论文部分内容阅读
雷电是一种落雷时间、地点、能量均无法预测的随机产生的自然现象,这对雷电过电压识别带来了难度,针对传统识别方法使用单一特征信息作为判断依据导致的识别率低等问题,本文针对时域波形、雷击波头特征以及HHT时频谱这三个方面进行特征提取,建立的了普通短路过电压、感应雷击过电压、直击雷过电压、反击雷过电压以及绕击过电压这5种过电压类型与与特征对应关系,并提出一种结合协同量子粒子群优化算法和最小二乘支持向量机的雷击过电压识别模型,通过仿真研究,与基于QPSO-LSSVM算法进行对比,本文研究的识别模型的准确率提高了6.46%。
Thunder and lightning is a natural phenomenon of unpredictable lightning strike time, place and energy, which makes it difficult to identify lightning overvoltage. In view of the low recognition rate caused by using single feature information as the basis for the traditional recognition method, In this paper, the characteristics of time-domain waveforms, lightning strike wave header and HHT time-frequency spectrum are extracted. The common short-circuit overvoltage, induced lightning overvoltage, direct lightning overvoltage, overvoltage strikeback and overvoltage Five kinds of overvoltage types and their corresponding characteristics, and presents a lightning stroke overvoltage identification model combined with cooperative quantum particle swarm optimization algorithm and least square support vector machine. Compared with QPSO-LSSVM algorithm, The accuracy of the recognition model studied increased by 6.46%.