A Hybrid Artificial Neural Network Gravitational Search Algorithm for Rainfall Runoffs Modeling and Simulation in Hydrology

Abstract: Artificial Neural Network (ANN) as a method of data processing and inspired by studies of the nervous systems – has become a robust tool for modeling complex, non-linear and dynamic processes due to its flexible mathematical structure that easily generalize patterns with results even with imprecise, noisy and ambiguous input data. This work describes ANN’s application to implement a model to simulate runoff at the Benin Owena River Basin Developmental Agency (BORDA) – with data collected from four (4) gauge and six (6) stream-flow stations namely: Benin, Ekpoma, Sapele and Agbor catchments respectively. The study uses the 4G SE-design; the structured analysis of the existing is based on the lumped, conceptual hybrid (HBV and TOPMODEL) used for calibration and validation. The existing system’s bottleneck includes large computational demand, excessive parameter requirement with validation, still an on-going process. Its mean annual rainfall of Benin, Ekpoma, Sapele and Agbor stations are 823, 732, 962 and 734mm respectively with computed COE of 58, 24, 56 and 42% respectively – indicating strong inter-annual and spatial variability in sub-catchments. Variation in the annual rainfall was observed and long-term runoff trend reflects, the effect of variation cycle with significant correlations between rainfall and runoff as observed in sub-catchments (via historic dataset obtained for the period). The study contributes as: (a) soft computing (a branch of Artificial Intelligence) with an aim to create a synergy with other fields/disciplines, and in this case (hydrology) in its bid to implement the hybrid ANNGSA algorithm for RR process. It also contributes in Artificial Intelligence (AI) – as it aims to create machine/system that mimics the human brain – so that such systems (in this case – hybrid ANNGSA model) that will train the ANN network to simulate future flood occurrence, provide lead time warning for flood management.

Keywords: Catchment, Algorithms, Evolutionary, Fitness Function

^1,2Mathematics/Computer Science and Earth Science Departments, Federal University of Petroleum Resources Effurun, Delta State. ojugo_arnold@yahoo.com, emusjul@yahoo.co.uk
³Department of Computer Science, Delta State Polytechnic Ogwashi-Uku, Delta State, rumerisky@yahoo.com
^4,5Department of Computer Science Edu., Federal College of Education-Technical, Asaba, Delta State, okeyokonta@yahoo.com, an_drey2k@yahoo.com

[PDF]

Cite: A.A. Ojugo, Emudianughe J., R.E Yoro, E.O. Okonta and A.O. Eboka, "A Hybrid Artificial Neural Network Gravitational Search Algorithm for Rainfall Runoffs Modeling and Simulation in Hydrology," Progress in Intelligent Computing and Applications , vol. 2, no. 1, pp. 22-33, March 2013.