and device based on an AI model.

[0005] In a first aspect of the present invention, a method for determining the battery status of an electric vehicle based on an AI model is provided, comprising:

[0006] Import the trained battery charging algorithm model;

[0007] Monitor the curve change of the charging pulse in real time, analyze and extract the characteristics of the curve change of the charging pulse, determine whether there is any abnormality in charging, and obtain the charging pulse curve data;

[0008] Acquire sensor data in real time based on multi-dimensional sensors, including environment and vehicle status data, and multi-dimensional sensors include temperature, humidity and spectral sensors;

[0009] Input charging pulse curve data and sensor data into the battery charging algorithm model to judge the battery status;

[0010] In response to the determination that the battery is abnormal, an emergency process is executed, including automatic power off, issuing an alarm signal, and notifying relevant personnel to perform manual recovery.

[0011] Furthermore, the battery abnormality includes charging overcurrent, overheating and environmental abnormality.

[0012] Furthermore, analyzing and extracting features of the curve change of the charging pulse includes:

[0013] Derivative the curve change f(t) of the charging pulse to obtain the first-order derivative f′(t);

[0014] Take the second-order derivative of f′(t) to obtain the second-order derivative f″(t);

[0015] The rate of change of the charging current is analyzed based on f′(t), and the change of the rate of change is analyzed based on f″(t).

[0016] Furthermore, the method also includes

[0017] The sensor data is processed using the multivariate linear regression algorithm, as shown in formula (1):

[0018] y=β0+β1x1+β2x2+…+β n x n (1)

[0020] Among them, y is the judgment result of sensor data, x1, x2, ..., x n are the data of temperature, humidity and spectrum sensors, β0, β1, β2, …, β n is the regression coefficient.

[0021] Furthermore, the method also includes calculating the optimal regression coefficient by the least square method so that the sum of squares of errors between the predicted value and the actual value is minimized;

[0022] The optimal regression coefficient is calculated by the least squares method, as shown in formula (2):

[0023]

[0024] Among them, S(β) is the error sum of squares function, y i is the actual value of the i-th observation; is the predicted value of the i-th observation; x 1i ,x 2i ,...,x niis the value of each independent variable corresponding to the i-th observation;

[0025] Taking the partial derivative of formula (2) and setting it equal to 0, we get formula (3):

[0026]

[0027] Solve equation (3) to get the optimal regression coefficient.

[0028] Furthermore, constructing a battery charging algorithm model includes the following steps:

[0029] Collect detailed data on different types, usage levels, and brands of electric vehicle batteries under various charging conditions;

[0030] Preprocessing the detailed data, including data cleaning, denoising, and missing value processing;

[0031] The battery charging algorithm model is constructed by combining the convolutional neural network (CNN) and the recurrent neural network (RNN) in deep learning;

[0032] The battery charging algorithm model includes an input layer and a circulation layer.

[0033] Furthermore, the input layer of the battery charging algorithm model vectorizes the preprocessed data and inputs it into the network in time series arrangement; the recurrent layer of the battery charging algorithm model uses the long short-term memory network LSTM and or the gated recurrent unit GRU to process the time series data and learn the long-term dependencies of the data.

[0034] Furthermore, the internal calculation of LSTM includes:

[0035] Forget gate calculation, input gate calculation, candidate cell state calculation, cell state update, output gate calculation, and hidden state update.

[0036] Furthermore, the battery charging algorithm model adopts the adaptive moment estimation Adam optimization algorithm to update parameters, and introduces L1 and L2 regularization techniques.

[0037] In a second aspect of the present invention, a device for analyzing and judging the battery status of an electric vehicle based on an AI model is provided, comprising:

[0038] A data collection module, which is used to collect detailed data of electric vehicle batteries under various charging conditions;

[0039] Data preprocessing module, used to clean, denoise and process missing values ​​of collected data;

[0040] A model building module is used to build a battery charging algorithm model by combining the convolutional neural network (CNN) and the recurrent neural network (RNN) in deep learning;

[0041] Real-time monitoring module, used to monitor the curve changes of charging pulses;

[0042] The comprehensive judgment module is used to make comprehensive judgments based on sensor data and take corresponding actions.

