The intake system is an important component for filtering air impurities in gas turbine power plants. Blockage of intake filter reduces overall thermal efficiency and may cause compressor stall, equipment damage, and unstable combustion. The existing method determines the severity of blockage based on intake system pressure difference, which is also affected by the opening position of compressor inlet guide vanes, atmospheric temperature, etc., thereby easily leading to misleading diagnostic results. To address this issue, a quantitative diagnosis method for assessing the severity of intake filter blockage was proposed for the first time. First, a dimensionless mathematical model of intake system health parameter, represented by measurable parameters, was derived from the relationship between component fault modes and fault symptoms of measurements. Second, using historical measurements taken when the intake system was healthy, a deep learning model was designed to calculate the theoretical healthy pressure difference of the intake system. Subsequently, a diagnostic strategy for variable operating conditions was proposed. Finally, both steady-state and transient simulation experiments show that the diagnosed health parameter effectively reflects the fundamental physical changes in the blockage status of the intake filter with improved robustness, and the accuracy of quantitative diagnosis exceeds 97%.
