Abstract
The prediction of electricity demand plays an essential role in the building environment. It strongly contributes to making the building more energy-efficient, having the potential to increase both thermal and visual comfort of the occupants, while reducing energy consumption, by allowing the use of model predictive control. The present article focuses on the use of computational intelligence methods for prediction of the power consumption of a case study residential building, during a horizon of 12 hours. Two exogeneous variables (ambient temperature and day code) are used in the NARX model Two different time steps were considered in the simulations, as well as constrained and unconstrained model design. The study concluded that the smaller timestep and the constrained model design obtain the best power demand prediction performance. The results obtained compare very favourably with similar approaches in the literature
Highlights
The present study aims to discuss the prediction of energy consumption in a residential building, having as a case study the Honda Smart Home, located in Davis, United State
To develop the present study, and based on the background information previously presented, from the data set provided by the Honda Smart Home (HSH) group, two parameters will be used from the HSH data set
The Artificial Neural Networks (ANN) structure used in the present work uses as model type Radial Basis Function (RBF)
Summary
The present study aims to discuss the prediction of energy consumption in a residential building, having as a case study the Honda Smart Home, located in Davis, United State. The objective is to develop an accurate prediction model, in order to be subsequently used in decision making and model predictive control. Section two presents the case study, the Honda Smart Home architecture and the dataset available by their experimental campaigns. The data used as input for the models developed by this work are presented . Section three presents the predictive model design, concerning problem formulation, the prediction horizon validation method, and design experiments. Results are shown and discussed in Section four, and Section five presents the conclusions and points out future research
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
