Optimal multi-reservoir network control by two-phase neural network

Abstract

An approach based on two-phase neural network (TPNN) is proposed for the optimal operation of multi-reservoir network control problems. The advantage of the proposed technique is that it takes into account the concurrent interaction among all the water release variables of the problem. Here, the main objective of this work is to figure out the optimal amounts of water releases from each hydro-plant during each interval in the interconnected system and to minimize and distribute uniformly the energy deficit if any. This TPNN approach is basically a two-stage solution method. In stage 1, the neural network is developed to bring the solution trajectory close to the boundary of the feasible region. In stage 2, the directional vector of the constraints is slowly shifted to the corresponding Lagrange multipliers and this moves the solution trajectory to the feasible region which satisfies all practical constraints. Application of this technique to a 10-reservoir network demonstrates efficacy of the proposed algorithm. It is concluded from the results that the proposed method with proper selection of network control parameters is very effective in providing a good optimal solution.