[0043] According to the above-mentioned embodiments of the present invention, at least the following beneficial effects are achieved: the present invention can realize accurate judgment of the battery status of electric vehicles by adopting a trained battery charging algorithm model, combined with real-time monitoring of charging pulse curve changes and multi-dimensional sensor data. This judgment method can not only monitor abnormal conditions in the charging process in real time, such as charging overcurrent, overheating and environmental abnormalities, but also intelligently analyze the battery status through deep learning technology, thereby improving the safety and reliability of battery use. In addition, the method also includes automatically executing emergency procedures when the judgment result is battery abnormality, such as automatic power off, issuing alarm signals and notifying relevant personnel to perform manual recovery, which can further enhance the safety and response speed of the battery management system. Further, by constructing a battery charging algorithm model, the present invention uses the technology of combining convolutional neural networks and recurrent neural networks in deep learning to deeply analyze and process the data in the battery charging process. This method can not only improve the accuracy of battery status monitoring, but also optimize the parameters of the model and improve the generalization ability of the model through the adaptive moment estimation Adam optimization algorithm and L1 and L2 regularization technology. At the same time, the method also includes processing sensor data using a multivariate linear regression algorithm and calculating the optimal regression coefficient using the least squares method to minimize the error between the predicted value and the actual value. The combination of these technologies makes the present invention have significant technical advantages and practical value in monitoring and managing the battery status of electric vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0045] Figure 1 A schematic diagram of a flow chart of a method for analyzing the battery status of an electric vehicle based on an AI model provided in one embodiment of the present invention;

[0046] Figure 2 A schematic diagram of the structure of an electric vehicle battery status analysis system based on an AI model provided by an embodiment of the present invention;

[0047] Figure 3The schematic diagram schematically shows the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the technical means, creative features, objectives and effects achieved by the embodiments of the present invention easy to understand, the embodiments of the present invention are further described below in combination with diagrams and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not used to limit the embodiments of the present invention.

[0049] A specific embodiment of the present invention discloses a method for analyzing the battery status of an electric vehicle based on an AI model, comprising the following steps:

[0050] Step 101, importing a trained battery charging algorithm model;

[0051] Step 102, real-time monitoring of the curve change of the charging pulse, analyzing and extracting features of the curve change of the charging pulse, determining whether there is an abnormality in charging, and obtaining charging pulse curve data;

[0052] Step 103, acquiring sensor data in real time based on a multi-dimensional sensor, the sensor data including environment and vehicle status data, the multi-dimensional sensor including temperature, humidity and spectrum sensors;

[0053] Step 104, inputting the charging pulse curve data and the sensor data into the battery charging algorithm model to analyze the battery status;

[0054] Step 105, in response to the determination result that the battery is abnormal, executing the emergency process, including automatic power off, issuing an alarm signal, and notifying relevant personnel to perform manual recovery.

[0055] In some embodiments, the electric vehicle battery status analysis method of the present invention realizes intelligent analysis and abnormality detection of the battery charging status by combining AI models and real-time monitoring technology. This method can not only monitor the curve changes of the charging pulse in real time, but also obtain environmental and vehicle status data through multi-dimensional sensors, thereby providing comprehensive data support for battery status analysis. When the analysis results show that the battery is abnormal, the system will automatically execute the emergency process to ensure safety and timely response.

[0056] Specifically, the battery charging algorithm model can be a deep neural network that has been trained with a large amount of battery charging data and can identify normal and abnormal charging patterns. When monitoring the changes in the charging pulse curve in real time, a high-speed data acquisition system can be used to capture subtle changes in the charging process. Multidimensional sensor data includes but is not limited to temperature, humidity, spectrum, etc. The data of these sensors can be pre-processed through specific algorithms to improve the accuracy and availability of the data. For example, a temperature sensor can be set to record data once a minute, a humidity sensor can be set to record every 5 minutes, and a spectral sensor can monitor the color changes on the battery surface in real time.

[0057] Preferably, when the result of the analysis is that the battery is abnormal, the emergency process can include multiple levels of safety measures. First, the system will automatically cut off the power to prevent further damage; second, an alarm signal will be issued to alert users and relevant personnel; finally, relevant personnel will be notified through wireless communication technology to perform manual recovery. In addition, in order to improve the robustness of the system, multiple battery charging algorithm models can be set, each model focusing on different types of batteries or charging conditions to adapt to different application scenarios. In some cases, machine learning algorithms can also be used to update the model online to adapt to the impact of battery aging or environmental changes.

