Advanced Reactive Power Compensation of Wind Power Plant Using PMU Data

This paper introduces a new model to improve the wind power plant performance by modeling its reactive power demand. It develops a probabilistic model based on prediction interval to help better modeling of the reactive power demands of wind unit which needs to be compensated by the static VAr compensator (SVC). This is made possible by the use of a non-parametric neural network (NN) based model using the lower and upper bound estimation (LUBE) method. To avoid the instability arising due to the nonlinear and complex nature of NN, the idea of combined prediction intervals is used here. Due to the highly nonlinear and non-stationary characteristics of the reactive power pattern consumed in the wind power plant, a new optimization algorithm based on theta -symbiotic organisms search (theta -SOS) is proposed to train the LUBE model parameters in the polar coordinates. In addition, a two-phase modification method is developed to enhance the local search ability of SOS and avoid premature convergence issue. The performance of the proposed model on the experimental Phasor Measurement Unit (PMU) data of a wind unit shows that the model can help to improve the performance of the wind SVC, effectively.