The invention discloses a method and device for discriminating collusion behaviors of power generators and a medium. The method comprises the steps that quotation curves of the power generators are acquired; wherein the quotation curve reflects the relationship between the quotation capacity and the electricity price of the unit; carrying out segmentation quantification and per unit processing onthe quotation curve; wherein the quotation capacity of the unit is based on the unit available capacity, and the electricity price is based on the electricity market price upper limit; According to aset difference integral formula, calculating a representation vector of a quotation curve after segmentation quantization and per unit processing, namely a quotation sequence of each power generator;calculating the degree of association between every two quotation sequences of the power generators, and constructing a degree of association matrix according to the degree of association; performingclustering analysis on the correlation degree matrix, and screening out a power generator group with the correlation degree greater than or equal to a correlation degree threshold; And searching related factors of the power generator set, carrying out collusion analysis on the power generator set according to the related factors, and outputting an analysis result. The method can accurately discriminate the collusion behavior of the power generator in the power spot market transaction.