Optimization of unit commitment by improved Hopfield neural network algorithm

Unit commitment is a very important issue of generation scheduling in electric power systems. A rational scheme will lead to cost savings. Based on the principle of Hopfield neural network and the characteristics of unit commitment, an improved Hopfield neural network algorithm is presented, which can be applied to the optimal generation unit commitment. The improved neural network combines good solution quality of simulated annealing with rapid convergence of Hopfield neural network. The method replaces the digital neurons with two binary states by analogue neurons with continuous output and dispenses with additional economic power dispatch, and can quickly search high-quality optimal solution of the system. The simulation results from a practical power system prove the method is very effective in reaching proper unit commitment. It has good parallel characteristics, and easily implemented on parallel computer. The possibility of the application to practical power systems is obvious.