Battery system state monitoring equipment

By monitoring battery parameters in real time and combining multiple models for prediction and optimization, the problems of low battery monitoring accuracy and untimely early warning are solved, efficient battery status monitoring and early warning are achieved, and the safety and economics of the battery system are improved.

CN119064789APending Publication Date: 2024-12-03山西省能源互联网研究院 +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411130637.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing battery monitoring technology has low monitoring accuracy, inaccurate prediction, untimely warning and high operating costs, making it difficult to achieve accurate assessment and effective early warning of battery status.

Method used

The data acquisition module is used to monitor the battery parameters in real time, denoising and normalization through the data preprocessing module, combining convolutional neural network, long and short-term memory network and Gaussian process regression model for state prediction, and using multi-objective genetic algorithm to optimize system parameters and Early warning threshold, implement a multi-level early warning mechanism.

Benefits of technology

It improves battery monitoring accuracy and prediction reliability, realizes timely warning and control of operating costs, enhances battery safety and system reliability, provides flexible data acquisition frequency and parameter settings, and meets different monitoring needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119064789A_ABST
    Figure CN119064789A_ABST
Patent Text Reader

Abstract

The invention relates to the field of battery system monitoring, and discloses a battery system state monitoring device, comprising: a data acquisition module for monitoring the voltage, current, temperature and electrolyte concentration of a battery system in real time; the data preprocessing module is used for denoising and normalizing the collected data; the model prediction module is used for comprehensively analyzing and predicting the state of the battery system; the parameter optimization module is used for realizing optimal balance among accuracy, timeliness, false alarm rate and operation cost; and the early warning execution module is used for implementing a multi-level early warning mechanism based on model prediction and parameter optimization results and timely notifying a user of potential problems of the battery system. According to the method, the convolutional neural network, the long-short-term memory network and the Gaussian process regression model are integrated, so that the early warning timeliness and the operation cost control of multi-target genetic algorithm optimization are realized while the monitoring precision and the prediction reliability of the battery system are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of battery system monitoring, and in particular to a battery system state monitoring device. Background Art

[0002] With the rapid development and widespread application of battery technology, battery systems are playing an increasingly important role in the field of energy storage and power supply. However, battery systems face many challenges in actual use, such as battery performance degradation, overcharge and overdischarge, battery failure and other problems. These problems not only affect the service life and performance of the battery, but may also bring safety hazards. Therefore, real-time monitoring and status prediction of battery systems are particularly important.

[0003] Existing battery monitoring technologies mainly rely on simple parameter detection and threshold warning methods. Although they can detect abnormal battery conditions to a certain extent, their monitoring accuracy and prediction accuracy are greatly limited. These traditional methods can often only capture the surface changes in battery conditions and cannot deeply analyze the complex dynamic characteristics inside the battery. In addition, existing monitoring systems also have shortcomings in terms of the timeliness of warnings and control of false alarm rates, making it difficult to achieve accurate evaluation of battery conditions and effective warnings.

[0004] In current technologies, monitoring systems often lack the ability to integrate multiple models. A single model is difficult to cope with the variability and complexity of battery status, resulting in insufficient reliability of prediction results. At the same time, the lack of flexible adaptability to different monitoring needs and application scenarios limits the system's versatility and user experience. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a battery system status monitoring device, which solves the problems of low monitoring accuracy, inaccurate prediction, untimely warning and high operating cost in traditional battery monitoring technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A battery system status monitoring device, comprising:

[0007] The data acquisition module is equipped with various types of sensors to monitor the voltage, current, temperature and electrolyte concentration parameters of the battery system in real time;

[0008] A data preprocessing module, configured to perform denoising and normalization processing on the collected data to improve the accuracy and reliability of the data;

[0009] Model prediction module, which integrates convolutional neural network, long short-term memory network and Gaussian process regression technology to comprehensively analyze and predict the status of the battery system;

[0010] Parameter optimization module, which uses a multi-objective genetic algorithm to optimize system parameters and warning thresholds to achieve the best trade-off between accuracy, timeliness, false alarm rate and operating costs;

[0011] The early warning execution module implements a multi-level early warning mechanism based on model prediction and parameter optimization results to promptly notify users of potential problems in the battery system.

[0012] Preferably, the data acquisition module includes:

[0013] A voltage sensor for measuring the terminal voltage of the battery;

[0014] A current sensor to measure the current flowing through the battery;

[0015] Temperature sensor, used to monitor the operating temperature of the battery;

[0016] Electrolyte concentration sensor, used to measure the ion concentration of electrolyte.

[0017] Preferably, the data preprocessing module includes:

[0018] Data cleaning unit, used to delete invalid or noisy data;

[0019] Data normalization unit, used to normalize data of different dimensions to the same range;

[0020] The data denoising unit uses the moving average method and the Kalman filter to denoise the data, wherein the moving average method is used to smooth the data and the Kalman filter is used to eliminate dynamic noise;

[0021] The data storage unit is used to store the preprocessed data for subsequent analysis.

