Modelling of Hydropower Reservoir Variables for Energy Generation: Neural Network Approach
Efficient management of hydropower reservoir can only be realized when there is sufficient understanding of interactions existing between reservoir variables and energy generation. Reservoir inflow, storage, reservoir elevation, turbine release, net generating had, plant use coefficient, tail race level and evaporation losses are the major hydropower reservoir variables affecting the energy generation. Thus, this paper presents the modelling of reservoir variables of two hydropower dams along the River Niger (Kainji and Jebba dams) in Nigeria for energy generation using multilayer perceptron neural network. Total monthly historical data of Kainji and Jebba hydropower reservoirs’ variables and energy generated were collected from Power Holding Company of Nigeria respectively for a period of (1970-2011) and (1984-2011) for the network training. These data were divided into training, testing and holdout data set. The neural network summary yielded a good forecast for Kainji and Jebba hydropower reservoirs with correlation coefficients of 0.89 and 0.77 respectively. These values of the correlation coefficient showed that the networks are reliable for modeling energy generation as a function of reservoir variables for future energy prediction.
Key words: Hydropower, reservoir variables, neural network, energy generation, coefficient of correlation