Abstract:Nowadays, with the goals of "Carbon Peaking and Carbon Neutrality," the economic and reliable operation of gas turbine generator sets has gradually become a focal point in the industry. Therefore, it is of great value to quickly and accurately detect faults and improve operational levels. In this paper, we address the issue of fault detection in gas turbine power plants. By combining wavelet denoising with a fault detection method based on Nonlinear State Estimation Technology (NSET), we propose an approach that involves constructing a memory matrix, establishing a mathematical model, comparing actual and predicted curves, and analyzing residual values to achieve effective fault detection, with the big data in electric power. Taking a gas turbine power plant in Zhejiang Province as an example, we verify the method using a simulation platform, which successfully detect all 33 types of faults on the simulation platform. Moreover, the method has been effectively applied using real industrial data, demonstrating its ability to detect faults such as those in the gas turbine speed ratio valve and thermocouples. Finally, the method has been implemented in the E-powerplant project, serving as an engineering demonstration.