(2020) Development of Alarm Prediction System for Monitoring Steam Turbine Based on SCADA Data
Abstract: A steam turbine is a critical machinery in power plants for generating a driving force of a generator. The high reliability of the turbine is a must to guarantee the availability of power plant in producing electricity. Therefore, a turbine condition monitoring (CM) system is needed to access real conditions and the health state of such equipment. Even though many CM systems have been developed, however, the current system was set up manually based on a determined threshold that was adopted from standard or best practice. The CM system that automatically generates the true alarm based on the performance of the generator is still rare. In this paper, an automated alarm system that uses the power output of a generator as a reference has been developed. The data for developing the alarm system is vibration data acquired by the SCADA system that is a very famous data acquisition system used in the industry. Furthermore, the method of developing such an alarm system is machine learning (ML) through a long short-term memory (LSTM) network. Validation of the proposed method has been conducted using a real system of SCADA data for training the LSTM. The trained LSTM is then used to generate an alarm system based on predicted data for turbine condition monitoring. The results show that the alarm generation and prediction give a plausible performance measured by RMSE.
You can view the details on this paper.