Application of Long short-term Memory Neural Networks in Dynamic State Estimation of Generators Subjected to Ageing in Complex Power Systems

In this paper, Long short-term memory(LSTM) neural networks based techniques for estimating dynamic states of generators in highly complex power systems is presented. It is proven that time-series prediction techniques can be used for dynamic state estimation. The most benefit that proposed method offers, is its independency from the mathematical model of the generators. The results proves superiority of the proposed technique over particle filter and unscented Kalman filter when parameters of the generators alter. The proposed scheme sustain its accuracy and precision even in the presence of unobservable variances in generator parameters. Parameter alterations in generators usually happen due to ageing of the equipment and environment impacts, and so on.