A Semi-supervised Approach for Steam Turbine Health Prognostics Based on GAN and PF

Steam turbines are indispensable part of power plants, they will cause serious impact once fail. Health prognostics is an effective technology to keep steam turbine in healthy condition and avoid unnecessary losses. This paper proposed a semi-supervised framework for health prognostics on steam turbines. The framework is based on Generative Adversarial Networks (GAN) and Particle Filter (PF). In this framework, anomaly detection and remaining useful life (RUL) prediction are conducted. During anomaly detection, a health index (HI) is constructed to indicate steam turbine operation conditions and an anomaly threshold is used to detect anomalies. During predicting the RUL, an approach based on PF is employed to predict the RUL. Real case are demonstrated to validate the effectiveness of the proposed framework. Results indicate that the proposed framework has good performance on anomaly detection and RUL prediction.