Adaptive lighting control in intelligent buildings using artificial neural networks: design, implementation, and experimental validation

This study presents the design, implementation, and experimental validation of an intelligent lighting control system based on artificial neural networks. The proposed system optimizes energy consumption while maintaining the required illuminance levels by adaptively adjusting artificial lighting in response to daylight variations and occupancy conditions. The control framework employs a feedforward neural network trained offline and deployed on an embedded lighting control unit. Rather than introducing a new neural network architecture, the work focuses on the embedded deployment and long‑term field operation of an ANN‑based controller in a real office environment. Experimental validation in a real‑world intelligent building demonstrated consistent energy savings of 7.4-12.5% during continuous operation and 8.5% under occupancy‑driven conditions while maintaining the required illuminance levels. The ANN model achieved high predictive accuracy (R² = 0.91) and stable performance under dynamic lighting scenarios. The implemented control algorithm ensured smooth dimming transitions and reduced abrupt illuminance fluctuations typically associated with conventional switching systems. The architecture enables real‑time operation on low‑cost hardware without high‑performance computing resources, making it suitable for retrofit applications. Overall, the findings confirm that ANN‑based adaptive lighting control provides a practical, scalable, and energy‑efficient solution for modern intelligent building automation and provide a basis for further validation across different building types and climatic conditions.