基于极端学习机的SLFN谐波检测法.pdfVIP

学兔兔 第50卷 总第572期 电测与仪表 Vo1．50 No．572 2013年 第8期 E]ectricai MeasurementInstrumentation Aug．2013 基于极端学习机的SLFN谐波检测法术 项子旋，王振宇，刘国海，奚家健 (江苏大学电气信息工程学院，江苏镇江212013) 摘要：为了解决传统电网谐波检测法的实时性和精确度问题，提出一种基于极端学习机(ELM)单隐层前馈神经 网络(SLFN)的谐波检测法。该方法把ELM用于SLFN，将傅里叶变换系数选作ELM模型的检测参数，用广义逆矩 阵计算输出层权值的最小二乘解，以获得该模型的最优输出向量，从而改善各次谐波电流的检测性能。运用 MATLAB仿真软件进行了仿真实验。结果表明：与基于BP算法的sLFN谐波检测法相比，所提出ELM算法的SLFN 谐波检测法达到稳定检测的时间缩短，且检测误差的精度有所提高，达到了有效提升谐波检测实时『生和精确度 的目的。 关键词：极端学习机(ELM)；单隐层前馈神经网络(SLFN)；谐波电流；实时检测；电力网 中图分类号：TM93 文献标识码：A 文章编号：1001—1390(2013)08—0048—05 Harmonic Detecting Approach of SLFN Based on ELM XIANG Zi一-xuan，WANG Zhen-yu，LIU Guo-hai，XI Jia-jian (School of Electrical Information Engineering,．Jiangsu University，Zhenjiang 212013，Jiangsu，China) Abstract：In order to solve the accuracy and practicability in harmonic detection，a kind of SLFN based on extreme learning machine algorithm is proposed．This paper properly applies the ELM to the SLFN，which means adopting the Fourier transform coefficients as the test parameter of ELM model，and employing generalized inverse matrix to calculate the least squares solutions of the output layer weights．On this basis．this mode]【will【not only obtain the optimal output vector，butalso improve the performance detecting of each harmonic current wil1．Based on the MATLAB simulation，the results of contrasting ELM with BP show that the former has good stability and high measurement accuracy，and its harmonic detection of real-time and accuracy has effectively improved． Key words：e