Machine Learning Forecasting for Optimisation of Green Energy Generation in Non-domestic Buildings

Abstract

To mitigate their carbon print, buildings must have on-site renewable energy generation systems to supply energy for the buildings without relying on the national grid. Unfortunately, the unreliability of renewable sources such as solar and wind energy makes it difficult to rely on them for the only source of energy as they depend on many weather features. If the power output of the renewables can be accurately forecasted, a building management system (BMS) can be equipped to optimise on-site renewable energy generation. Various solutions are available for the prediction of renewables but, current BMSs don’t equip forecasting and optimisation algorithms for demand and renewables. This work presents a method to forecasting renewable generation through Random Forest (RF) and Neural Network (NN) machine learning classification and regression using collected weather data and building features. Data was collected from an operational University campus located in central Manchester-North West England-for validating the developed method. Results show that solar energy is able to be forecasted with less than 7% prediction error with wind and additional renewable sources theoretically tested with the same method. This proves that the developed method is an effective tool for mitigating carbon prints of public buildings.