基于GRNN的二次设备在线监测信息预测

摘要

In allusion to continuous information in online monitoring information of grid secondary equipment such as power source voltage,device temperature,CPU usage rate and so on,generalized regression neural network (GRNN)is used for forecasting changes of the information. By comparing predicted value and actual value,relative error curve is obtained. It is verified higher veracity of GRNN in forecast by comparing with back propagation (BP)neural network and radial basis function (RBF)neural network. Meanwhile,this paper discusses to take relevance of online monitoring information of sec-ondary equipment into consideration as well as relate temperature to CPU usage rate for prediction. According to compari-son results,it is proved that it is useful to improve veracity of state prediction by taking relevance of information into consid-eration.%针对电网二次设备在线监测信息中连续变化的信息,如电源电压、装置温度、CPU 使用率等,应用广义回归神经网络(generalized regression neural network,GRNN)对这些信息在线监测值进行预测,通过对比预测值和实际值,得出相对误差曲线；再与反向传播算法(back propagation,BP)神经网络和径向基函数(radial basis function,RBF)神经网络进行对比,验证了 GRNN在预测中具有更高的准确性。同时,探讨了将二次设备在线监测信息的关联性考虑进预测中,将温度和CPU使用率关联并进行预测,与之前的预测结果进行对比,证明将信息之间的关联性考虑到预测中去有助于提高状态预测的准确性。