Self-adaptive and Self-learning Lighting System: Integrating LSTM and RL for Energy Efficiency and Personalized Visual Comfort

Optimizing lighting systems is crucial for reducing energy consumption and enhancing occupant well-being in sustainable building design. A key challenge is creating energy-efficient lighting systems that adapt to individual users' visual comfort needs. This paper proposes a two-phase approach for a self-adaptive and self-learning lighting control system. In the first phase, Long Short-Term Memory (LSTM) networks optimize the placement of photosensors by modelling dynamic lighting conditions over time. In the second phase, Reinforcement Learning (RL) enables real-time adaptation of lighting based on occupant preferences, maximizing energy efficiency and visual comfort. This system ensures personalized, efficient lighting in office environments while minimizing energy waste.