Load disaggregation in non-intrusive load monitoring based on random forest optimized by particle swarm optimization

• Published in: 2017 IEEE Conference on Energy Internet and Energy System Integration (EI2) pp. 1 - 6
• Main Authors: Feixiang Gong, Chang Liu, Linru Jiang, Hao Li, Lin, J. Y., Bo Yin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-11-2017
• Subjects: Classification algorithms; Home appliances; Load disaggregation; Monitoring; NILM; Particle swarm optimization; Power demand; Power difference; PSO-RF; Switches; Transient analysis
• DOI: 10.1109/EI2.2017.8245609

Fine-grained load disaggregation based on smart metering plays an important role in energy consumption optimization on appliance level. This paper proposed a load disaggregation in non-intrusive load monitoring based on random forest optimized by particle swarm optimization(PSO-RF) in order to real-time monitor and recognize the type and working state of smart appliances in the smart grid user side. The method was mainly based on event detection and multi-feature classification system to do load disaggregation. Firstly, the method of identifying the switching state of power fluctuation signals based on the power difference was proposed. Then, the multi-feature parameters of the switching events were extracted to train the PSO-RF in order to identify the type and switching the state of the appliance. The experimental results showed that the recognition accuracy can approach to 98.9% for switched-on state and 97.5% for the switched-off state, which are both higher than GA-BP and GA-SVM.