重庆理工大学学报2025，Vol.39Issue(21)：30-38,9.DOI:10.3969/j.issn.1674-8425(z).2025.11.004
电驱动系统效率试验数据质量评估方法研究
Research on data quality evaluation method for efficiency test of electric drive system
摘要
Abstract
As the primary power source of electric vehicles,the E-drive system directly determines the vehicle'sdynamic performance and energy economy.The development of E-drive system requires continuous testing to evaluate actual performance,identify weaknesses,and support ongoing optimization and validation.Thus,efficiency testing has become an indispensable component of E-drive system development.However,each efficiency test generates a large volume of high-dimensional data,and with repeated testing,the cumulative data continue to expand dramatically.Conventional data processing methods struggle to extract meaningful features or actionable insights from such massive datasets,leading to data redundancy and low utilization efficiency. Data mining technology employs algorithmic learning models to process large-scale datasets rapidly,thereby,enchancing both computational speed and analytical accuracy.Moreover,it captures and learns feature relationships within high-dimensional datasets,enabling precise data prediction and in-depth analysis of underlying factors that support decision-making. To address the challenges posed by high-dimensional data generated in E-drive system efficiency tests,this paper proposes a data mining-based method for preprocessing and quality assessment of efficiency test data. First,efficiency measurement principles,measurement parameters,and test methodologies are analyzed.Then,an efficiency testing method is formulated,an E-drive system efficiency test platform built,and tests conducted to collect data for subsequent data mining analysis. Next,to effectively reduce the noise in the efficiency test data,the acquired dataset is examined from two perspectives:test data characteristics and data representation.Computing methods and visualization tools are employed to analyze data features,on the basis of which criteria for selecting appropriate preprocessing algorithms are developed.The efficiency data undergoes preprocessing.Considering the distribution characteristics of noise in the data and integrating the concepts of IQR and MAD,a probability distribution-based noise reduction method suitable for efficiency test data is proposed to validate its effectiveness by employing multiple sets of test data. Furthermore,to address the challenge of outlier handling,a two-stage outlier detection method based on data mining technology is introduced.It improves the parameter selection approach of the traditional DBSCAN model by applying an improved particle swarm optimization(IPSO)algorithm.The enhanced DBSCAN model clusters data according to different speed conditions and detects obvious outliers outside clusters,improving model accuracy and efficiency.A combined LOF-iForest model performs secondary inspection,leveraging the respective characteristics of each algorithm to identify both obvious and complex outliers within clusters. Finally,based on quality assessment principles and data characteristics,a multi-dimensional data quality evaluation model is built and corresponding metrics are designed across different dimensions.By integrating the analytic hierarchy process(AHP)and the entropy weight method,it achieves a unification of subjective and objective weighting.The effectiveness of the quality assessment model is validated with data both before and after pre-processing.关键词
电驱动系统/试验数据/数据预处理/异常值检验/质量评估Key words
electric drive system/test data/data preprocessing/outlier test/quality assessment分类
交通工程引用本文复制引用
ZOU Xihong,WANG Xiaoli,YUAN Dongmei,ZHOU Qing,XIONG Feng,ZHOU Zhen,WANG Wanying..电驱动系统效率试验数据质量评估方法研究[J].重庆理工大学学报,2025,39(21):30-38,9.基金项目
国家自然科学基金项目(52202437) （52202437）
