Please login first
Energy Efficiency Through an On-Line Learning Approach for Forecasting of Indoor Temperature
Abstract: University CEU Cardenal Herrera (CEU-UCH) has constructed a Solar house, known as SMLSystem, to participate in the Solar Decathlon Europe 2012 competition. Such construction becomes a research facility that University employs in order to test innovative solutions around the area of energy efficiency. A lot of technologies have been integrated to help to reduce the overall power consumption of the house. Among them, a predictive system, based on Artificial Neural Networks (ANNs), has been developed using the data acquired in Valencia, where the house is placed. Such system produces short-term forecast of indoor temperature, using as input the data captured by a complex monitoring system. The system expects to reduce the power consumption mainly related to Heating, Ventilation and Air Conditioning (HVAC) systems because of the following assumptions: the high power consumption for which HVAC is responsible (53,9% of the overall consumption); and the energy needed to keep the temperature is less tan the energy required to lower/increase it.This paper studies the viability of the development of such kind of predictive systems but for totally unknown environments, that is, without historical data. To do that it is possible to apply on-line learning approaches, where the model parameters are estimated following Bayesian methods or Gradient Descent (GD) methods, starting from an unbiased a-priori knowledge, or from a totally random model. These forecasting measures could allow the house to adapt itself to future temperature conditions by using home automation in an energy-efficient manner. Preliminary experimental results show a high forecasting accuracy with simple models and with a short training time of 4-5 days. The final idea is to develop intelligent agents, with the minimum resources, to be implemented in very cheap computer architectures.
Keywords: Energy efficiency; Time series forecasting; Bayesian models; Gradient descent; Artificial neural networks