Simulation of Outflow Reservoir Level, and Hydro-Power at Dukan Dam Using Artificial Neural Network

摘要

Dam constructions and the reservoir of resulting water are considered as common implemented methods on river systems, but still the biggest challenge is the accurate simulation of reservoir parameters and the interaction between them. A modeling of reservoir parameters suitable fordifferent and unexpected variations of that parameter can improve the operation understandings of dams. This paper develops a Custom Neural Network (CNN) to simulate the relations between the reservoir parameters for Dukan Dam, which may add a contribution to monthly routing dam models. Threecategories of reservoir parameter interactions have been discussed, we analyzed the distribution patterns for Dukan Dam observations and simulated the outflow, reservoir level, and the hydro-power as outcomes of the proposed models. The results have been evaluated by the correlation coefficientbetween the predicted and the observed monthly data over the given time. Lower correlation coefficients (0.85537) due to the sharp and the non-periodic variations in the observations, while a higher value were attained when the data tend to be smoother and relatively periodic. This researchproves the effectiveness of the forward CNN for such application, and the using of one hidden layer with number of neurons equivalent to the peaks of parameter variations may be sufficient to lead the optimal results.