A high accurate user-friendly energy audit platform of a university building using ANN Bayesian regularization and Levenberg-Marquardt algorithm

The population is directly proportional to the energy demand. With the development of the economic level, the demand for electrical energy also increases. In order to achieve effective planning and investment, it is crucial to accurately ascertain the energy demand. This may be accomplished by utilizing reliable software that can predict energy usage. In this context, the integration of Artificial Intelligence (AI) emerges as a transformative force, promising unparalleled precision, speed, and depth in analyzing and optimizing energy consumption. With the increasing use of AI in various industries, there is a need for further research on designing an energy efficient building. This study explores the compelling reasons why AI is needed in energy audits to revolutionize our approach to sustainable energy practices. This paper presents the development of an electrical energy audit application using MATLAB (R) R2020a platform, to display the energy consumption in the selected building of one State University in the Philippines. The researchers employed empirical research. The conventional walk-through process in energy audit was used to collect data as a basis for the model training parameters. The predictability accuracy of artificial neural networks (ANNs) can vary depending on the specific problem and dataset they are applied to. In achieving best validation, in voltage unbalance, BR (0) and LM (616); for the current unbalance, BR (996) and LM (239); for the lighting consumption BR (32) and LM (868); for the plug loads BR (1000) and LM (75). In the training state, BR shows more stable than LM. In the regression plot (All), in the voltage unbalance, BR (0.99463) and LM (0.99012); for the current unbalance, BR (0.92784) and LM (0.96943); for the lighting consumption BR (0.9925) and LM (0.99888); for the plug loads BR (1) and LM (0.91329). In this study, the indicative results have shown that Bayesian Regularization ANN (BRANN) training technique had a better performance than Levenberg-Marquardt algorithm. It is concluded that BRANN reveals potential complex relationships and can provide a robust model. This study introduces ANN in lighting design and load balancing for an energy efficient building. Further study on using ANN in voltage and current balancing is recommended.