Daily water level forecasting using adaptive neuro-fuzzy interface system with different scenarios: Klang Gate, Malaysia

Abstract

Forecasting the level of reservoir has been a significant subject in the management of reservoirs and water resource. For many years, estimation of reservoir water level was primary based on operator’s experience, curves and mathematical models. Recently, Artificial Intelligence (AI) methods are developed in several hydrological aspects, such as classification and forecasting parameters. The major advantage of AI modeling is the considerable ability to map input-output pattern without requiring prior knowledge about the factors that affect the forecasting parameters. This study attempts to forecast the daily level of Klang Gate dam using adaptive neuro fuzzy interface system (ANFIS) in two different scenarios and various time delays in inputs. In the first scenario, daily rainfall is used solely as an input in different time delays from the time (t) to the time (t-4) that is illustrated in spite of the reasonable performance of error, less than 10% of solely rainfall data could not have reasonable response in fluctuations to forecast accurately. Increasing the level of reservoir beside precipitation as inputs in both sets of models could enhance the fitness of the estimated and observed data dramatically. Due to the fact that the distance of gauges of stations are unknown, using various models in different time delays of inputs could demonstrate the distance between gauges; moreover, it shows the reasonable duration in inputs and outputs to have accurate prediction. Key words: Klang Gate, adaptive neuro fuzzy interface system (ANFIS), forecasting model.