Load prediction for aptimal thermal storage: comparison of three kinds of model application

It is necessary to predict the load of the following day and hours to establish optimal thermal storage. In this paper, three kinds of load prediction models, the Kalman filter, GMDH and neural network are used and characteristics and usability of each method were compared. It has been shown that proper selection of input variables, method of preprocessing the meas- urement data and the form of prediction equation gave a large influence on the prediction accuracy, and that each of them could predict the cooling load for thermal storage operation with sufficient accuracy.