Truck lithium battery fault detection method

Abstract

The invention discloses a truck lithium battery fault detection method comprising the following steps: 1, data acquisition; step 2, data processing; 3, constructing an initial confidence rule base expert system model; step 4, optimizing the initial confidence rule base expert system model through a genetic algorithm; step 5, fault detection and isolation; and step 6, performing feedback and adaptive adjustment. The system has the advantages of high safety, high reliability, low maintenance cost and capability of prolonging the service life of the battery.

Description

Technical Field

[0001] The present invention relates to the field of truck lithium battery fault detection, and particularly to a method for detecting faults in truck lithium batteries. Background Art

[0002] There are multiple battery modules in a truck lithium battery. During long-term use, these modules may malfunction, such as problems like overheating, overcharging, short-circuiting, or capacity degradation of individual battery cells. To ensure the safety and reliability of the use of truck lithium batteries and effectively detect and isolate malfunctioning battery modules, common methods in the prior art for truck lithium battery fault detection include:

[0003] The first is fault identification based on voltage monitoring and fault judgment based on abnormal temperature. These methods issue a fault alarm signal by determining whether the voltage value or temperature value of the battery module exceeds the safe range, and control the isolation device to disconnect the faulty module;

[0004] However, there are obvious defects in fault detection. Firstly, there is a lack of intelligent and adaptive adjustment capabilities. Many battery management systems still rely on fixed preset rules and thresholds, which makes it easy for the system to have situations of mis-isolation or missed isolation when dealing with the behavior of battery modules under different working conditions. For example, under different environmental temperatures or load conditions, the performance of battery modules may vary greatly, and traditional fault detection methods based on fixed thresholds cannot flexibly handle these differences, potentially missing some potential faults or wrongly isolating normal modules, thus affecting the overall performance and safety of the battery pack;

[0005] The second is that some technical solutions also introduce intelligent algorithms, such as Bayesian inference, artificial neural networks, etc., to improve the accuracy and response speed of fault diagnosis;

[0006] However, Bayesian inference has limitations in dealing with uncertain information; the artificial neural network solution belongs to a black box model and has a slow convergence speed when dealing with complex data.

[0007] Therefore, it is necessary to further improve the intelligent level and integrate adaptive adjustment capabilities to improve the accuracy, real-time performance, and efficiency of fault detection and fault diagnosis. Summary of the Invention

[0008] The purpose of the present invention is to provide a method for detecting faults in truck lithium batteries, which has the advantages of high safety, high reliability, low maintenance cost, and extended battery life.

[0009] The above technical objective of the present invention is achieved through the following technical solutions:

[0010] A method for detecting faults in truck lithium batteries, comprising the following steps,

[0011] Step 1, data acquisition: The battery management system (BMS) collects the voltage, current, and temperature data of the battery pack. The data acquisition frequency is dynamically adjusted according to the working state of the battery pack to ensure timely capture of potential fault signals.

[0012] Step 2, data processing: The collected battery data is filtered, denoised, and normalized to remove environmental interference and ensure the accuracy and consistency of the data.

[0013] Step 3, construct an initial belief rule base expert system model: The initial belief rule base expert system model includes various possible fault modes, including short circuit, over-temperature, over-current, over-voltage, and under-voltage. Using the collected data, an initial belief rule base expert system model for fault detection is constructed through historical data, including the discrimination conditions corresponding to the normal and fault states of each module of the battery.

[0014] Step 4, optimize the initial belief rule base expert system model using the genetic algorithm.

[0015] Step 5, fault detection and isolation: The belief rule base expert system optimized by the genetic algorithm is used to monitor the battery pack in real time. The collected data is input into the belief rule base, and based on rule-based reasoning, the current fault state of the battery pack is output.

[0016] Step 6, feedback and adaptive adjustment: Continuously monitor the battery state during operation, and dynamically adjust the rules and weights of the belief rule base expert system for fault determination according to real-time feedback, further feeding back to optimize the belief rule base expert system and the genetic algorithm model.

[0017] The preferred solutions are as follows:

[0018] Preferably: In step 3, in order to accurately reflect the relationship between the input and output of the belief rule base expert system, the reference value of the input battery data information is divided into three, and the corresponding three fuzzy semantic values are small, medium, and large. The output is divided into five fault types: short circuit, over-temperature, over-current, over-voltage, and under-voltage.

[0019] Preferably: In step 4, when optimizing the initial belief rule base expert system model using the genetic algorithm, the parameter vector set U of the belief rule base system includes the rule weights, premise attribute weights, and confidence degrees of the rule base.

