This paper analyzes the dynamic nature of real-time traffic network performance measurement, with particular focus on random traffic flow fluctuation, traffic deferred conduction effect and traffic periodicity. The limitations of the traditional traffic network evaluation methods based on traffic V/C is discussed; followed by the presentation of a novel approach for real-time traffic network performance evaluation, which solves the previous technical issues such as data collection and processing with no blind region, dynamic spatial-temporal traffic network performance analysis and evaluation threshold calibration. The proposed approach is validated in a case study using the Beijing traffic network, which proves its rationality, efficiency and practicality. The implications of the research and methodologies presented in this paper provide new theories and perspectives for infrastructural planning, dynamic road network performance evaluation, etc.