The widespread deployment of renewable energies and non-linear loads has led to serious harmonic pollution in the electrical distribution networks. Evaluation of the harmonic contribution (HC) of each customer is a significant task for power quality management. Most previous studies focus on periodic evaluation methods, where numerous data have to be collected in advance over a period (e.g., one day). However, customer behaviors are time-varying and would lead to dynamic HCs, which can not be captured by traditional periodic evaluation methods. To address this issue, this paper presents a novel real-time HC evaluation method considering multiple dynamic customers. First, a two-stage iteration estimator is proposed based on the information fusion technique to quantify the real-time HC of each customer. Then, to mitigate the negative effect of unknown background harmonics, a dominant index method is developed to determine the credibility of the measurement data. On this basis, an adaptive gain selection strategy is proposed to improve the accuracy of real-time HC evaluation. By doing so, the major harmonic contributor can be identified for implementing harmonic suppression and improving power quality. Finally, a typical IEEE system is utilized to verify the proposed methods. The results show that using the proposed method, the evaluation errors can be reduced from about 10% to 2.5%. Moreover, the total harmonic distortion of voltage can be suppressed from 5.564% to 0.702%. Therefore, this research can be helpful for guiding harmonic problems in power systems.