Implementation and visualization of artificial intelligent ventilation control system using fast prediction models and limited monitoring data

Abstract

Coupling modelling and monitoring has always been a very challenging task aiming for indoor air pollution control towards the construction of safe, healthy and energy-efficient building environments. Currently, numerous fast prediction models were proposed but facing a dilemma of prediction accuracy and speed to satisfy the engineering application. This work will systematically introduce the establishment of dynamic ventilation online control system, which relies on a newly developed faster-than-real-time prediction model by incorporating limited monitor data, and further put it forward to the real application. An experimental ventilation chamber was considered. To rapidly predict non-uniform indoor pollutant concentration, the fast prediction method was adopted by incorporating monitoring data (ZigBee-based monitoring sensors) to LLVM (low-dimensional linear ventilation model)-based ANN (artificial neural network) model. Visualization of the control effect was also displayed. It was found that based on the evaluation index EV, the ventilation control system could show favorable performance with energy conservation and indoor pollutant level (e.g., CO2) respectively decreasing up to 43.8% and 28%. This work will provide a significant guidance for the development and application of intelligent ventilation online control system in the perspectives of constructing healthy and energy-efficient indoor environments for sustainable buildings.