Aiming at the problems of single dimension and lack of completeness in assessing the short term service efficiency of wind turbines, a method for assessing the short term service efficiency of wind turbine and cluster is constructed based on SCADA data. First, the core concept of short term service efficiency is defined, and the calculation method of short term service efficiency is determined as the ratio of the actual power generation to the expected power generation in a specific period of time. Second, a power prediction model based on the combination of heuristic optimization algorithm and machine learning is proposed to accumulate the power prediction values in order to provide the short term expected power generation. Finally, using data from an actual wind farm, the power prediction model is verified to have good prediction results, and the results of the short term service efficiency of single wind turbine and cluster are visualized in the form of wind power scatters, which verifies the validity of the proposed short term service efficiency calculation method.