Prediction of office building electricity demand using artificial neural network by splitting the time horizon for different occupancy rates

Due to the impact of occupants’ activities in buildings, relationship between electricity demand and ambient temperature will show different trends long-term short-term, which seasonal variation hourly variation, respectively. This makes it difficult for conventional data fitting methods accurately predict short-term power buildings at same time. In order solve this problem, paper proposes two approaches predicting office buildings. The first proposed approach splits into fixed time periods, containing working hours non-working hours, reduce activities. After finding most sensitive weather variable hour demand, building baseload occupant can be predicted separately. second uses artificial neural network (ANN) fuzzy logic techniques fit baseload, peak load, occupancy rate with multi-variables variables. approach, is split a narrower range as no-occupancy full-occupancy them, varying depending on are verified by real from University Glasgow case study. simulation results that, compared traditional ANN method, both have less root-mean-square-error (RMSE) demand. addition, based regression reduces average RMSE 35%, while 42%, comparing prediction method. provide more information energy management, including rate, without requiring additional parameters.