[0022] Preferably, the convolutional neural network model includes:

[0023] Three convolutional layers, each with a kernel size of 3x3 and channels of 32, 64, and 128 respectively;

[0024] Each convolutional layer is followed by a ReLU activation function and a 2x2 pooling layer;

[0025] The input layer receives the feature data output by the data preprocessing module;

[0026] The fully connected layer maps the output of the convolutional layer to a specific state space;

[0027] The output layer is used to provide the predicted value of the battery system status.

[0028] Preferably, the long short-term memory network model includes:

[0029] Two layers of LSTM, each containing 128 units, are used to capture the dependencies of battery system states over time;

[0030] The input layer receives the output of the convolutional neural network model;

[0031] The fully connected layer maps the output of the LSTM layer to the state prediction space;

[0032] The output layer is used to output the state prediction results of the battery system;

[0033] The activation function of the LSTM layer is the sigmoid function, and the activation function of the output layer is the linear function.

[0034] Preferably, the kernel function of the Gaussian process regression model is a Gaussian kernel function, and its formula is as follows:

[0035]

[0036] Among them, k is the kernel function, x i and x j are any two data points, and l is the length scale parameter.

[0037] Preferably, the parameter optimization module uses a multi-objective genetic algorithm for optimization to achieve the following optimization goals:

[0038] Accuracy f1(x): The accuracy of battery status warning, defined as Where TP represents the number of faults that are correctly warned, and FN represents the number of faults that are not predicted;

[0039] Timeliness f2(x): The response time of the battery fault warning, defined as f2(x) = - mean fault detection time, where the negative sign indicates that the optimization goal is to maximize timeliness;

[0040] False alarm rate f3(x): The ratio of false alarms, defined as Where FP represents the number of non-fault events that are incorrectly marked as faults, and TN represents the number of non-fault events that are correctly predicted;

[0041] Operating cost f4(x): the total cost of system operation and maintenance, defined as f4(x)=C, where C represents the operating cost.

[0042] Preferably, the warning execution module includes a multi-level warning mechanism, which implements different levels of warnings according to the prediction results obtained from the model design and prediction module and the parameters optimized by the parameter optimization module:

[0043] Caution level warning: issued when the system detects a slight degradation in battery performance, with the purpose of notifying the user to perform a preliminary check;

[0044] Warning level alert: issued when the system detects that the battery performance has significantly deteriorated, requiring the user to perform immediate maintenance;

[0045] Emergency level warning: Issued when the system detects that a serious battery failure may occur, and the user is advised to stop the machine immediately for inspection.

[0046] Preferably, the parameter constraints of the device include:

[0047] Voltage threshold setting range: defined as V min ≤V≤V max , where V min and V max They represent the minimum and maximum voltage thresholds acceptable to the battery system, respectively, and are used to ensure that the battery system operates within the operating voltage range and avoids overvoltage or undervoltage working conditions;

[0048] Current threshold setting range: defined as I min ≤I≤I max , where I min and I max They represent the minimum and maximum current thresholds that the battery system can accept, respectively, to prevent battery damage from excessive current or battery performance from being affected by insufficient current.

[0049] Data collection frequency: defined as f min ≤f≤f max , where f min and f max Respectively represent the minimum and maximum frequencies of data collection, ensuring the real-time and accuracy of data collection while adapting to different monitoring needs and battery status;

[0050] Budget constraint: defined as C total ≤C budget , where C total represents the actual operating cost of the system, C budget represents the predetermined operating budget, which ensures the economy and cost-effectiveness of the system.

[0051] The present invention also provides a battery system status monitoring method, comprising the following steps:

[0052] Use the data acquisition module to collect the voltage, current, temperature and electrolyte concentration parameters of the battery system;

[0053] Use the data preprocessing module to clean, normalize and denoise the collected data;

[0054] Use convolutional neural networks, long short-term memory networks, and Gaussian process regression models to predict the state of the battery system;

[0055] Use a multi-objective genetic algorithm to optimize the prediction results and adjust the warning threshold according to the optimization results;

[0056] Based on the comprehensive evaluation of real-time data and historical data, the battery system status assessment and fault warning are carried out through a multi-level early warning mechanism.

[0057] The present invention provides a battery system status monitoring device, which has the following beneficial effects:

[0058] 1. The present invention integrates convolutional neural networks, long short-term memory networks and Gaussian process regression models to improve the accuracy of battery system monitoring and the reliability of prediction, while achieving the warning timeliness and operating cost control of multi-objective genetic algorithm optimization. Specifically, it enhances battery safety and system reliability, provides a timely and effective multi-level warning mechanism, and meets different monitoring needs through flexible data acquisition frequency and parameter setting optimization, making the present invention not only economical and practical, but also highly adaptable and user-friendly. These comprehensive advantages make the present invention have significant application value in the field of battery status monitoring and management.

