As the existing power system peak load regulation scheme is designed based on the experience of dispatcher mainly, it has the disadvantages of strong subjectivity, inaccurate scheme evaluation system and so on. Coal consumption, power purchase cost and SO2 emission are used as power system peak load regulation scheme evaluation indexes. Combined the attribute recognition mode of ordered segmentation problems with the entropy value of information theory, the power system peak load regulation scheme evaluation system is proposed based on the entropy weight attribute recognition model. Firstly, the attribute space of the peak load regulation scheme is set up and partitioned in order. After using entropy theory to determine the evaluation index coefficient, the attribute measurement of peak load regulation scheme can be worked out with weight sum. Finally, by confidence criterion, the investigated scheme is evaluated. In the numerical examples, Guangdong power grid is used as application background, peak load regulation schemes of six typical days are evaluated to testify the proposed evaluation system.