[0020] Input the given input training samples x into the actual system and the belief rule base expert system model i (i = 1, 2,..., M);

[0021] The output data y of the actual system r and the output y of the belief rule base expert system model mThe mean value ξ(U) of the sum of squared differences is optimized for the parameter set U through a genetic algorithm to minimize ξ(U) and ultimately reach a relatively optimal state to adapt to different working environments and battery states;

[0022] The actual system sample output data y r And the mean value ξ(U) of the sum of squared differences between the output of the confidence rule base expert system model y m The formula for the mean value ξ(U) of the sum of squared differences between the outputs is shown in the following formula

[0023]

[0024] Where m is the number of battery data types and N is the number of training samples.

[0025] Preferably: The process of optimizing parameters by the genetic algorithm includes the following steps

[0026] S1. Set the algorithm parameters and initialize the population

[0027] Encode the values in the parameter set U that need to be optimized for the confidence rule base expert system and set their constraints. Use binary encoding, then set the population size n, the maximum number of iterations Gmax, the crossover probability Pc, and the mutation probability Pm, and randomly initialize the population according to the constraints;

[0028] S2. Calculate the individual fitness

[0029] Take the mean value ξ(U) of the sum of squared differences between the output of the actual system and the output of the confidence rule base expert system model as the individual fitness value;

[0030] S3. Determine the selection algorithm

[0031] The roulette wheel selection method determines the probability of each individual being selected according to the ratio of the individual fitness to the overall fitness of the population. The specific calculation expression is as follows

[0032]

[0033] In the formula, i represents the individual serial number, F is the fitness of this individual, Pi represents the probability of each individual being selected into the next generation population, and Qi is to calculate the cumulative probability of each individual;

[0034] S4. Determine the crossover algorithm

[0035] Select the single-point crossover algorithm that is easy to implement and has high efficiency. This algorithm randomly selects a point from the paired chromosomes as the crossover position, and then performs gene exchange on the paired chromosomes at this point. This method helps to find better solutions;

[0036] S5. Determine the mutation algorithm. Use basic bit mutation, that is, only mutate a certain bit in the gene sequence, which helps to promote extensive exploration of the solution space;

[0037] S6. Judge the termination condition. When the preset maximum number of iterations Gmax is reached, output the optimal result, and then decode it into the actual parameters corresponding to the belief rule base expert system for model establishment.

[0038] In summary, the present invention has the following beneficial effects:

[0039] 1. High safety. By real-time monitoring the health status of the battery pack modules and isolating the faulty modules in a timely manner when a fault is detected, the present invention can effectively avoid the expansion of faults, prevent potential safety hazards caused by faults in the battery system, such as overheating, overvoltage or short circuit, etc., and ensure the safe operation of the truck under various working conditions;

[0040] 2. High reliability. The present invention adopts a belief rule base model optimized by a genetic algorithm, which can dynamically adjust the fault detection strategy, significantly improve the response speed and accuracy of the system to battery module faults. Compared with the traditional fault detection method based on threshold judgment, the present invention can more accurately identify and isolate faulty modules, reduce false alarms and missed alarms, improve the reliability of the system, and ensure the continuous and stable operation of the battery pack under complex working conditions;

[0041] 3. Low maintenance cost. Through intelligent fault detection, the present invention can detect potential problems of battery modules in advance, avoid the expansion of faults, reduce the risk of complete damage of the battery pack. Maintenance personnel can quickly locate the faulty module and replace it, shorten the fault repair time, reduce the maintenance cost and maintenance frequency. In addition, intelligent control also reduces the dependence on manual operation, further reducing the maintenance difficulty and cost;

[0042] 4. Prolong the battery life. Through precise fault detection and intelligent management, the present invention can avoid the influence of faulty modules on other normal modules in the battery system, thereby reducing the overall load of the battery pack, delaying the battery aging process, and significantly prolonging the service life of the battery, which provides higher economy and long-term benefits for the truck.

[0043] The present invention adopts a belief rule base model optimized by a genetic algorithm for fault determination and isolation of truck lithium batteries. This model demonstrates its advantages in efficiently processing various data types and establishing non-linear relationships between inputs and outputs. At the same time, it provides higher transparency compared to artificial neural networks, simulates human thinking reasoning, and has strong interpretability. Description of the Drawings

[0044] Figure 1 is the workflow framework diagram of the embodiment;

[0045] Figure 2 is the expert system model of the initial confidence rule base optimized by the genetic algorithm in the embodiment;

[0046] Figure 3 is the flow framework diagram of the expert system of the initial confidence rule base optimized by the genetic algorithm in the embodiment. Specific implementation manners

[0047] The present invention will be further described in detail below with reference to the accompanying drawings.