[0059] 2. By integrating convolutional neural networks, long short-term memory networks and Gaussian process regression models, the present invention can analyze and predict battery status from multiple dimensions, enhancing the accuracy of monitoring. This method of using multiple models in combination not only makes the prediction accurate, but also can evaluate the uncertainty of the prediction, providing users with more detailed and reliable data support.

[0060] 3. The multi-level early warning mechanism of the present invention can provide corresponding levels of early warning according to different levels of battery performance degradation, ensuring that users are notified in time to take measures. This timely feedback mechanism is crucial to preventing the development of battery failures, helping to intervene in the early stages of the problem and avoiding possible system downtime or greater damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the device structure of the present invention;

[0062] Figure 2 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] Please see attached Figure 1, an embodiment of the present invention provides a battery system status monitoring device, which monitors the voltage, current, temperature and electrolyte concentration parameters of the battery system in real time by configuring various types of sensors, denoises and normalizes the data through a data preprocessing module, and then integrates convolutional neural networks, long short-term memory networks and Gaussian process regression technology through a model prediction module to comprehensively analyze and predict the battery system status. Subsequently, the parameter optimization module uses a multi-objective genetic algorithm to optimize system parameters and warning thresholds to achieve the optimal trade-off between accuracy, timeliness, false alarm rate and operating costs. Finally, the warning execution module implements a multi-level warning mechanism based on the model prediction and parameter optimization results to promptly notify users of potential problems with the battery system.

[0065] Specifically, the equipment includes:

[0066] Data acquisition module

[0067] The data acquisition module is responsible for configuring various types of sensors to monitor the key parameters of the battery system in real time. The collected data is the basis for subsequent analysis and prediction, so the accuracy and real-time performance of this module are crucial.

[0068] In this embodiment, the data acquisition module includes the following parts:

[0069] Voltage sensor: used to measure the terminal voltage of the battery to ensure that the battery is within the safe operating range.

[0070] Current sensor: used to measure the current flowing through the battery and monitor the battery's charge and discharge status.

[0071] Temperature sensor: used to monitor the operating temperature of the battery to prevent damage caused by overheating.

[0072] Electrolyte concentration sensor: used to measure the ion concentration of the electrolyte and monitor the chemical state of the battery.

[0073] Data transmission unit: transmits the collected data to the data preprocessing module in real time to ensure the real-time and integrity of the data.

[0074] The data acquisition module monitors the key parameters of the battery system in real time through a high-precision sensor array, sampling 10 times per second and transmitting the data to the data preprocessing module for further processing. Through the combination of high-frequency sampling and multiple sensors, it can effectively capture subtle changes in the battery status and improve the monitoring accuracy and response speed of the system.

[0075] Data preprocessing module

[0076] The data preprocessing module cleans, denoises and normalizes the collected raw data to improve the accuracy and reliability of the data and provide high-quality input data for the model prediction module.

[0077] In this embodiment, the data preprocessing module includes the following functions:

[0078] Data cleaning: Delete invalid or abnormal data to ensure the validity of the data.

[0079] De-noising: Moving average method and Kalman filter are used to denoise the data. Moving average method is used to smooth data and eliminate short-term fluctuations; Kalman filter is used for dynamic denoising to adapt to the changing trend of data.

[0080] Data normalization: Standardize data of different dimensions to the same range for subsequent model processing.

[0081] The data preprocessing module first cleans the raw data, removes outliers and noise data, and then performs denoising through moving average method and Kalman filter to smooth data fluctuations and eliminate dynamic noise. Finally, the data is normalized to convert data from different sensors to a unified range to improve the accuracy and stability of subsequent model predictions.

[0082] Model prediction module

[0083] The model prediction module integrates convolutional neural network (CNN), long short-term memory network (LSTM) and Gaussian process regression (GPR) technology to comprehensively analyze and predict the status of the battery system.

[0084] In this embodiment, the model prediction module includes the following parts:

[0085] Convolutional Neural Network (CNN): It is used to extract key features from preprocessed data. The model consists of three convolutional layers, each with a kernel size of 3x3 and channels of 32, 64, and 128. Each convolutional layer is followed by a ReLU activation function and a 2x2 pooling layer.

[0086] f k =σ(W k *X+b k )

[0087] Among them, f k is the feature map after the k-th convolution, W k is the convolution kernel weight, b k is the bias, * denotes the convolution operation, and σ is the ReLU activation function.

[0088] Long Short-Term Memory (LSTM): used to process time series data and capture changes in battery status over time. The model includes two layers of LSTM, each with 128 units.

