On the basis of the uncertainty theory

a multi-variable steady-state detection method is proposed. This method extracts steady state information by controlling the influence of each main operating parameter on random uncertainty of the performance index within a specified limits

which can adaptively determine the steady-state interval length. Moreover

this method is applied to carry out steady-state detection for operating data of an ultra-supercritical 1 000 MW thermal power unit. The results show that

the power and feed water flow are key indicators for the system's steady state evaluation

the performance index calculated by the detected steady state condition has high accuracy and good consistency

which can be applied for performance monitoring and diagnosis of steam turbines. By comparing the steady-state detection results with the sliding window method

it finds that this method has higher detection sensitivity and accuracy. Therefore

the steady state detection method proposed in this paper lays a foundation for thermal power units' performance monitoring and evaluation

fault detection and diagnosis by using historical operational data

which alleviates many problems faced by current steam turbines' thermal performance tests and has certain engineering application values.