Combination of generator capability curve constraint and statistic-fuzzy load clustering algorithm to improve NN-OPF performance

The inclusion of statistic-fuzzy Load clustering method in algorithm of NN-OPF is intended to make the NN-OPF robust in load changing at a certain load range. Whereas, The inclusion of Generator Capability Curve (GCC) as a constraint in NN-OPF is to ensure cheap and safe operation of generators. NN-OPF is built with reference to a Particle Swarm Optimization Optimal Power Flow (PSO-OPF). There are three stages in Designing NN-OPF. The first stage is design of PSO-OPF with generator capability curve constraint. The second stage is load clustering using statistic-fuzzy method. The third stage is training NN-OPF using constructive back propagation method. In training process of Neural Network (NN), the nearness index of load curve (FW _ik) resulting from statistic-fuzzy method, and total load (active power, reactive power), are used as input. The pattern of generator scheduling resulting from PSO-OPF is used as outputs. In this paper, the Java-Bali power system is used as sample system to verify the validity of this method. The simulation results using MATLAB software have shown that the proposed method has good performance. The proposed method is possible to apply in on-line system in normal condition, especially in representing non linear generation operation limit near steady state stability limit and under excitation operation area.