Inertia Estimation of Islanded Power System With Distributed Generation Using Long Short Term Memory Algorithm

Due to enhanced penetration of renewable energy sources (RESs) in modern power grids, the inertia of power system has become a time-varying parameter. Moreover, estimating inertia using dynamic power system models is inappropriate, since converter-dominated grids exhibit very different dynamics than the conventional one. In this paper, the model includes Distributed Generation (DG) along with islanded thermal power system and is exploited to get local frequency measurements. The disturbance in the form of change in disturbance signal is generated by a pulse generator. Long Short Term Mem-ory (LSTM) algorithm, an extension of the Recurrent Neural Network (RNN), is proposed for estimating inertia using local frequency measurements. The study achieved a testing accuracy of 99.84 percent, while evaluating the prediction model.