Using machine learning techniques for the sizing energy storage systems coupled to offshore windfarms

Abstract

Machine learning techniques have been extended to the use of Measure-Correlate-Predict methodologies. The same techniques are also used for energy demand forecasting. MCP techniques and energy demand forecasting are combined to determine the optimum size of an Energy Storage System, which is coupled to an offshore windfarm and an onshore electricity grid. Three machine learning methodologies are compared in this study. Each methodology is used to create two time series data sets which are the predicted wind speed, from which the energy yield from the windfarm is derived, and the forecasted energy demand. The time series are used to predict the behaviour of the windfarm and the ESS. This resulted in the establishment of a relationship between the curtailed energy from the wind farm and the ESS capacity. The optimal ESS capacity is derived from this relationship. The optimal ESS capacity is derived for both measured wind speed and measured energy demand time-series, and from the predicted time series to estimate the error in the ESS capacity.