Short-term load forecasting with high levels of distributed renewable generation

Traditionally, load forecasting tools include weather variables as model inputs. Northern Ireland has seen a major increase in weather dependent, renewable generation over the last number of years creating a double impact by weather parameters in the load profile. The new generation is not visible to, or controllable by, the system operator and is presenting major challenges to traditional load forecasting tools which are struggling to account for the fluctuations in demand. In this paper a simple linear method, using the previous week demand value and a correction for the weather and renewable generation is investigated within a sliding window parameter updating framework. Selection of variables are based on correlation analysis of supplied and derived meteorological variables. Model performance is evaluated on load data for the period 2011-16, split into several periods due to different levels of installed renewable energy. A 4-input model with parameters updated on the basis of a 44-day sliding window of historical data, is shown to give a mean absolute percentage error of 2.6% overall.