Optimal operation of reservoir based on neural network analysis of downstream flooding

The largest reservoir of Bangladesh - the Karnafulli is generally operated on an ad hoc basis resulting into a non-optimal power generation and sometimes causes severe flooding of downstream areas including the city of Chittagong. An integrated approach, combining. flood management, optimization, and simulation models, was developed and applied for the optimal operation of the Karnafulli reservoir. Particular attention was paid to incorporate flood damage functions of downstream areas in reservoir operation. Potential flood, damages were assessed by using a neural network which was trained by an extensive data set generated by a combined hydrodynamic-GIS model (MIKE11-GIS) for the rivers and flood plain. An optimal operation policy was derived from the optimization-simulation model results. The performance of the suggested policy was evaluated by comparing with the existing operation rules. Using the model derived policies, the reservoir operation would result in higher energy generation while keeping the downstream flooding to a minimum.