Evaluation of loadability limit of pool model with TCSC using optimal featured BPNN

This paper presents an approach for online evaluation of loadability limit of Pool Model with Thyristor Controlled Series Compensator (TCSC) for various load patterns using Back Propagation Neural Network (BPNN) with optimal feature set. Differential Evolution (DE) algorithm is employed to find out optimal location and control of TCSC. This approach uses AC load flow equations with constraints on real and reactive power generations, transmission line flows, magnitude of bus voltages and TCSC settings. The input parameters are real and reactive power loads at all buses. The BPNN is trained through off-line simulation using DE algorithm and tested with new load patterns. The optimal feature set for training BPNN is obtained by a wrapper model of feature selection called Sequential Forward Selection (SFS). Simulations are performed on 39 bus New England test system. The performance of the proposed model is compared with unified BPNN trained with full feature set. The selection of optimal features with SFS has significantly reduced the training time of BPNN with minimal Mean Squared Error (MSE) for the evaluation of loadability limit of pool model with TCSC.