The Forecast of Energy Demand on Artificial Neural Network

Traditional method about forecast of energy demand, Trend Extrapolation, can´t study the information supplied with date effectively, and BP neural network has the great power of goal learning, which can dig potential function in the date. The article design the GDP and other factors as input variables, and use steepest descent back propagation to adjust the weight and threshold of network. We choose the optimal number of hide layer via experimentation, and achieve the train and simulate of network with MATLAB. The final result shows that the forecast of neural network has much higher precision than the forecast of trend extrapolation. The article indicates that BP neural network has the higher precision.