Abstract
To operate energy supply systems optimally from the viewpoints of energy and cost savings, it is important to predict energy demands accurately as basic conditions. Several methods of predicting energy demands have been proposed, and one of them is to use multi-layered neural network models. Although gradient methods have conventionally been adopted in the back propagation procedure to identify the values of model parameters, they have the significant drawback that they can derive only local optimal solutions. In this paper, a global optimization method named "Modal Trimming Method" for nonlinear programming problems is adopted in the back propagation procedure to derive global quasi-optimal solutions. The multi-layered neural network model is applied to the prediction of the cooling demand in a building used for a bench mark test of a variety of prediction methods, and its validity and effectiveness are clarified.