[0089] h t =σ(W ihx t +b ih +W hh h t-1 +b hh )

[0090] Among them, h t is the hidden state at time step t, x t is the input, W ih , W hh are the weight matrices from input to hidden and hidden to hidden, respectively, and b ih , b hh is the corresponding deviation.

[0091] Gaussian Process Regression (GPR): used to estimate the uncertainty of model output and provide prediction results with high confidence. The kernel function is the Gaussian kernel function:

[0092]

[0093] Among them, k is the kernel function, x i and x j are any two data points, and l is the length scale parameter.

[0094] The model prediction module makes a high-precision prediction of the battery system status through integrated analysis of the above three models. CNN is used for feature extraction, LSTM is used for time series analysis, and GPR is used for uncertainty estimation. The combination of the three improves the accuracy and robustness of the prediction.

[0095] Parameter Optimization Module

[0096] The parameter optimization module uses a multi-objective genetic algorithm (MOGA) to optimize system parameters and warning thresholds to achieve the optimal trade-off between accuracy, timeliness, false alarm rate and operating costs.

[0097] In this embodiment, the parameter optimization module includes the following parts:

[0098] Multi-objective genetic algorithm (MOGA): used to optimize multiple performance indicators, including accuracy, timeliness, false alarm rate, and operating cost. The objective function is defined as follows:

[0099] F(x)=(f1(x),f2(x),f3(x),f4(x))

[0100] in,

[0101] Accuracy, TP is a true positive example, and FN is a false negative example.

[0102] f2(x) = -mean fault detection time: timeliness, the negative sign indicates that the optimization goal is maximization.

[0103] False positive rate, FP is a false positive example, and TN is a true negative example.

[0104] f4(x)=operating cost: cost.

[0105] Genetic algorithm operations: including population initialization, fitness evaluation, selection, crossover and mutation. The optimization process is as follows:

[0106] Initialization: Randomly generate a set of initial populations, each individual represents a set of parameter settings.

[0107] Fitness evaluation: Calculate the objective function value of each individual.

[0108] Selection: Roulette wheel selection based on fitness.

[0109] Crossover: A single-point crossover operation is used to generate new individuals, and the crossover rate is 0.8.

[0110] Mutation: Randomly mutate some new individuals with a mutation rate of 0.1.

[0111] Iteration: Repeat the above steps until the maximum number of iterations or convergence condition is reached.

[0112] The parameter optimization module achieves the optimal balance of multiple performance indicators through MOGA to ensure the best performance of the system under different operating conditions.

[0113] Early warning execution module

[0114] The early warning execution module implements a multi-level early warning mechanism based on model prediction and parameter optimization results to promptly notify users of potential problems in the battery system.

[0115] In this embodiment, the early warning execution module includes the following functions:

[0116] Multi-level early warning mechanism: According to the prediction results and optimization parameters, different levels of early warning are implemented, including attention, warning and emergency levels.

[0117] Caution level warning: Issued when system performance degrades slightly and the user is advised to perform preliminary inspection.

[0118] Warning level alert: Issued when system performance degrades significantly, requiring users to perform immediate maintenance.

[0119] Emergency level warning: Issued when a serious system failure may occur, and users are advised to stop the system immediately for inspection.

[0120] Dynamically adjust warning thresholds: Comprehensively evaluate warning levels based on real-time and historical data to ensure the timeliness and accuracy of warnings.

[0121] The early warning execution module ensures that users can promptly understand potential problems in the battery system and take appropriate countermeasures to ensure the safety and stability of the system through a multi-level early warning mechanism and dynamic threshold adjustment.

[0122] Through the above modules, the battery system status monitoring device of the present invention realizes real-time monitoring and accurate prediction of the battery system status. The data acquisition module ensures the comprehensiveness and real-time nature of the data, the data preprocessing module improves the quality of the data, the model prediction module provides high-precision status prediction, the parameter optimization module achieves the optimal balance of multiple performance indicators, and the early warning execution module ensures that users can understand the potential problems of the battery system in a timely manner and take countermeasures.

[0123] In a preferred embodiment of the present invention, the model prediction module integrates a convolutional neural network (CNN) model to extract key features from preprocessed data and predict the battery system state. Through the combination of multiple convolutional layers, activation functions and pooling layers, the CNN model can effectively capture the complex features of battery data and improve the accuracy and reliability of state prediction.

[0124] The CNN model specifically includes:

[0125] 1. Three convolutional layers:

[0126] In this embodiment, the convolutional neural network (CNN) model includes three convolutional layers, each of which has a convolution kernel size of 3x3 and channels of 32, 64, and 128, respectively.

[0127] The first convolution layer: The input feature data passes through the first convolution layer, uses 32 3x3 convolution kernels for convolution operation, and outputs 32 feature maps.

