Constrained power system state estimation on recurrent neural networks

A new method for power system state estimation which combines robust M-estimation with treatment of the inequality constraints is presented. The main advantage of the method is that most expensive computation is performed in a neural network which is amenable to parallel implementation. The designed recurrent neural networks are based on differential equations and realize searching a saddle point for appropriate Lagrangian function. Test results on standard test system are used to illustrate the effectiveness of the method.