Examination of Essential Factors for Chilled and Hot Water Demand Load and Adaptive Load Prediction Using Neural Networks

Abstract

This paper presents a following day's hot and chilled water load forecasting system by using artificial neural networks. The neural network can realize the nonlinear relationship between the load and the environmental factors. Load forecasting is very important for the economic operations of air-conditioning system. Many statistical methods have been developed and used for such forecasting. It has been difficult to construct a proper functional model or needed huge efforts. We propose the optimum neural-network structure and sufficient training method in concern of learning speed and categorizing ability, under the investigation on the effective factors for the forecasting system. From the simulation experiments, we obtained the successful results comparing with conventional methods.