Evaluating the Capability of Adaptive Neuro-Fuzzy Inference System to Predict of Flushing Half-Cone Volume in Reservoirs

摘要

Volume of flushed sediment from reservoir can be considered as one of the main points in sediment management issues. Retrieving the storage capacity, cleaning adjacent of power plant intakes and sediment replenishment at downstream area are relatively related to precise prediction of flushed sediment volume. Therefore presenting the intelligent models is necessary to increase the accuracy of calculations, decrease of response time and avoid the uncertainties of multiple regressions models. In this paper, the Adaptive Neuro-Fuzzy Inference System (ANFIS) was developed on a wide database of 3 different experimental studies. The results show ANFIS model simulated the actual experimental data quite successfully. The predicted flushed sediment volume in ANFIS was more accurate than multiple regressions results. Finally sensitivity analysis was conducted on the best ANFIS model to select the key parameter affected on flushing processes. It was found that the height ratio of sediment to water is an important parameter to predict the flushing half-cone volume.