Radiant heating systems control in buildings via Inverse Conformable Artificial Neural Networks and optimization techniques

Abstract

This study introduces an innovative methodology that integrates Inverse Conformable Artificial Neural Networks (CANNi) with Genetic Algorithms (GA) or Particle Swarm Optimization (PSO) to optimize thermal comfort in buildings. Emphasizing the efficacy of conformable transfer functions within CANNi, the research relies on a comprehensive dataset to forecast heat transfer across diverse climates. Notably, the methodology stands out for its meticulous model selection process, employing slope-intercept tests to ensure robust predictability (99%) and strong correlation between input and output variables. The selected model exhibits an optimal equilibrium between predictive precision and computational efficiency, reaching an R-value of 0.9992 with low RMSE (0.0078) and MAPE (2.1099). Such performance enables precise calibration of Radiant Floor Heating Systems (RFH) for enhanced comfort and marks a significant stride toward bolstering energy efficiency and sustainability within the construction sector. These findings advocate for more efficient utilization of energy resources in buildings, adeptly accommodating climatic fluctuations and enhancing the inhabitants’ quality of life.