摘要
配网环境噪声以及复杂传输路径等干扰因素使得配网告警信号中伴生信号特征,往往表现得非常微弱,采用传统的信号剔除方法难以从告警信号中准确提取出伴生信号特征向量,伴生信号剔除准确率较低。针对此问题,提出一种基于强化学习支持向量机的配网告警信号中伴生信号剔除方法。采用小波分析算法对原始配网告警信号进行分解,获得若干子带信号,以峭度准则为指导标准对子带信号进行合并处理,并从所得合并结果中过滤出峭度极值分量,通过分析剩余信号分量的包络谱来提取伴生信号特征向量。将提取的伴生信号特征向量作为学习样本,强化学习支持向量机,并依据得到的伴生信号分类结果"对"或"错"的反馈对支持向量机分类器进行改进,达到从告警信号中有效剔除伴生信号的目的。实验结果表明,所提方法能够准确滤除配网告警信号中的伴生信号,具有一定的可靠性。
Interference factors such as distribution network environmental noise and complex transmission paths make the associated signal characteristics in the distribution network alarm signal often very weak. It is difficult to accurately extract the associated signal feature vector from the alarm signal by the traditional signal rejection method, and the associated signal rejection is accurate. The rate is lower. Aiming at this problem , a method of culling the associated signals in the distribution network alarm signal based on the enhanced learning support vector machine is proposed. The wavelet analysis algorithm is used to decompose the original distribution network alarm signal to obtain several subband signals. The sub-band signals are combined by the kurtosis criterion as the guiding standard , and the kurtosis extreme components are filtered out from the obtained combined results. The envelope spectrum of the residual signal component is used to extract the associated signal feature vector. The extracted associated signal feature vector is taken as a learning sample, the learning support vector machine is strengthened, and the support vector machine classifier is improved according to the feedback of the resulting signal classification result “pair” or “wrong”,so as to effectively liminate the warning signal. The purpose of the associated signal. The experimental results show that the proposed method can accurately filter the associated signals in the distribution network alarm signal and has certain reliability.
作者
刁柏青
姚刚
董学新
李晓志
DIAO Boqing;YAO Gang;DONG Xuexin;LI Xiaozhi(Shandong Electric Power Co., Ltd., Jinan 250001, China;Shandong Power Company Weihai Power supply Company, Weihai Shandong 264200,China)
出处
《自动化与仪器仪表》
2019年第4期192-195,200,共5页
Automation & Instrumentation
关键词
强化学习
配网
告警信号
伴生信号剔除
支持向量机
Reinforcement learning
distribution network
alarm signal
associated signal rejection
support vector machine
