Harmonic Source State Identification Using Random Forest

Identification of harmonic source types in the power grid is a prerequisite for comprehensive system rectification solutions to address grid harmonics and is of great significance. This article investigates the problem of harmonic pollution in the power system and proposes a machine learning-based method using a random forest model to identify five types of harmonic sources, including single-phase rectification, three-phase rectification, inverter rectification, wind power generation, and photovoltaic power generation. By measuring the current and voltage characteristics of harmonic sources and combining the accuracy and interpretability of the random forest algorithm, classification of harmonic source types is achieved. Experimental results show that the classification accuracy of wind power generation and photovoltaic power generation harmonic sources is 100%, while the classification effects of single-phase rectification, three-phase rectification, and inverter rectification harmonic sources are 90.5%, 98.2%, and 95.8%, respectively. Overall, this method has achieved good classification effects for harmonic sources.