[0128] f1=σ(W1*X+b1)

[0129] Among them, f1 is the output feature map of the first convolutional layer, W1 is the weight matrix of the first convolution kernel, b1 is the bias, * represents the convolution operation, and σ is the ReLU activation function.

[0130] Second convolutional layer: The output of the first convolutional layer passes through the second convolutional layer, which uses 64 3x3 convolution kernels for convolution operation and outputs 64 feature maps.

[0131] f2=σ(W2*f1+b2)

[0132] Among them, f2 is the output feature map of the second convolution layer, W2 is the weight matrix of the second convolution kernel, and b2 is the bias.

[0133] The third convolutional layer: The output of the second convolutional layer passes through the third convolutional layer, which uses 128 3x3 convolution kernels for convolution operation and outputs 128 feature maps.

[0134] f3=σ(W3*f2+b3)

[0135] Among them, f3 is the output feature map of the third convolutional layer, W3 is the weight matrix of the second convolution kernel, and b3 is the bias.

[0136] 2. Activation function and pooling layer:

[0137] Each convolutional layer is followed by a ReLU activation function and a 2x2 pooling layer for nonlinear mapping and feature dimensionality reduction.

[0138] ReLU activation function: The formula of the activation function is:

[0139] σ(x)=max(0,x)

[0140] Here, x is the input feature value.

[0141] 2x2 pooling layer: The pooling layer uses the maximum pooling operation, the formula is:

[0142]

[0143] Among them, p is the eigenvalue after pooling, and f(i,j) is the eigenvalue in the pooling area.

[0144] 3. Input layer:

[0145] In this embodiment, the input layer receives feature data output by the data preprocessing module, which has been denoised and normalized to ensure that the data input to the CNN model is of high quality and consistent.

[0146] 4. Fully connected layer:

[0147] The fully connected layer maps the output features of the convolutional layer to a specific state space for further processing and state prediction.

[0148] Fully connected layer formula:

[0149] y=W fc f pool +b fc

[0150] Among them, y is the output of the fully connected layer, W fc is the weight matrix of the fully connected layer, f pool is the output feature of the pooling layer, b fc It's a deviation.

[0151] 5. Output layer:

[0152] The output layer is used to provide the predicted value of the battery system status. By processing the output of the fully connected layer, the output layer generates the predicted result of the battery status.

[0153] Output layer formula:

[0154]

[0155] in, is the predicted value, W out is the weight matrix of the output layer, b out is the bias and σ is the activation function.

[0156] Through the convolutional neural network model with the above structure, the data characteristics of the battery system can be effectively extracted and analyzed, and accurate status prediction can be provided, providing reliable support for system monitoring and early warning.

[0157] In a preferred embodiment of the present invention, a long short-term memory network (LSTM) model is used to capture the dependency of battery system states over time. Through the two-layer LSTM structure, the model can effectively process time series data and provide accurate state prediction. The input layer receives the output from the convolutional neural network model, the fully connected layer maps the output of the LSTM layer to the state prediction space, and the final output layer is used to provide the state prediction result of the battery system.

[0158] The long short-term memory network model specifically includes:

[0159] Two-layer LSTM: In this embodiment, the long short-term memory network (LSTM) model includes two layers of LSTM, each layer contains 128 units. The LSTM layer can capture the long-term and short-term dependencies in the time series data and enhance the model's ability to perceive changes in the battery state.

[0160] 1. The first layer of LSTM:

[0161]

[0162] in, is the hidden state of the first layer LSTM at time step t, x t is the input, are the weight matrices from input to hidden and hidden to hidden, respectively. is the corresponding bias and σ is the activation function.

[0163] 2. Second layer LSTM:

[0164]

[0165] in, is the hidden state of the second layer LSTM at time step t, is the output of the first layer of LSTM, are the weight matrices from input to hidden and hidden to hidden, respectively. is the corresponding deviation.

[0166] 3. Input layer:

[0167] In this embodiment, the input layer receives output feature data from a convolutional neural network (CNN) model. These feature data have been processed by the CNN and have been extracted and converted into a higher-level representation, which is suitable as an input for the LSTM.

[0168] 4. Fully connected layer:

[0169] The fully connected layer maps the output features of the LSTM layer to the state prediction space for further processing and state prediction.

[0170] Fully connected layer formula:

[0171]

[0172] Among them, y is the output of the fully connected layer, W fc is the weight matrix of the fully connected layer, is the output feature of the second layer LSTM, b fc It's a deviation.

[0173] 4. Output layer:

[0174] The output layer is used to provide the battery system status prediction results. By processing the output of the fully connected layer, the output layer generates the battery status prediction results.

[0175] Output layer formula:

[0176]

[0177] in, is the predicted value, W out is the weight matrix of the output layer, b out is the bias and f is the linear activation function.

