With the rapid advance of the spot market, it is critical to identify the abuse of market power accurately and in a timely fashion. This paper combines an improved support vector machine with a variational inequality solving algorithm to realize the accurate identification of the abuse of market power by power generation enterprises when only a few power generation enterprises have label data. First, based on the actual situation of the power market, it constructs an index system to identify the abuse of market power, and reduces the dimension of the high-dimensional data of the power market. Secondly, in view of the fact that the label data of power generation enterprises account for a small part of the total data and the data are unbalanced, a method to identify the abuse of market power of power generation enterprises based on the improved cost-sensitive transductive support vector machine is proposed. Considering the long solution time of the semi-supervised algorithm, the problem is transformed into an efficient solving problem of variational inequality, and a customized proximal point algorithm is used. Finally, UCI data set, power market simulation data and actual power market data are used to carry out experiments. The results show that this method can quickly and effectively identify power generation enterprises that abuse market power. This work is supported by the Science and Technology Project of State Grid Corporation of China (No. 52094020001A).