Comparison of Two Charateristics Extended Process Neural Networks

The input of process neural networks changed from discrete data to veritable data.As an extension to process input from instantaneous input on essence,the input can be considered as generalization of learn sample characters,and aggrandizing the sample information.The two different ways based on the time-domain feature expansion and the orthogonal decomposition feature expansion are used to establish process neural networks model.According to the data of student electric consumption in Jiangnan University,the model training and the accuracy of load forcasting are investigated.The simulation results show that the orthogonal decomposition feature expansion based process neural networks is supprior to time-domain feture expansion based proscess neural networks in training speed,veracity in prediction,and more suitable for the application of electric load forecasting.