Artificial Neural Networks Based Parametric Curve Generation for Health Assessment of Industrial Gas Turbine Systems

• Published in: Process integration and optimization for sustainability Vol. 8; no. 2; pp. 577 - 590
• Main Authors: Raghavan, Shaiju M., Palatel, Arun, Simon, Jayaraj
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.05.2024; Springer Nature B.V
• Subjects: Algorithms; Alternative energy sources; Artificial neural networks; Combined cycle power generation; Combustion; Combustion chambers; Damage assessment; Degradation; Economics and Management; Energy Policy; Engineering; Feature selection; Filtration; Gas turbines; Gas-fired power plants; GE engines; Heat; Humidity; Industrial and Production Engineering; Industrial Chemistry/Chemical Engineering; Inspection; Machine learning; Neural networks; Original Research Paper; Overhauling; Parameter identification; Performance assessment; Power plants; Renewable resources; Sustainable Development; Turbines; Waste Management/Waste Technology; Compressor fouling; Gas turbine; Artificial neural network; Combustion chamber; Degradation assessment
• ISSN: 2509-4238; 2509-4246
• DOI: 10.1007/s41660-023-00372-5

Degradation assessment of gas turbine components in ‘off-design’ operating conditions is one of the main concerns of gas turbine users. This paper proposes a systematic methodology to promptly identify the fast degradation of gas turbine module components under any specified operating condition based on operational data of the machine. It is attempted to segregate the performance deviation due to part-load operation and due to other degradation effects, taking the case of typical General Electric Frame 9E machine components used in a combined cycle power plant in India. The development of artificial neural network (ANN) models of inlet filtration system, compressor, combustion chamber and turbine based on process history data, covering 60 to 100% load operation is described initially. The methodology for generation of parametric curves for performance assessment of individual components is illustrated further. It is shown that the sufficiently trained ANN models ( R 2  > 0.95) can be used for performance assessment of inlet filtration system, identification of fast rate of compressor fouling, combustion nozzle crack discovery and turbine angel wing seal damage. The degradation assessment tool assisted in the identification of the necessity for rectification works such as inlet filter replacement, compressor wash, combustion inspection and hot gas path inspection. The residuals of critical parameters identified in the study are found to be helpful in estimating the performance of components after machine overhauling. The proposed steady-state operational data based generic algorithm can be applied for the performance assessment of any similar gas turbine power plant.