Load modeling is a fundamental issue in the operation and control of power system, which has very important significance in theory and practice. First, this study reviews existing load modeling achievements, provides an overview of existing load models and methodologies for parameters estimation, identifies current industry practice on load modeling for power system, and summarizes the shortcomings of existing load modeling theory. Then, future prospects are analyzed, including new challenges and new requirements for modeling work resulted from major changes in load characteristics and grid structure, and new opportunities with the rapid development of artificial intelligence and big data. Finally, the future work is summarized as a twostep problem that is “qualitative correctness” of model structure and “quantitative accuracy” of model parameters. The structure of a generalized synthesis load model is proposed considering the active load. To establish the longterm mechanism of load modeling, a “type classified and time classified” load model database, taking the advantage of a variety of online modeling methods as well as the artificial intelligence algorithm, is proposed.