Integrated Artificial Neural Network (ANN) and Stochastic Dynamic Programming (SDP) Model for Optimal Release Policy

Complexicity in reservoir operation poses serious challenges to water resources planners and managers. These challenges of water reservoir operation are illustrated using a simulation to aid the development of an optimal operation policy for dam and reservoir. To achieve this, a Comprehensive Stochastic Dynamic Programming with Artificial Neural Network (SDP-ANN) model were developed and tested at Sg. Langat Reservoir in Malaysia. The nonlinearity of the natural physical processes was a major problem in determining the simulation of the reservoir parameters (elevation, surface-area, storage). To overcome water shortages resulting from uncertainty, the SDP-ANN model was used to evaluate the input variable and the performance outcome of the Model were compared with the Stochastic Dynamic Programming integrated with auto-regression (SDP-AR) model. The objective function of the models was set to minimize the sum of squared deviation from the desired targeted supply. Comparison result on the performance between SDP-AR model policy with SDP-ANN model found that the SDP-ANN model is a reliable and resilience model with a lesser supply deficit. The study concludes that the SDP-ANN model performs better than the SDP-AR model in deriving an optimal operating policy for the reservoir. Copyright Springer Science+Business Media Dordrecht 2013