The application of the multi-scale wavelet neutral network energy consumption forecast model in the steel corporation

This paper puts forward a kind of multi-scale wavelet neutral network which can forecast energy waste with high frequency noise signal on the basis of analyzing the wavelet neutral network and multi-scale space decomposability. In the paper, we have analyzed the network structure and training algorithm and made the simulation research. Overcoming the inhere defect of the BP network, it is that the training speed is slow, easy convergence the part infinitesimal dot and the whole searching ability is feebleness. Through the simulation of the algorithm, the experiment result indicated the model that this paper advanced had the high forecast precision and the fast convergent speed. The maximum relative error is 0.62%, the minimum relative error is 0.09% and the average relative error is 0.29% between forecast value and actual value. The system has better forecast effect and offers the new method for making energy forecast by using the neutral network method.