Intelligent Alarm Method of Ultra-Supercritical Units Based on Kernel Density Estimation and Convolutional Optimization Algorithm

In order to improve the performance and operational safety of ultra-supercritical (USC) units alarm system, the problem of unreasonable setting of the traditional alarm threshold is addressed. In this paper, an alarm threshold optimization method based on kernel density estimation (KDE) and convolutional optimization algorithm (COA) is proposed. Firstly, the KDE method is used to estimate the data in normal and abnormal states. Secondly, in view of the traditional method that only considers missed alarms and false alarms, this paper introduces the detection delay expectation, and comprehensively considers the above three evaluation indicators to establish the objective function of optimizing the alarm threshold. Then, the COA is used to solve the objective function optimally. Finally, the simulation results show that the method ensures low false alarms while greatly reducing missed alarms and detection delay expectation, which enhances the alarm system performance.