Abstract: Neural Network (NN) has been the most popular technique used in predicting Reservoir Water Level (RWL). However, NN is a black-box modelling technique where the model can be established without knowledge of the mathematical relationship between the inputs and the corresponding outputs. Most researches on reservoir water release applied the NN techniques using discretized data. To discover the current Reservoir Water Level at time t (RWLt) in relation to the previous rainfall event, this paper proposed a predictive t modelling for RWL using regression and the temporal pattern of both RWL and rainfall. The sliding window technique has been used to segment the temporal data into various slices. The finding shows that the best input scenario for the current RWL is one day delay for RWL and two days delay for rainfall; comparing this to the actual data, the model has an error of 0.1628%. The model can be used to guide the reservoir operator predicting the present and immediate decisions on reservoir water release, especially in the absence of the supervisor or during emergency situations.
Siti Rafidah M-Dawam and Ku Ruhana Ku-Mahamud, 2016. Predictive Modelling for Reservoir Water Level. Research Journal of Applied Sciences, 11: 851-857.