Daylight Illuminance Prediction in a TehranContemporary Building Using a MLP MachineLearning Algorithm

One of the most important issues today is the preservation of the environment throughenergy consumption reduction, with a significant portion of global energy consumptionattributed to residential buildings. Among these consumption patterns, a considerableamount is allocated to electricity consumption for artificial lighting or reducing heatingload resulting from daylight. Therefore, predicting daylight illuminance as a controllingfactor in lighting energy consumption in each thermal zone of a building is essential.Accurate prediction can assist civil engineers, architects, or urban designers in buildingenergy management, both during the design phase prior to construction and postconstruction.One of the best methods for predicting daylighting illuminance is fitting amachine learning model using multi-layer neural network regression, where selectingappropriate geometric, physical, and climatic factors as input features can achieve goodaccuracy for predicted daylight illuminance values in each zone of the building within thetrained region's climate. In this study, the final model evaluation of simulating daylightillumination in buildings using a multi-layer perceptron regression neural network led tostatistical performance metrics such as coefficient of determination (R²) = 0.8651, rootmean squared error (RMSE) = 233.25, and mean absolute error (MAE) = 115.27 [lux].

Daylight Illuminance Prediction in a TehranContemporary Building Using a MLP MachineLearning Algorithm Keywords:

Daylight illuminance , Machine Learning , Multilayer perceptron , Tehran , building