Nonintrusive load disaggregation is a kind of method for recognizing the state of appliances by using the current and voltage in the power bus. However, the obtained results are usually not coincided with the actual results, because of the fluctuation of current and voltage. To promote the performance of disaggregation, the effective features are built by the harmonic of current and the feature of real power. And particularly, the measurement function which is utilized to combine the harmonic of current and the feature of real power together is introduced into the fitness function inherent in the particle swarm optimization (PSO) algorithm, thus finding the optimal results of energy disaggregation using PSO algorithm. Finally, experiments are performed on the hardware equipment developed by our Labs. The results demonstrate the good performance for energy disaggregation online.