Testing the Feasibility of Dynamic Mode Decomposition for Short Term Load Forecast

Concurrently with the electrification of many segments of modern life, such as transport and heating, the problem of short term load forecasting becomes more actual and important. For each power system management, it is crucial to have parameters that display how much electricity is consumed and generated, so the estimation of future consumption can be developed. During the last years, short term load forecast is obtained by various methodologies; each presenting their advantages compared to the others. In this paper we investigate the feasibility of short term load forecasting using dynamic mode decomposition (DMD). The main idea is to model the consumption space as a linear dynamic system using the measured data. A test data-set is collected from 2009 to 2011 in Ireland, consumption of power was recorded every half hour, the experiments performed by manipulating data, and arranging it for easier application of the algorithm. It is shown that using timestamps with DMD provides easy and fast computation. The obtained results are proven to be promising and comparable with others from the literature.