梯度下降优化模糊系统的接触电阻预测方法  

Method for Predicting Contact Resistance of Optimizing Fuzzy System by Gradient Descent Algorithm

作　　者：王刚 谭盛武 何子博[2] 林生军 王之军 常林晶 

机构地区：[1]平高集团有限公司国家电网高压开关设备绝缘材料实验室,河南平顶山467001 [2]西安工业大学材料与化工学院,陕西西安710021

出　　处：《华侨大学学报（自然科学版）》2017年第1期86-90,共5页Journal of Huaqiao University(Natural Science)

基　　金：国家高技术研究发展(863)计划项目(2014AA051802);国家自然科学基金资助项目(51477120);国家电网科技计划项目(SGNXJX00YJJS1400105)

摘　　要：根据接触电阻的特点,将结合模糊逻辑的预测方法引入电气领域,提出基于模糊系统的接触电阻预测新方法.根据接触电阻与各影响因素之间的关系及研究目的进行试验,得到足量试验数据,将所有试验数据分成两部分,训练数据和测试数据.通过训练数据运用梯度下降算法训练模糊系统,调整模糊系统参数,建立相应的接触电阻模型,利用训练数据建立接触电阻的回归分析模型.通过测试数据对两种模型进行测试,基于模糊系统的接触电阻模型的测试结果优于回归分析.预测与比较结果表明:若能得到足量训练数据,用梯度下降算法训练模糊系统建立的接触电阻模型精确可靠.Based on the characteristics of the contact resistance, the predicting method combined fuzzy logic is introduced to electric field,the new method for predicting the contact resistance based on fuzzy system is de-veloped. According to relationship between contact resistance and influence factors, as well as the research tar-get, the test is processed. The enough data is obtained and all data is divided into two parts, namely, training data and testing data. Fuzzy system is trained by gradient descent algorithm through training data, the system-ic parameter is adjusted, the corresponding model of contact resistance is found. The regression analysis model of contact resistance is built by training data. The two models are tested through testing data, the prediction effect of the model based on fuzzy system is better than that of regression analysis. The prediction results show that if enough training data is obtained, the fuzzy system trained by gradient descent algorithm is reliable to predict the contact resistance.

关 键 词：接触电阻 模糊系统 梯度下降算法 回归分析 

分 类 号：TM11[电气工程—电工理论与新技术]