Abstract: For a heat and power cogeneration unit, analyses were conducted on energy efficiency characteristics based on adaptive multi-level filtering fuzzy C-means (FCM) algorithm with the optimal clustering number. Through improving the traditional clustering algorithm, an adaptive optimal clustering method was proposed based on multi-level strict screening mechanism, so as to obtain the energy efficiency characteristics of unit under different load conditions. The curves of unit operation characteristics were fitted by least squares method, while comparisons and analyses were carried out based on the surfaces of unit operation energy efficiency characteristics from Newton interpolation algorithm. Results show that, the unit performances have certain differences with different external constraints. From the surfaces of energy efficiency characteristics, the difference between the energy efficiency characteristics of two units under different loads and the load condition with a better operation performance of each unit can be visually displayed. Under the condition of 225 MW electric load and 72.5-88.5 t/h heat load (heat-supply steam mass flow rate), the net coal consumption rate of No.1 unit is higher than that of No.2 unit, and with the increase of heat load, the difference in coal consumption rate firstly increases and then decreases. Therefore, No.2 unit can be prioritized to operate under low load condition.