[0178] 5. Activation function:

[0179] In this embodiment, the activation function of the LSTM layer is a sigmoid function, which is suitable for capturing the dependency relationship of time series data:

[0180]

[0181] The activation function of the output layer is a linear function, which is suitable for outputting state prediction results of continuous values:

[0182] f(x)=x

[0183] Through the long short-term memory network model of the above structure, the time series data of the battery system can be effectively processed, accurate state prediction results can be provided, and reliable support can be provided for system monitoring and early warning.

[0184] In a preferred embodiment of the present invention, a Gaussian process regression (GPR) model is used to estimate the uncertainty of the battery system state prediction. The GPR model calculates the similarity between data points through a kernel function and uses these similarities to make predictions and uncertainty estimates. This model has significant advantages in handling nonlinear relationships and providing prediction results with high confidence.

[0185] 1. Kernel function:

[0186] In this embodiment, the kernel function of the Gaussian process regression model is a Gaussian kernel function, and its formula is as follows:

[0187]

[0188] Among them, k is the kernel function, x i and x j are any two data points, and l is the length scale parameter.

[0189] 2. The role of Gaussian kernel function:

[0190] Gaussian kernel function calculates the data point x i and x j The similarity between them. A smaller Euclidean distance will result in a larger similarity value, indicating that there is a stronger correlation between the data points. The length scale parameter l controls the decay speed of the similarity. A larger l value indicates a slower similarity decay, and a smaller l value indicates a faster similarity decay.

[0191] 3. Choice of length scale parameters:

[0192] The choice of the length scale parameter l has an important impact on the performance of the Gaussian process regression model. The optimal value of l can be determined by cross-validation or maximum likelihood estimation, thereby improving the prediction accuracy and robustness of the model.

[0193] 4. Output and application:

[0194] The Gaussian process regression model outputs the predicted value and prediction uncertainty of the battery system status. By estimating the uncertainty, it can provide the early warning execution module with a high confidence prediction result, helping users to better judge the battery status and take corresponding measures.

[0195] Through the application of Gaussian process regression model and its Gaussian kernel function, it can play an important role in processing the nonlinear relationship of battery system status and providing high-confidence prediction, providing strong support for system monitoring and early warning.

[0196] In a preferred embodiment of the present invention, the parameter optimization module uses a multi-objective genetic algorithm (MOGA) to optimize system parameters and warning thresholds. This optimization process aims to achieve the best performance of the battery status monitoring and warning system, and improve the accuracy and timeliness of the system, reduce the false alarm rate, and control the operating cost by balancing multiple objectives.

[0197] 1. Multi-Objective Genetic Algorithm (MOGA):

[0198] The parameter optimization module uses a multi-objective genetic algorithm to optimize multiple performance indicators, including accuracy, timeliness, false alarm rate, and operating cost. The optimization objective function is defined as follows:

[0199] Accuracy f1(x):

[0200] The accuracy of battery status warning is defined as:

[0201]

[0202] Among them, TP represents the number of faults that are correctly warned, and FN represents the number of faults that are not predicted;

[0203] Timeliness f2(x):

[0204] The response time of the battery failure warning is defined as:

[0205] f2(x)=-mean fault detection time

[0206] Among them, the negative sign indicates that the optimization goal is to maximize timeliness, so that the system can respond quickly and issue warnings;

[0207] False alarm rate f3(x):

[0208] The ratio of false alarms is defined as:

[0209]

[0210] Among them, FP represents the number of non-fault events that are incorrectly marked as faults, and TN represents the number of non-fault events that are correctly predicted. This objective function is used to minimize the false alarm rate of the system.

[0211] Operating cost f4(x):

[0212] The total cost of operating and maintaining the system is defined as:

[0213] f4(x)=C

[0214] Among them, C represents the operating cost. The objective function is used to control the operating cost of the system so that the system can operate economically and efficiently.

[0215] 2. Genetic algorithm operation:

[0216] The multi-objective genetic algorithm optimizes system parameters and warning thresholds through the following steps:

[0217] Initialization: Randomly generate a set of initial populations, each individual represents a set of parameter settings. The parameters of the individual include warning thresholds, sensor sampling frequencies, etc.

[0218] Fitness evaluation: Calculate the fitness value of each individual according to the defined objective function. The fitness value is used to measure the performance of the individual on various objectives.

[0219] Selection: Selection is based on fitness values, usually using a roulette wheel selection method. Individuals with higher fitness are selected to enter the next generation.

[0220] Crossover: Perform a crossover operation on the selected individuals to generate new individuals. The crossover operation can mix the characteristics of the parent individuals and improve the diversity of the population. The crossover rate is set to 0.8.

[0221] Mutation: Perform mutation operations on some new individuals to introduce random changes to increase the diversity of the population and prevent premature convergence. The mutation rate is set to 0.1.