[0058] The battery abnormalities include charging overcurrent, overheating and environmental abnormalities.

[0059] In some embodiments, several specific situations of battery abnormality include overcurrent, overheating, and environmental abnormalities. The identification of these abnormal conditions is crucial for the safe use of batteries. Through the analysis and judgment of the AI ​​model, these potential risks can be discovered in a timely manner and corresponding preventive measures can be taken.

[0060] Specifically, charging overcurrent means that the current of the battery during charging exceeds its maximum design tolerance value, which may cause damage to the internal structure of the battery. Overheating means that the battery temperature exceeds the safety threshold, which may cause serious safety problems such as thermal runaway. Environmental anomalies may include excessive humidity or spectral anomalies, which may affect the performance and life of the battery. In order to detect these anomalies, specific thresholds can be set. For example, when the current exceeds 110% of the battery rating, the system will identify it as an overcurrent anomaly; when the battery temperature exceeds the set maximum temperature threshold, the system will identify it as an overheating anomaly.

[0061] More specifically, the system can detect overcurrent anomalies by monitoring the battery current and voltage in real time, and use temperature sensors to monitor the battery temperature to identify overheating. Detection of environmental anomalies can be achieved through humidity sensors and spectral sensors. For example, when the humidity sensor reading exceeds the set safety range, the system will identify it as an environmental humidity anomaly. In addition, a variety of algorithms can be used to improve the accuracy of anomaly detection, such as using statistical methods to determine the normal fluctuation range of current and temperature, or using machine learning models to predict and identify abnormal patterns. In some cases, wireless sensor networks can also be used to enhance data collection and transmission capabilities to achieve more extensive environmental monitoring.

[0062] In some embodiments, analyzing and extracting features from the curve variation of the charging pulse includes:

[0063] Derivative the curve change f(t) of the charging pulse to obtain the first-order derivative f′(t);

[0064] Take the second-order derivative of f′(t) to obtain the second-order derivative f″(t);

[0065] The rate of change of the charging current is analyzed based on f′(t), and the change of the rate of change is analyzed based on f″(t).

[0066] The sensor data is processed using the multivariate linear regression algorithm, as shown in formula (1):

[0067] y=β0+β1x1+β2x2+·…+β n x n (1)

[0069] Among them, y is the judgment result of sensor data, x1, x2, ..., x n are the data of temperature, humidity and spectrum sensors, β0, β1, β2, …, β n is the regression coefficient.

[0070] The optimal regression coefficient is calculated by the least square method to minimize the sum of square errors between the predicted value and the actual value;

[0071] The optimal regression coefficient is calculated by the least squares method, as shown in formula (2):

[0072]

[0073] Among them, S(β) is the error sum of squares function, y i is the actual value of the i-th observation; is the predicted value of the i-th observation; x 1i , x 2i , ..., xni is the value of each independent variable corresponding to the i-th observation;

[0074] Taking the partial derivative of formula (2) and setting it equal to 0, we get formula (3):

[0075]

[0076] Solve equation (3) to get the optimal regression coefficient.

[0077] In the present invention, the core lies in the in-depth analysis of the charging pulse curve and the comprehensive processing of multi-sensor data.

[0078] For the charging pulse curve, assume that it can be represented by a function f(t), where t is time. In order to more accurately capture the characteristics and changing trends of the curve, we not only derive the function to obtain the derivative f′(t), but also analyze its second-order derivative f″(t). The first-order derivative f′(t) reflects the rate of change of the charging current, while the second-order derivative f″(t) can reveal the change in the rate of change of the current. For example, if f′(t) continues to increase within a certain period of time and exceeds a preset threshold, it is preliminarily judged that the charging current has increased abnormally. If f″(t) is positive and large at the same time, it means that the trend of current increase is accelerating, further confirming the severity of the charging anomaly.

[0079] In order to determine the optimal regression coefficient, we first collect a large amount of charging pulse curve data under normal and abnormal charging conditions and the corresponding multi-sensor data as a training set. Then, the regression coefficient is calculated by the least squares method to minimize the sum of squared errors between the predicted value and the actual value.

