International Journal of Engineering and Technology

Power networks modernization toward smart grids has encouraged the use of smart meters, opening opportunities for the development of new applications to improve energy efficiency of consumers, as nonintrusive load monitoring. This paper presents a nonintrusive load identification and characterization model with real power and impedance by using artificial neural network. The approach involved the use of real power and impedance measures obtained from an educational building to represent the consumption behaviour of loads and to identify the operation states. Transitions between operation states of devices are identified as events. Common operation states transitions were used to create a representative set of events to train the multilayer feed-forward neural network designed. Comparisons between the conventional power model and the proposed mixed power-impedance model were achieved and results showed that the proposed method was better to identify loads and characterize the consumption behaviour. This model has been proposed to detect load variation in real applications using smart meters.