As the increased frequency, intensity and duration of extreme weather events have significantly challenged power systems, greater attention has been focused on the development of resilient power systems. Taking a physical-cyber-human system perspective, this paper establishes a multi-criteria resilience evaluation framework for urban power systems, in which two principal elements responsible for power system function degradation are described, and fifteen (eleven objective and four subjective) power system resilience evaluation indicators are identified. Fuzzy hesitant judgement and a TOPSIS aggregation method are applied for the evaluation to minimize expert divergence and maximize group consensus. The evaluation method is then applied to four Chinese municipalities: Shanghai, Beijing and Chongqing, and Tianjin. It was found that Beijing’s resilience was the best of the four but overall the urban power system resiliencies were not enough in the face of extreme event challenges.