Building energy prediction using artificial neural networks and analysis of covariance in the six thermal zones of Morocco

Abstract

Energy management is essential in light of the current energy issues, particularly in the building sector, which accounts for a sizable amount of global energy consumption. Incorporating intelligent energy management solutions in this sector is one of the answers that could assist the nation in achieving its goals for environmental changes and overcoming its energy challenges. Energy prediction in buildings is therefore essential for efficient management. The present paper aims to evaluate the performance of two different approaches: Artificial intelligence (AI) and General Linear model (GLM) approaches, in predicting heating energy consumption of an administrative building for which localization concerns the six Moroccan thermal zones. The chosen models for AI and GLM are respectively ANN and ANCOVA. In the TRNSYS environment, building energy simulation was carried out in order to produce a database for the models training and validation. Only two meteorological data were used: External temperature and internal temperature. While the ANN model was trained on mixed data of different thermal zones, ANCOVA model established an equation for each thermal zone separately, with respect of model hypothesis. Results show that ANN model outperforms the ANCOVA model with a correlation coefficient of 0.95 and 0.90 respectively. Thus, the use of Artificial Neural Network for building energy prediction is more accurate than ANCOVA model.