Augmented Hopfield network for unit commitment and economic dispatch

The Hopfield network has been applied to the power system economic dispatch problem with very promising results. However, it has been found that the unit commitment problem cannot be tackled accurately within the framework of the conventional Hopfield network. This is due to the fact that both discrete and continuous terms must be considered to fully model the problem. This paper presents an augmented network architecture with a new form of interconnection between neurons giving a more general energy function containing both discrete and continuous terms. A comprehensive cost function for the unit commitment problem is developed and mapped to this energy function. Results show that this technique outperforms previous neural network methods. The new method also compares favourably with Lagrangian relaxation. Detailed results for a power system with thermal, hydro and pumped storage units are presented