[0048] The method for detecting faults in a truck lithium battery, as Figures 1 - 3 shown, includes the following steps

[0049] Step 1, data acquisition. The voltage, current, and temperature data of the battery pack are collected through the battery management system BMS. The data acquisition frequency is dynamically adjusted according to the working state of the battery pack to ensure timely capture of potential fault signals;

[0050] Step 2, data processing. The collected battery data is subjected to filtering, denoising, and normalization processing to remove environmental interference and ensure the accuracy and consistency of the data;

[0051] Step 3, constructing an expert system model of the initial confidence rule base. The expert system model of the initial confidence rule base includes various possible fault modes, including short circuit, over-temperature, over-current, over-voltage, and under-voltage. Using the collected data, an initial expert system model of the confidence rule base for fault detection is constructed through historical data, including the discrimination conditions corresponding to the normal and fault states of each module of the battery;

[0052] In order to accurately reflect the relationship between the input and output of the confidence rule base expert system, the reference value of the input battery data information is subdivided into three, and the corresponding three fuzzy semantic values are small, medium, and large, and the output is divided into five fault types: short circuit, over-temperature, over-current, over-voltage, and under-voltage;

[0053] Step 4, optimizing the expert system model of the initial confidence rule base by the genetic algorithm;

[0054] In the process of optimizing the expert system model of the initial confidence rule base by the genetic algorithm, the parameter vector set U of the confidence rule base system includes the rule weights, premise attribute weights, and confidence degrees of the rule base;

[0055] Input the given input training samples x i (i = 1, 2,..., M) into the actual system and the expert system model of the confidence rule base;

[0056] Compare the output data y r of the actual system with the output y mThe mean value ξ(U) of the sum of squared differences is optimized for the parameter set U through a genetic algorithm to minimize ξ(U) and ultimately reach a relatively optimal state to adapt to different working environments and battery states;

[0057] The actual system sample output data y r And the mean value ξ(U) of the sum of squared differences between the output of the belief rule base expert system model y m The formula for the mean value ξ(U) of the sum of squared differences between the outputs is shown in the following formula

[0058]

[0059] where m is the number of battery data types and N is the number of training samples;

[0060] The process of optimizing parameters by the genetic algorithm includes the following steps

[0061] S1. Set the algorithm parameters and initialize the population

[0062] Encode the values in the parameter set U that need to be optimized for the belief rule base expert system and set their constraints. Use binary encoding, then set the population size n, the maximum number of iterations Gmax, the crossover probability Pc, and the mutation probability Pm, and randomly initialize the population according to the constraints;

[0063] S2. Calculate the individual fitness

[0064] Take the mean value ξ(U) of the sum of squared differences between the output of the actual system and the output of the belief rule base expert system model as the individual fitness value;

[0065] S3. Determine the selection algorithm

[0066] The roulette wheel selection method determines the probability of each individual being selected according to the ratio of the individual fitness to the overall fitness of the population. The specific calculation expression is as follows

[0067]

[0068] In the formula, i represents the individual serial number, F is the fitness of this individual, Pi represents the probability of each individual being selected into the next generation population, and Qi is the cumulative probability calculated for each individual;

[0069] S4. Determine the crossover algorithm

[0070] Select the single-point crossover algorithm that is easy to implement and has high efficiency. This algorithm randomly selects a point as the crossover position from the paired chromosomes, and then performs gene exchange on the paired chromosomes at this point. This method helps to find better solutions;

[0071] S5. Determine the mutation algorithm, and use basic bit mutation, that is, only mutate a certain bit in the gene sequence, which helps to promote the extensive exploration of the solution space;

[0072] S6. Judge the termination condition. When the preset maximum number of iterations Gmax is reached, output the optimal result, and then decode it into the actual parameters corresponding to the belief rule base expert system for model establishment.

[0073] Step Five, fault detection and isolation. Use the belief rule base expert system optimized by the genetic algorithm to monitor the battery pack in real time. Input the collected data into the belief rule base, and based on rule-based reasoning, output the current fault state of the battery pack.

