Recurrent neural network for solving the short-term hydrothermal scheduling problem

The short-term hydrothermal generation scheduling problem is aimed to minimize the total thermal plant cost while satisfying diverse hydraulic, thermal and load balance constraints. Firstly, a multi-reservoir cascaded hydro power system is set up. The power system consists of several cascaded hydro plants and several thermal plants. The power generation function is the function of water discharge rate and storage volume. Then, a recurrent neural network is proposed to optimize the power generation function. Simulation results show the validity of the proposed method.