Urban microclimate prediction based on weather station data and artificial neural network

Abstract

Urban microclimate has a significant impact on building energy consumption. Building energy modeling (BEM) requires accurate local weather conditions near a target building, whereas Typical Meteorological Year (TMY) weather inputs often use remote airport weather data. An artificial neural network (ANN) model is presented in this study to predict urban microclimates based on long-term measurements from local weather stations near urban buildings and their significance in analyzing building energy consumption. By utilizing only a few months of data, the ANN model could connect local and remote meteorological parameters for a whole year. The 20-year historical weather data at the airport was then used to generate a local TMY. Based on the original and local TMYs, this study compared building heating and cooling loads. This method was evaluated for five weather stations within the city of Montreal to assess the impact of the local microclimate on the energy consumption of buildings. Based on locations, urban microclimate contributed to an additional 2 % to 14 % cooling energy consumption and a reduction of 1 % to 10 % winter heating energy consumption.