Analysis and Prediction of Water Temperatures in a Reservior by the Fuzzy GMDH

Abstract

The GMDH (Group Method of Data Handling) is well-known technique for obtaining a mathematical model of complex phenomenon. The fuzzy GMDH has been reformulated by the possibility measure. In the fuzzy GMDH algorithm, possibilistic linear systems with fuzzy parameters are used to obtain a possibilistic linear model by using a multilayer procedure, given the observed data.Water temperature in dam reserviors exerts influence on water supply, irrigation and so on. Accordingly, the prediction of water temperature is needed to avoid bad influences. In this paper, the fuzzy GMDH is applied to prediction of water temperatures. The data of air temperatures and water temperatures in the Daiichi Yahagi Dam are observed during from July 1st to July 31th, 1980. By the fuzzy GMDH algorithm, the model of water temperatures is built from these data. The estimated model is used to forecast water temperatures from August 1st to August 31th, 1980. As a result, 71% of all observed data are included in the estimated fuzzy outputs. Therefore good estimated intervals are obtained. The fuzzy GMDH can be used as interval analysis. A vague phenomenon must be forcasted by intervals which represent our ignorance.