[0222] Iteration: Repeat the above steps until a predetermined number of iterations (e.g., 100) or convergence condition (e.g., the fitness value no longer changes significantly) is reached.

[0223] 3. Optimize the application of results:

[0224] Through the optimization of multi-objective genetic algorithms, the system can achieve the best balance between accuracy, timeliness, false alarm rate and operating costs. The optimized parameters and warning thresholds will be applied to the battery status monitoring and warning system to improve the overall performance and reliability of the system.

[0225] By adopting a multi-objective genetic algorithm, the parameter optimization module can effectively balance multiple performance indicators, ensure the optimal performance of the battery status monitoring and early warning system under different operating conditions, and improve the safety and economic benefits of the system.

[0226] In a preferred embodiment of the present invention, the parameter constraints of the device include voltage threshold, current threshold, data acquisition frequency, operation budget, etc. The setting range of these parameters ensures that the battery system operates under safe and effective conditions and optimizes the real-time nature of data acquisition and the economy of the system.

[0227] Voltage threshold setting range:

[0228] Defined as V min ≤V≤V max , where V min and V max Respectively represent the minimum and maximum voltage thresholds acceptable to the battery system, which are used to ensure that the battery system operates within the operating voltage range and avoids overvoltage or undervoltage working conditions. The specific range of the voltage threshold is determined according to the battery type and application scenario, and is usually set through experiments and actual operating data.

[0229] Current threshold setting range:

[0230] Defined as I min ≤I≤I max , where I min and I max They represent the minimum and maximum current thresholds that the battery system can accept. The current threshold is used to prevent excessive current from damaging the battery or too low current from affecting battery performance. Excessive current may cause the battery to overheat, while too low current may not meet the power requirements of the device. The specific range of the current threshold is set according to the discharge capacity of the battery and the power requirements of the device to ensure safe and efficient operation of the battery.

[0231] Data collection frequency:

[0232] Defined as f min ≤f≤f max , where f min and f max Respectively represent the minimum and maximum frequency of data collection. The data collection frequency ensures the real-time and accuracy of data collection, while adapting to different monitoring needs and battery status. A higher acquisition frequency can capture more instantaneous change information, while a lower frequency helps save data storage and transmission resources. The specific range of data collection frequency is set according to the monitoring needs and the dynamic changes of battery status. The optimal frequency range is generally determined through experiments.

[0233] Budget constraint: defined as C total ≤C budget , where C total represents the actual operating cost of the system, C budget Represents a predetermined operating budget. This constraint ensures the economy and cost-effectiveness of the system, prevents operating costs from exceeding the budget, and ensures the sustainable operation of the system. The specific value of the operating budget is determined based on factors such as the scale of the system, operating time, and maintenance costs, and is usually derived from financial planning and cost analysis.

[0234] By setting the above parameter constraints, the equipment can achieve real-time data collection and system economy while ensuring safety and effectiveness. These parameter constraints provide an important basis for the optimized operation of the system, making the battery status monitoring and early warning system more reliable and efficient.

[0235] Please see attached Figure 2 The present invention also provides a battery system status monitoring method, comprising the following steps:

[0236] S1. Use the data acquisition module to collect the voltage, current, temperature and electrolyte concentration parameters of the battery system;

[0237] S2, using the data preprocessing module to clean, normalize and denoise the collected data;

[0238] S3, predict the state of the battery system using convolutional neural network, long short-term memory network and Gaussian process regression model;

[0239] S4. Use a multi-objective genetic algorithm to optimize the prediction results and adjust the warning threshold according to the optimization results;

[0240] S5. Based on the comprehensive evaluation of real-time data and historical data, the battery system status assessment and fault warning are carried out through a multi-level early warning mechanism.

[0241] The method of this embodiment is executed based on the above-mentioned system embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0242] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A battery system status monitoring device, characterized in that: include: The data acquisition module is equipped with various types of sensors to monitor the voltage, current, temperature and electrolyte concentration parameters of the battery system in real time; A data preprocessing module, configured to perform denoising and normalization processing on the collected data to improve the accuracy and reliability of the data; Model prediction module, which integrates convolutional neural network, long short-term memory network and Gaussian process regression technology to comprehensively analyze and predict the status of the battery system; Parameter optimization module, which uses a multi-objective genetic algorithm to optimize system parameters and warning thresholds to achieve the best trade-off between accuracy, timeliness, false alarm rate and operating costs; The early warning execution module implements a multi-level early warning mechanism based on model prediction and parameter optimization results to promptly notify users of potential problems in the battery system.

2. A battery system status monitoring device according to claim 1, characterized in that: The data acquisition module comprises: A voltage sensor for measuring the terminal voltage of the battery; A current sensor to measure the current flowing through the battery; Temperature sensor, used to monitor the operating temperature of the battery; Electrolyte concentration sensor, used to measure the ion concentration of electrolyte.

