Climate-responsive machine learning-based control of switchable glazing towards human-centric lighting

Abstract

This work presents a climate-responsive control of switchable glazing using machine learning models. Here carried out experimental investigations in daylight artificial light integrated room on Polymer-dispersed liquid crystals (PDLCs) for collecting data towards modelling. A support vector machine learning classification algorithm is designed to model the transparency change in PDLC. A tunable LED luminaire is used to adjust the brightness and circadian effectiveness in the test room. The real-time implementation of the model gives the states of switchable glazing under various climate conditions. Real-time experimentation is carried out to verify the results. It is observed that the Interior light colour characteristics were satisfactory under different PDLC states. Correlated colour temperature (CCT) and circadian stimulus (CS) were similar to daylight. In the available literature, simulation results show PDLC as satisfactory for visual comfort. However, the experimental investigations and the prediction models give a range of sunlight on the window and solar altitude for which it satisfactorily works; outside this range, PDLC acts as a luminous source. The feasibility of PDLC in all window orientations is also simulated. The Simulink model can give the states of switchable glazing by optimising visual comfort, thermal comfort and energy effectiveness.