An enhanced neural network load frequency control technique

We investigate the use of neural networks (NN) to act as the control intelligence in conjunction with a standard adaptive load frequency control scheme. In this approach a NN is operated in parallel with a full load frequency adaptive control scheme. The NNs are able to monitor the system frequency as the controller issues its control commands. If there is a condition where the frequency values are corrupted, or the system is not sufficiently exciting, then the NNs will be able to provide the power set points that may be directly communicated to those gensets that are providing system frequency control. This neural control approach is shown to have several advantages over the basic fixed parameter schemes and the more advanced adaptive control techniques that have been recently developed. The results of a number of real-time power system simulations are presented.