Research On Fault Diagnosis Method Of Electric Vehicle Battery Based On Data Mining

With the increasingly prominent global energy and environmental problems,it is imperative to reduce carbon emissions.China has also set the goal of carbon emission peak and carbon neutrality in 2020.In order to achieve this important goal,the automotive industry is bound to vigorously develop new energy vehicles.With the support of national policies and the traction of market demand,the new energy vehicle industry has ushered in major opportunities.However,electric vehicle safety accidents mainly caused by battery failure occur frequently and increase year by year,which brings great challenges to the new energy vehicle industry.Therefore,focusing on the problems of difficult,untimely and even misdiagnosis of battery fault diagnosis,it is of great strategic significance to carry out battery fault diagnosis research to achieve fast,accurate,safe,and efficient fault diagnosis and troubleshooting.Data mining integrates multiple disciplines and technical means,and has become an effective method for fault diagnosis in the field of artificial intelligence.Based on the multi-dimensional data and fault mechanism of lithium-ion batteries,this paper uses data mining algorithms such as classification,clustering,and anomaly detection to achieve single battery fault diagnosis,battery pack inconsistency detection,and battery pack fault diagnosis.Firstly,the basic structure,working principle and performance parameters of lithium-ion battery are comprehensively considered,and the fault causes and types of battery are deeply analyzed.By introducing two fault injection methods of software injection and hardware injection,battery fault tests are designed to provide data support for subsequent research.Secondly,the basic structure of RBF neural network and the advantages of subtractive clustering algorithm are studied.For single battery,an improved RBF neural network battery fault diagnosis scheme based on the subtractive clustering algorithm is proposed,and the single battery fault diagnosis process is designed.Simulation results show that this scheme can accurately diagnose common battery faults,and is faster and more efficient than traditional methods.Then,the causes of the inconsistency of the battery pack and the performance of the inconsistent monomer in the battery pack are studied.Several classical anomaly detection algorithms are introduced,and two anomaly detection algorithms based on Euclidean distance and K-means clustering are used to detect the battery pack.The results show that both methods can detect the abnormal monomer in the battery pack.Compared with the algorithm based on distance,the anomaly detection algorithm based on clustering is faster and more intuitive.Finally,aiming at the problem of battery pack fault diagnosis,combined with the research results of single battery fault and battery pack inconsistency,the following battery pack fault diagnosis scheme is designed:firstly,the inconsistency detection is carried out to locate the fault battery,and then the single battery diagnosis scheme is used to diagnose the fault cause.Three groups of faulty ternary lithium-ion battery packs are actually tested under DST condition.The test results show that the designed battery pack fault diagnosis scheme can accurately and quickly locate the fault battery and diagnose the fault cause.