[0080] In practical applications, the sensor data collected in real time is substituted into the model and the prediction results are calculated to achieve accurate judgment of the battery status.

[0081] In some embodiments, constructing a battery charging algorithm model includes the following steps:

[0082] Collect detailed data on different types, usage levels, and brands of electric vehicle batteries under various charging conditions;

[0083] Preprocessing the detailed data, including data cleaning, denoising, and missing value processing;

[0084] A battery charging algorithm model is constructed by combining a convolutional neural network (CNN) and a recurrent neural network (RNN) in deep learning; the battery charging algorithm model includes an input layer and a recurrent layer.

[0085] Specifically, the collected data may include battery voltage, current, temperature, charging time and other parameters, which can be obtained from electric vehicle batteries of different brands and different usage levels under various charging conditions such as standard charging and fast charging. Data preprocessing is a key step to ensure data quality, where data cleaning may involve removing outliers or incomplete records, denoising may use filters to reduce noise in the data, and missing value processing may use interpolation or prediction algorithms to fill in missing data points. When building a model, CNN can be used to extract local features of the battery charging curve, while RNN can capture the time series characteristics of the battery charging process.

[0086] Preferably, the design of the input layer can use standardization or normalization techniques to process time series data to adapt it to the input requirements of the model. In the design of the recurrent layer, LSTM or GRU units can be used to learn long-term dependencies in battery charging data. For example, multiple LSTM layers can be set, each with a different time step to capture features on different time scales. In addition, the training of the model can use cross-validation to improve its generalization ability, and the model performance can be optimized by adjusting hyperparameters such as the number of layers and the number of neurons of CNN and RNN. As an alternative, other types of recurrent neural networks, such as variants of long short-term memory networks (LSTM) or bidirectional recurrent neural networks, can be considered to further improve the model's prediction accuracy of battery state changes.

[0087] In some embodiments, the input layer of the battery charging algorithm model vectorizes the preprocessed data and inputs it into the network in time series arrangement; the recurrent layer of the battery charging algorithm model uses the long short-term memory network LSTM and or the gated recurrent unit GRU to process the time series data and learn the long-term dependencies of the data.

[0088] In some embodiments, the battery charging algorithm model is a working mechanism of an input layer and a recurrent layer. The input layer is responsible for converting the preprocessed data into a vector form suitable for network processing and arranging it in the order of the time series. The recurrent layer uses two advanced recurrent neural network structures, LSTM and GRU, to process time series data, thereby capturing and learning long-term dependencies in the battery charging process.

[0089] Specifically, vectorization processing may involve feature extraction and feature selection, converting raw data into a set of feature vectors that can represent key information in the battery charging process. Time series arrangement means that the data will be input into the network in the order in which they appear in time, which is essential for capturing time dependencies. LSTM and GRU are the core of the recurrent layer. They process sequence data by maintaining internal states. LSTM controls the flow of information through forget gates, input gates, candidate cell states, and output gates, while GRU implements similar functions through update gates and reset gates.

[0090] Preferably, the vectorization processing of the input layer can use principal component analysis (PCA) or autoencoder to reduce the dimension of the data while retaining the most important features. Time series data can be divided into sequences of fixed length for input into the recurrent layer. In the recurrent layer, multiple LSTM or GRU units can be set, and each unit can have different parameter settings, such as the size of the hidden layer, the learning rate, etc., to adapt to different learning tasks. In addition, in order to improve the predictive ability of the model, an attention mechanism can be introduced to enable the model to pay more attention to the key parts in the time series. As an alternative, other types of recurrent neural network structures, such as bidirectional LSTM or GRU, can be considered to capture both forward and reverse time dependencies. Different optimization algorithms, such as Adam or RMSprop, can also be explored to improve the training efficiency and performance of the model.

[0091] For example, first, a large number of detailed data of electric vehicle batteries of different types, different usage levels, and different brands under various charging conditions are collected, including but not limited to charging current I(t), charging voltage V(t), charging time t, battery temperature T b 、Ambient temperature T e , humidity H, spectrum S and other multi-dimensional data. Then, the collected data is preprocessed, including data cleaning, denoising, missing value processing and other operations to ensure the quality and reliability of the data. Then, the model is constructed by combining the convolutional neural network (CNN) and the recurrent neural network (RNN) in deep learning. In the input layer of the model, the preprocessed data is vectorized and input into the network in time series.

