Spain. Researchers from the Degree in Computer Engineering of Information Systems of the CEU Cardenal Herrera University have developed a module, based on artificial neural networks, which allows predicting the temperature of homes and adjusting the air conditioning based on this prediction, so that energy consumption is more efficient.
As reported by the university portal, the module, which combines microchip systems and artificial intelligence techniques, has been developed by researchers Juan Pardo, Francisco Zamora, Pablo Romeu and Paloma Botella, from the ESAI Group (Embedded Systems and Artificial Intelligence) of the CEU-UCH.
Air conditioning is responsible for more than half of the energy consumption in homes, accounting on average for 53.9% of total consumption. Therefore, the researchers consider that predicting the temperature is the best way to optimize the energy necessary to acclimatize a home, since, as explained by the CEU-UCH researcher, Juan Pardo, "the energy required to maintain the temperature in a home is less than what is needed to lower or raise it".
The predictive module designed by the ESAI Group of the CEU-UCH brings together parameters such as the amount of CO2, the number of people who inhabit the house or the outside temperature. "These are some of the data recorded by the module to predict when it is convenient to activate the air conditioning or heating or anticipate a possible excess of consumption and thus be more efficient: we reduce the electricity bill and at the same time we respect the environment more, all automatically," says Juan.
The module is an intelligent agent for the generation of predictions, which works from the data obtained through artificial neural networks, implemented in electronic systems of small dimensions, which are connected to each other wirelessly through a network of sensors installed inside a house.
"This format allows you to reduce the amount of wiring that is installed inside a home, which also means a considerable extra cost," says Juan Pardo. On the experimental results obtained, he highlights that "the designed module has achieved precision in the prediction in a relatively short training time: between four and five days. The possibilities of a system like this, capable of learning on its own in a new environment, are unimaginable." Image: courtesy of CEU Cardenal Herrera.