3. A battery system status monitoring device according to claim 1, characterized in that: The data preprocessing module comprises: Data cleaning unit, used to delete invalid or noisy data; Data normalization unit, used to normalize data of different dimensions to the same range; The data denoising unit uses the moving average method and the Kalman filter to denoise the data, wherein the moving average method is used to smooth the data and the Kalman filter is used to eliminate dynamic noise; The data storage unit is used to store the preprocessed data for subsequent analysis.

4. A battery system status monitoring device according to claim 1, characterized in that: The convolutional neural network model includes: Three convolutional layers, each with a kernel size of 3x3 and channels of 32, 64, and 128 respectively; Each convolutional layer is followed by a ReLU activation function and a 2x2 pooling layer; The input layer receives the feature data output by the data preprocessing module; The fully connected layer maps the output of the convolutional layer to a specific state space; The output layer is used to provide the predicted value of the battery system status.

5. A battery system status monitoring device according to claim 1, characterized in that: The long short-term memory network model includes: Two layers of LSTM, each containing 128 units, are used to capture the dependencies of battery system states over time; The input layer receives the output of the convolutional neural network model; The fully connected layer maps the output of the LSTM layer to the state prediction space; The output layer is used to output the state prediction results of the battery system; The activation function of the LSTM layer is the sigmoid function, and the activation function of the output layer is the linear function.

6. A battery system status monitoring device according to claim 1, characterized in that: The kernel function of the Gaussian process regression model is a Gaussian kernel function, and its formula is as follows: Among them, k is the kernel function, x i and x j are any two data points, and l is the length scale parameter.

7. A battery system status monitoring device according to claim 1, characterized in that: The parameter optimization module adopts a multi-objective genetic algorithm for optimization to achieve the following optimization goals: Accuracy f1(x): The accuracy of battery status warning, defined as Where TP represents the number of faults that are correctly warned, and FN represents the number of faults that are not predicted; Timeliness f2(x): The response time of the battery fault warning, defined as f2(x) = - mean fault detection time, where the negative sign indicates that the optimization goal is to maximize timeliness; False alarm rate f3(x): The ratio of false alarms, defined as Where FP represents the number of non-fault events that are incorrectly marked as faults, and TN represents the number of non-fault events that are correctly predicted; Operating cost f4(x): the total cost of system operation and maintenance, defined as f4(x)=C, where C represents the operating cost.

8. A battery system status monitoring device according to claim 1, characterized in that: The warning execution module includes a multi-level warning mechanism, which implements different levels of warnings according to the prediction results obtained from the model design and prediction module and the parameters optimized by the parameter optimization module: Caution level warning: issued when the system detects a slight degradation in battery performance, with the purpose of notifying the user to perform a preliminary check; Warning level alert: issued when the system detects that the battery performance has significantly deteriorated, requiring the user to perform immediate maintenance; Emergency level warning: Issued when the system detects that a serious battery failure may occur, and the user is advised to stop the machine immediately for inspection.

9. A battery system status monitoring device according to claim 1, characterized in that: The parameter constraints of the device include: Voltage threshold setting range: defined as V min ≤V≤V max , where V min and V max They represent the minimum and maximum voltage thresholds acceptable to the battery system, respectively, and are used to ensure that the battery system operates within the operating voltage range and avoids overvoltage or undervoltage working conditions; Current threshold setting range: defined as I min ≤I≤I max , where I min and I max They represent the minimum and maximum current thresholds that the battery system can accept, respectively, to prevent battery damage from excessive current or battery performance from being affected by insufficient current. Data collection frequency: defined as f min ≤f≤f max , where f min and f max Respectively represent the minimum and maximum frequencies of data collection, ensuring the real-time and accuracy of data collection while adapting to different monitoring needs and battery status; Budget constraint: defined as C total ≤C budget , where C total represents the actual operating cost of the system, C budget represents the predetermined operating budget, which ensures the economy and cost-effectiveness of the system.

10. A method for monitoring the state of a battery system, using the device according to any one of claims 1 to 9, characterized in that: The following steps are involved: Use the data acquisition module to collect the voltage, current, temperature and electrolyte concentration parameters of the battery system; Use the data preprocessing module to clean, normalize and denoise the collected data; Use convolutional neural networks, long short-term memory networks, and Gaussian process regression models to predict the state of the battery system; Use a multi-objective genetic algorithm to optimize the prediction results and adjust the warning threshold according to the optimization results; Based on the comprehensive evaluation of real-time data and historical data, the battery system status assessment and fault warning are carried out through a multi-level early warning mechanism.

Citation Information

Cited By

  • Multi-source monitoring and dynamic regulation and control method and system for mill, terminal and medium

    CN120143720A

  • Electronic apparatus having voltage drop management mechanism and voltage drop management method

    TWI938824B