[0092] Assume that the input data vector is x = [x1, x2, ..., x n ]The convolution kernel of the convolution layer is w=[w1,w2,…,w m ];

[0093] Then the output y of the convolution operation is:

[0094]

[0095] in,

[0096] In some embodiments, the internal computation of the LSTM includes:

[0097] Forget gate calculation, input gate calculation, candidate cell state calculation, cell state update, output gate calculation, and hidden state update.

[0098] The internal calculation process of the LSTM unit is the key to LSTM's ability to effectively process sequence data and learn long-term dependencies. The forget gate, input gate, candidate cell state, cell state update, output gate, and hidden state update together constitute the core mechanism of LSTM.

[0099] Specifically, the forget gate is responsible for deciding which information to discard from the cell state at the previous time step, usually implemented through a sigmoid activation function. The input gate consists of two parts: a sigmoid layer that decides which values ​​will be updated, and a tanh layer that creates a new candidate cell state. The cell state is the main memory component in the LSTM cell, which is updated by combining the outputs of the forget gate and the input gate. The output gate determines the value of the next hidden state, which can be regarded as the output of the LSTM cell, which carries information about the previous time step.

[0100] Forget gate: Forget gate f t Used to control the cell state at the last moment How much information needs to be forgotten? The calculation formula is:

[0101] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0102] Among them, W f is the weight matrix of the forget gate, b f is the bias term, σ is the sigmoid function, and its output value is between 0 and 1. t-1 is the hidden state at the previous moment, x t is the input at the current moment. t When the value of is close to 0, it means that most of the cell state information at the previous moment is forgotten; when f t When the value is close to 1, it means that most of the cell state information at the previous moment is retained.

[0103] Input gate: Input gate i t It is used to control how much input information at the current moment needs to be updated to the cell state. The calculation formula is:

[0104] i t=σ(W i ·[h t-1 , x t ]+b i )

[0105] Among them, W i is the weight matrix of the input gate, b i is the bias term. t The value of is also between 0 and 1, which determines the degree of update of the input information.

[0106] Candidate cell state: Candidate cell state It is the possible cell state update value calculated based on the current input and the hidden state at the previous moment. Its calculation formula is:

[0107]

[0108] Among them, W C is the weight matrix of the candidate cell state, b C is the bias term, and the tanh function maps values ​​to between -1 and 1.

[0109] Cell state: Cell state C t The update of comprehensively considers the effects of the forget gate and the input gate. Its calculation formula is:

[0110]

[0111] In the formula, ⊙ represents element multiplication, that is, the forgetting degree of the cell state at the previous moment is controlled by the forget gate, and the updating degree of the candidate cell state is controlled by the input gate.

[0112] Output gate: Output gate o t It is used to control how much information of the cell state needs to be output as the hidden state at the current moment. The calculation formula is:

[0113] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0114] The final output is h t The calculation formula is:

[0115] h t =o t ☉tanh(C t )

[0116] In order to optimize the model, the adaptive moment estimation (Adam) optimization algorithm is used to update the model parameters. The Adam algorithm estimates m by calculating the first-order moment of the gradient. t and the second-order moment estimate vt to update the parameters θ.

[0117] First-order moment estimate: First-order moment estimate m t It is an estimate of the first-order statistics of the gradient, which is similar to the concept of momentum and helps to accelerate the convergence of gradient descent. Its calculation formula is:

[0118] m t =β1m t-1 +(1-β1)g t

[0119] Among them, β1 is the decay rate of the first-order moment estimate, m t-1 is the first-order moment estimate of the previous moment, g t is the gradient at the current moment.

[0120] Second-order moment estimate: Second-order moment estimate v t It is an estimate of the second-order statistics of the gradient, which is similar to the variance of the gradient and is used to adaptively adjust the learning rate. Its calculation formula is:

[0121]

[0122] Where β2 is the decay rate of the second-order moment estimate, v t-1 is the second moment estimate of the previous moment.

[0123] In order to correct the first-order moment estimate and the second-order moment estimate to obtain a more accurate estimate:

[0124]

[0125]

[0126] Finally, the parameter θ is updated according to the corrected first-order moment estimate and second-order moment estimate:

[0127]

[0128] Here, α is the learning rate and ∈ is a small constant to prevent the denominator from being zero.

