A novel scheme to supervise the harmonic pollution effectively and reasonably in complex power networks was proposed based on the decision tree algorithm in this paper. It employs the inverse hierarchical topology search method to disintegrate the power networks into two parts, the main looped network and radial subnets, and then take the boundary nodes between two parts as the investigated point. The active power direction method is adopted to determine the harmonic pollution is mainly from the upper side or the lower side. The moving average linear filtering method is further used to extract the fast-varying components of the harmonics to calculate the harmonic impedance and corresponding harmonic emission voltage at the investigated point caused by the each equivalent harmonic source. Finally, the decision tree algorithm is employed to complete the equivalent harmonic source data classification, and to obtain the main influencing factor of the equivalent harmonic source among the related influence factors, and then to formulate the corresponding punishment mechanism. The proposed scheme simplifies the harmonic source regulation of a complex power grid, and overcomes the difficulties of the present harmonic power flow analysis. Simulation results prove the effectiveness of the proposed method.