Explaining Internal Behavior in a Fuzzy If-Then Rule-Based Flood-Forecasting Model

Abstract

This paper presents a popular fuzzy rule-based model for river flow forecasting for an Indian basin. To set up the fuzzy rules, a cluster estimation method is adopted to determine the number of rules and the membership functions of variables involved in the premises of the rules. The most appropriate set of input variables was determined by trial and error procedure to test the coherence of the different input variables in forecasting flood. It is observed that the last time steps of measured runoff are dominating the forecast. The developed model is used to forecast up to 12 h in advance. The values of three performance evaluation criteria namely, the coefficient of efficiency, the root-mean-square error and the coefficient of correlation, were found to be very good and consistent for flows forecasted 1 h in advance by the model. The performance is decreasing as the forecast horizon is increasing and a reasonable forecast is obtained up to 9 h ahead. A set of fuzzy rules is extracted and used for understanding of the behavior of the developed model. It is observed that the developed model follows the trend of the input membership grade in antecedent part of the fuzzy model.