[0129] In some embodiments, the battery charging algorithm model uses an adaptive moment estimation Adam optimization algorithm to update parameters, and introduces L1 and L2 regularization techniques.

[0130] In some embodiments, the optimization algorithm used for parameter update in the battery charging algorithm model can be an Adam optimization algorithm, and L1 and L2 regularization techniques are introduced. The combination of these techniques is intended to improve the training efficiency and generalization ability of the model while preventing overfitting.

[0131] Specifically, the Adam optimization algorithm is an adaptive learning rate optimization algorithm that combines the ideas of momentum and adaptive learning rate adjustment (AdaGrad). It adapts the learning rate of each parameter by maintaining two dynamically updated estimates. L1 regularization promotes the sparsity of the model by adding the sum of the absolute values ​​of the parameters to the loss function, while L2 regularization controls the complexity of the model by adding the sum of the squares of the parameters.

[0132] In terms of parameter settings, you can set hyperparameters such as learning rate, β1 (exponential decay rate of first-order moment estimate), β2 (exponential decay rate of second-order moment estimate) for the Adam algorithm. For L1 and L2 regularization, you need to determine the regularization coefficients λ1 and λ2, which control the weight of the regularization term in the total loss function.

[0133] For example, L1 and L2 regularization are used to prevent the model from overfitting. The L1 regularization term is The L2 regularization term is During the training process, batch normalization technology is used to accelerate the convergence of the model. Assume that the input of a layer is x = [x1, x2, ..., x m ], then the calculation of batch normalization is:

[0134]

[0135] Among them, μ is the mean, σ 2 is the variance, γ and β are learnable parameters, and ∈ is a small constant. Early Stopping is used to determine the optimal number of training rounds based on the performance of the validation set.

[0136] The above-mentioned embodiments of the present invention have the following beneficial effects: the electric vehicle battery state judgment method of the present invention can improve the accuracy of battery state monitoring by importing a trained battery charging algorithm model and monitoring the changes of the charging pulse curve in real time. This method can timely discover abnormal conditions in the charging process, such as charging overcurrent and overheating, by analyzing the changes and feature extraction of the charging pulse curve. In addition, through the environmental and vehicle state data acquired in real time by multi-dimensional sensors, the method can more comprehensively evaluate the health of the battery. When the judgment result indicates that the battery is abnormal, the system will automatically execute emergency procedures, such as automatic power off and alarm signals, thereby effectively preventing and reducing the risks caused by battery failure. Further, the application of the multivariate linear regression algorithm and the least squares method of the present invention makes the processing of sensor data more scientific and accurate. By calculating the optimal regression coefficient, the method can more accurately predict the battery state and reduce errors. When constructing the battery charging algorithm model, deep learning techniques such as convolutional neural networks and recurrent neural networks are used, so that the model can learn complex data patterns and improve the accuracy of prediction. In particular, the application of LSTM and GRU enables the model to capture long-term dependencies in time series data, thereby more effectively handling dynamic changes in the charging process. In addition, the application of adaptive moment estimation Adam optimization algorithm and L1, L2 regularization technology can further enhance the generalization ability and robustness of the model, ensuring the reliability and stability of battery status analysis.

[0137] like Figure 2 As shown, in some embodiments, an electric vehicle battery status analysis device 200 based on an AI model includes:

[0138] The data collection module 201 is used to collect detailed data of the electric vehicle battery under various charging conditions;

[0139] The data preprocessing module 202 is used to clean, remove noise and process missing values ​​of the collected data;

[0140] A model building module 203 is used to build a battery charging algorithm model by combining a convolutional neural network (CNN) and a recurrent neural network (RNN) in deep learning;

[0141] A real-time monitoring module 204 is used to monitor the curve change of the charging pulse;

[0142] The comprehensive judgment module 205 is used to make a comprehensive judgment based on the sensor data and perform corresponding operations.

[0143] It can be understood that the modules recorded in the electric vehicle battery status analysis device 200 based on the AI ​​model are similar to those in the reference Figure 1Therefore, the operations, features and beneficial effects described above for the electric vehicle battery state judgment method based on the AI ​​model are also applicable to the electric vehicle battery state judgment device 200 based on the AI ​​model and the modules contained therein, and will not be repeated here.

