Energy Forecasting Using Intelligent Models

This paper deals with the problem of intelligent energy forecasting in hotels. The aim is to provide reliable predictions of hotel energy demand to increase the efficiency of the management systems responsible for the energy resources planning. The immediate benefits will be related to reduction of energy consumption, decrease in associated greenhouse gases emissions and improvement of the sustainability indexes of hotel activities. The prediction algorithm is based on the use of intelligent methods. The objective is to provide a 24 h-prediction of consumed energy with a look ahead of 24 h. LSTM and GRU networks were used in the algorithm for this task. The algorithm, that includes the main specific variables affecting the consumption, is endowed with some important capabilities that improve the existing methods. In particular, the proposed model is able to adapt online to changes while maintaining a balanced trade-off between accuracy and simplicity. The evaluation of the proposal was done in a luxury hotel in Canary Islands. The results obtained show promising results for a real-time implementation of the method in an energy management system.