The air-conditioning of civil and industrial buildings during the startup and shutdown phases generally involves a certain degree of complexity. The evaluation of optimal starting and switching off times is a not easy choice, given the variability of the boundary conditions. The absence of a systematic rationalization of the management of the startup and shutdown of the plants does not allow to optimize the level of comfort in the environments compared to the real needs.
The ECO2A project, which stands for “Efficienza di Climatizzazione Ottimizzata con Auto Apprendimento”, aims to develop a system for the automatic control of the air conditioning during the startup and shutdown phases of the plants.
Temperature transients in air-conditioned rooms are influenced by a multitude of environmental and plant engineering factors: comfort is the result of many concomitant elements. The existence of many variables, which interact with each other, makes it impossible to achieve and maintain well-being through traditional methods.
The idea behind the ECO2A project is the self-learning of the evolution of these variables and, according to short-term weather forecasts, the determination of the optimal moment for the start-up and shutdown of the plants.
This kind of data acquisition is based on the adoption of Deep Learning logics, as the development of recurrent elastic and neural networks, with an architecture of the LSTM and GRU type.
Preliminary tests have shown energy savings of more than 50%, with consequent proportional reductions in carbon dioxide emissions.
The ECO2A project started in September 2017.