On comparison of two strategies in net demand forecasting using Wavelet Neural Network

Abstract

In this paper, direct and indirect net demand forecasting approaches are compared. Net demand is defined as the total system load minus total wind power generation of the system. Since volatility of wind power is added to the net demand, it is more volatile and uncertain than the load alone. This could make the results of direct and indirect net demand forecasting approaches different. Wavelet Neural Network (WNN) with Morlet Wavelet activation function is selected to be the forecasting engine for wind power, load, and net demand in this paper. For training the WNN, Levenberg-Marquardt algorithm is used. Simulations are performed using Alberta's and Ireland's wind and load data. The WNN forecasting engine is compared to MLP and RBF neural networks along with the persistence. Results showed the superiority of the WNN over other models for net demand forecasting application.