Data Driven Solar Forecasting Model for Northern Ireland

Solar forecasting is an increasing problem for power system operators due to its variability from day to day. This inconsistent power generation effects not only load forecasting, but can cause increased network stability problems due to changing system inertia. Ample research has been carried out surrounding wind generation, with several accurate forecast methodologies created and used across the world. However, as the solar industry is only now beginning to take off in Northern Ireland (NI), with its capacity increasing exponentially every year, forecasting methods have not yet been accurately implemented by the System Operator for Northern Ireland (SONI). This report documents the creation of a NI specific photovoltaic (PV) forecasting model, based on meteorological and historic PV generation data, from data collection to testing. It was found that partial least squares out-performed multiple linear regression and therefore was used as the forecast base. A large scale model was created, achieving an average Root Mean Square Error (RMSE) of 11.9MW and MAE of 7.7MW. The forecast also improved accuracy with the addition of temperature, wind speed and cloud cover data. A small scale forecast was also built and up-scaled to its representative capacity on the NI system. The model was tested using NI System Demand data, achieving positive results.