Deep Reinforcement Learning–Based Multiobjective Control of Hybrid Building Envelopes: Integrating Adaptive Façades, Passive Radiative Cooling, and Indoor Air Quality Management Across Iranian Climates
Randomized trial assesses a deep reinforcement learning system improving energy efficiency and indoor air quality in various climates, indicating significant health benefits.
Key Points
This research aims to enhance energy efficiency, thermal comfort, and indoor air quality in buildings through advanced control systems.
Utilized EnergyPlus–Python cosimulation across four Iranian climates (cold, hot-dry, humid, temperate)
Compared configurations: baseline, hybrid envelope with rule-based control, static envelope with DQN, and intelligent hybrid
Implemented a deep reinforcement learning agent to manage adaptive façades and passive cooling systems.
Intelligent hybrid system reduced annual energy use intensity by 31% and thermal discomfort by 70%–75% in cold and hot-dry climates.
Pollution-aware window control decreased PM 2.5 exposure by 48% during dust storms compared to baseline.
Health assessment estimated prevention of 5.8 respiratory illnesses and improved cognitive performance valued at 54–81 million IRR per worker/year.