[0074] Step Six, feedback and adaptive adjustment. Continuously monitor the battery state during operation, and dynamically adjust the rules and weights of the belief rule base expert system for fault determination according to real-time feedback, and further feedback to optimize the belief rule base expert system and the genetic algorithm model.

[0075] This specific embodiment is only an explanation of the present invention, and it is not a limitation of the present invention. Those skilled in the art can make modifications without creative contributions to this embodiment according to needs after reading this specification, but as long as it is within the scope of the claims of the present invention, it is protected by the patent law.

Claims

1. A truck lithium battery fault detection method, characterized in that: The following steps are included: Step 1: Data collection: The battery management system BMS collects the voltage, current, and temperature data of the battery pack. The data collection frequency is dynamically adjusted according to the working status of the battery pack to ensure that potential fault signals are captured in time. Step 2: Data processing: filtering, denoising and standardizing the collected battery data to remove environmental interference and ensure the accuracy and consistency of the data; Step three, construct an initial confidence rule base expert system model, which includes various possible failure modes, including short circuit, over-temperature, over-current, over-voltage, and under-voltage; it uses the collected data and historical data to construct an initial confidence rule base expert system model for initial fault detection, including the judgment conditions corresponding to the normal and fault states of each battery module; Step 4: Genetic algorithm optimizes the initial confidence rule base expert system model; Step 5: Fault detection and isolation: Use the confidence rule base expert system optimized by genetic algorithm to monitor the battery pack in real time, input the collected data into the confidence rule base, and output the current fault status of the battery pack based on rule reasoning; Step six, feedback and adaptive adjustment, continuously monitor the battery status during operation, and dynamically adjust the rules and weights of the confidence rule base expert system for fault judgment according to real-time feedback, and further feedback optimize the confidence rule base expert system and genetic algorithm model.

2. The truck lithium battery fault detection method according to claim 1, characterized in that: In step three, in order to accurately reflect the relationship between the input and output of the confidence rule base expert system, the input battery data information reference value is subdivided into three, and the corresponding three fuzzy semantic values ​​are small, medium, and large, and the output is divided into five fault types: short circuit, overtemperature, overcurrent, overvoltage, and undervoltage.

3. The truck lithium battery fault detection method according to claim 1, characterized in that: Step 4: In the genetic algorithm optimization of the initial confidence rule base expert system model, the parameter vector set U of the confidence rule base system includes the rule weight, premise attribute weight and confidence of the rule base; Input the given input training sample x to the actual system and the confidence rule base expert system model i (i=1, 2, ..., M); The output data y of the actual system r The output y of the expert system model with the confidence rule base m The mean value of the sum of squares of the differences between them is ξ(U), and the parameter set U is optimized by genetic algorithm to minimize ξ(U) and finally reach a relatively optimal state to adapt to different working environments and battery states; Actual system sample output data y r Expert system model y with confidence rule base m The formula for the mean value ξ(U) of the sum of squared differences between outputs is as shown below, Where m is the number of battery data types and N is the number of training samples.

4. The truck lithium battery fault detection method according to claim 3, characterized in that: The genetic algorithm optimization parameter process includes the following steps: S1. Set algorithm parameters and initialize the population. Encode the values ​​in the parameter set U that the confidence rule base expert system needs to optimize and set its constraints, using binary encoding, then set the population size n, the maximum number of iterations Gmax, the crossover probability Pc and the mutation probability Pm, and randomly initialize the population according to the constraints; S2, individual fitness calculation, The mean value ξ(U) of the sum of squares of the difference between the output of the actual system and the output of the expert system model of the confidence rule base is taken as the individual fitness value; S3, determine the selection algorithm, The roulette wheel selection method determines the probability of each individual being selected based on the ratio of the individual fitness to the overall fitness of the population. The specific calculation expression is as follows: Where i represents the individual number, F is the fitness of the individual, Pi represents the probability of each individual being selected into the next generation population, and Qi is the cumulative probability of calculating each individual; S4, determine the crossover algorithm, Choose a single-point crossover algorithm that is easy to implement and has high efficiency. This algorithm randomly selects a point from the paired chromosomes as the crossover site, and then performs gene exchange on the paired chromosomes at this point. This method helps to find a better solution; S5. Determine the mutation algorithm and use basic bit mutation, that is, only one bit in the gene sequence needs to be mutated, which helps to promote extensive exploration of the solution space; S6. Determine the termination condition. When the preset maximum number of iterations Gmax is reached, output the optimal result, and then decode it into the actual parameters corresponding to the confidence rule base expert system for model establishment.

Images