A multivariate non-linear regression model to predict the energy demand for lighting in rooms with different architectural features and lighting control systems

The paper presents a set of logistic mathematical models to quickly estimate, since the earliest design stages, the electric lighting energy demand for a generic room with different architectural features (site, orientation, external obstructing angle, window size, glazing visible transmittance and room depth), lighting system characteristics or users’ lighting requirements (working plane illuminance, lighting power density, type of lighting control). The models were built upon the data obtained from a parametric study, based on simulations carried out through Daysim of 828 case-studies. Two dataset were obtained: one for rooms with a manual on-off switch lighting control and one for rooms with an automatic daylight responsive lighting control. Consistently, two specific models were derived, one for each control system. The average error of the mathematical models in estimating the lighting energy demand with respect to the results of Daysim simulations was quantified in terms of the normalized Mean Bias Error value (0.66% – for the manual control system model and 0.29% – for the automatic daylight responsive control system model) and of the Coefficient of Variation of the Root Mean Squared Error (3.45% and 5.34% for the two models, respectively).