Feed-Forward and Long Short-Term Neural Network Models for Power System State Estimation

Abstract

The primary objective of this paper is to propose the two new combined approaches based on Feed-Forward and Long Short-Term Memory Neural Network models for Power System State Estimation. First, the Weighted Least Square method and the Generalized Maximum-Likelihood Estimator using the Projection statistics method are used to estimate the voltage magnitude and phase angle. Secondly, the Feed-Forward Neural Network model is proposed to combine the obtained voltages and angles. The optimal structure of the proposed Feed-Forward Neural Network model is defined based on the Akaike Information Criterion. Thirdly, the Long Short-Term Neural Network model is proposed as an alternative hybrid power system state estimation approach. Finally, the different case studies including IEEE 9-bus system and IEEE 14-bus system are used to validate the effectiveness of the proposed approaches. The final results imply that the proposed approaches can provide more effective solutions than the existing approaches according to Mean Absolute Percentage Error and Weighted Average Percentage Error criteria.