Experimental Assessment of Markov Chain Models for Data-Driven Voltage Forecasting

Regulation of voltage magnitude is one of the fundamental activities required to ensure safe and effective operation of the network, a task that is complicated by the large-scale integration of distributed energy resources at all system levels. Recently, data-driven (model free) forecasting of voltage magnitude has been emerging as a research area that may help system operators to improve situational awareness. In this paper, a probabilistic data-driven voltage forecasting methodology based on Markov chain models is proposed and experimentally assessed. The methodology is characterised by low input data requirements and three applications are considered and evaluated: a deterministic forecast, a probabilistic forecast, and an alarm system to provide early warning of possible voltage excursion events. The methodology and applications are demonstrated using real data from part of the Italian sub-transmission network for forecasting horizons of up to four hours.