[0144] Reference below Figure 3 , which shows a schematic diagram of a structure 300 of an electronic device suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0145] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0146] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0147] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0148] The above descriptions are only some preferred embodiments of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention to form a technical solution.

Claims

1. A method for judging the battery status of an electric vehicle based on an AI model, characterized in that: The following steps are involved: Import the trained battery charging algorithm model; Monitor the curve change of the charging pulse in real time, analyze and extract the characteristics of the curve change of the charging pulse, determine whether there is any abnormality in charging, and obtain the charging pulse curve data; Acquire sensor data in real time based on multi-dimensional sensors, including environment and vehicle status data, and multi-dimensional sensors include temperature, humidity and spectral sensors; Input charging pulse curve data and sensor data into the battery charging algorithm model to judge the battery status; In response to the determination that the battery is abnormal, an emergency process is executed, including automatic power off, issuing an alarm signal, and notifying relevant personnel to perform manual recovery.

2. The method according to claim 1, characterized in that The battery abnormalities include charging overcurrent, overheating and environmental abnormalities.

3. The method according to claim 1, characterized in that: Analysis and feature extraction of charging pulse curve changes include: Derivative the curve change f(t) of the charging pulse to obtain the first-order derivative f′(t); Take the second-order derivative of f′(t) to obtain the second-order derivative f″(t); The rate of change of the charging current is analyzed based on f′(t), and the change of the rate of change is analyzed based on f″(t).

4. The method according to claim 1, characterized in that: The method further comprises The sensor data is processed using the multivariate linear regression algorithm, as shown in formula (1): y=β0+β1x1+β2x2+…+β n x n (1) Among them, y is the judgment result of sensor data, x1, x2, ..., x n are the data of temperature, humidity and spectrum sensors, β0, β1, β2, …, β n is the regression coefficient.

5. The method according to claim 4, characterized in that The method further comprises calculating the optimal regression coefficient by the least square method so that the sum of squares of errors between the predicted value and the actual value is minimized; The optimal regression coefficient is calculated by the least squares method, as shown in formula (2): Among them, S(β) is the error sum of squares function, y i is the actual value of the i-th observation; is the predicted value of the i-th observation; x 1i , x 2i , ..., x ni is the value of each independent variable corresponding to the i-th observation; Taking the partial derivative of formula (2) and setting it equal to 0, we get formula (3): Solve equation (3) to get the optimal regression coefficient.

6. The method according to claim 1, characterized in that Building a battery charging algorithm model includes the following steps: Collect detailed data on different types, usage levels, and brands of electric vehicle batteries under various charging conditions; Preprocessing the detailed data, including data cleaning, denoising, and missing value processing; The battery charging algorithm model is constructed by combining the convolutional neural network (CNN) and the recurrent neural network (RNN) in deep learning; The battery charging algorithm model includes an input layer and a circulation layer.

7. The method according to claim 6, characterized in that The input layer of the battery charging algorithm model vectorizes the preprocessed data and inputs it into the network in time series arrangement; the recurrent layer of the battery charging algorithm model uses the long short-term memory network LSTM and the gated recurrent unit GRU to process the time series data and learn the long-term dependencies of the data.

8. The method according to claim 7, characterized in that The internal calculation of LSTM includes: Forget gate calculation, input gate calculation, candidate cell state calculation, cell state update, output gate calculation, and hidden state update.

9. The method according to claim 8, characterized in that The battery charging algorithm model uses the adaptive moment estimation Adam optimization algorithm to update parameters, and introduces L1 and L2 regularization techniques.

10. An electric vehicle battery status analysis device based on an AI model, characterized in that: include: A data collection module, which is used to collect detailed data of electric vehicle batteries under various charging conditions; Data preprocessing module, used to clean, denoise and process missing values ​​of collected data; A model building module is used to build a battery charging algorithm model by combining the convolutional neural network (CNN) and the recurrent neural network (RNN) in deep learning; Real-time monitoring module, used to monitor the curve changes of charging pulses; The comprehensive judgment module is used to make comprehensive judgments based on sensor data and take corresponding actions.

Images