Long-Range Inertia Prediction Considering Contemporary Evolution of Power Grid Networks

Abstract

Reduced network inertia due to high penetration levels of nonsynchronous generators in modern power systems is becoming a pressing issue. As a result, very quick inertial responses are observed after contingency events in networks. Due to quick inertial responses, there is a practically very limited time interval for control actions in real-time. Thus, system operators need to understand the prior inertia values to plan, control, and operate the network securely. Long-range forecasting of the network's inertia values, in contrast to short-range forecasting techniques, can pinpoint when the network is most likely to be vulnerable in a reasonable time ahead. Thus, in this research work, an improved ARIMA model (i-ARIMA) technique for long-range forecast inertia values in a modern network is proposed. To estimate future inertia values over a long period of time, the i-ARIMA model leverages strong periodic and seasonality characteristics of previous time series data. The i - ARIMAmethod is tuned for optimal values of a moving observant predictor P, periodicity and seasonality factor s and smoothing factor n that give the best forecasts with competitive accuracy. Rigorous evaluation and tests of the method, which are performed on the New Zealand network data using the Power Factory DigSilient platform, demonstrate that the proposed i-ARIMA is quicker, more reliable, more accurate, and better than other conventional forecasting methods.