Building Consumption Modeling for Fault Detection Analysis

The capability of different Artificial Neural Networks Ensemble (ANNE) approaches for artificial lighting fault detection of a real office building was demonstrated. Experimentation was carried out using the monitoring data sets (consumptions, occupancy, weather conditions) of an actual office building located at ENEA ‘Casaccia’ Research Centre in Rome. Using a fault free data set for the training, the ANNEs were used for the estimation of hourly lighting energy consumption in normal operational conditions. The fault detection was performed through the analysis of the magnitude of residuals generated by ANNE using a peak detection method. Moreover the peak detection method was applied directly to the testing data set. Finally, a Majority Voting method to ensemble the results of different Artificial Neural Network (ANN) classifiers was performed. Experimental results showed the effectiveness of ensembling approaches in automatic detection of two particular abnormal building lighting energy consumptions created in the testing period.

For more information see:

Lauro, Fiorella; Capozzoli, Alfonso; Pizzuti, Stefano, “Building Energy Consumption Modeling with Neural Ensembling Approaches for Fault Detection Analysis”, Sustainability in Energy and Buildings: Research Advances, KES Open Access LibrAry, Special Edition – Mediterranean Green Energy Forum 2013 (MGEF-13), Vol. 2, pp. 7-12, 2013.

URL: http://nimbusvault.net/publications/koala/SEBRA/166.html