A non-intrusive load monitoring system that estimates the behavior of individual electrical appliances from the measurement of the total household load demand curve is useful for the forecast of electric energy demand and better customer services. Furthermore, this system will become important for power companies to control peak electric energy demand in the near future. We have already reported the system using Support Vector Machines (SVM) and SVM could establish sufficient accuracy for the non-intrusive load monitoring system. However, SVM needs too much computational cost for training to establish sufficient accuracy. This paper shows Kernel based Subspace Classification can solve this problem with an equal accuracy of classification to SVM.
展开▼