Abstract: Accurate and reasonable assessment of power quality is of great significance to a fair power market. In view of the fact that power quality assessment are greatly influenced by artificial factors, the power quality comprehensive evaluation model is set up based on random forest with the help of high accuracy of random forest to deal with classification problems. Power quality is divided into five grades according to its six evaluation indexes, and then the random forest sample set is generated. Multiple power quality comprehensive evaluation tree classifiers is built based on the sample set and the rating system for the power quality is achieved by voting. The evaluation model is verified by two study cases. This model avoids the human disturbance in the evaluation process with the strong generalization and anti-noise capability. Therefore, it has a good application prospect in the comprehensive assessment of power grid power quality.