US 11,860,619 B2
Fault early-warning method and system applied to gas turbine unit, and apparatus
Kun Zhang, Hangzhou (CN); Hongren Li, Hangzhou (CN); Pingyang Zi, Hangzhou (CN); Wei Li, Hangzhou (CN); and Liang Sun, Hangzhou (CN)
Assigned to HUADIAN ELECTRIC POWER RESEARCH INSTITUTE CO., LTD., Hangzhou (CN)
Filed by HUADIAN ELECTRIC POWER RESEARCH INSTITUTE CO., LTD., Hangzhou (CN)
Filed on May 8, 2023, as Appl. No. 18/313,525.
Application 18/313,525 is a continuation of application No. PCT/CN2022/121824, filed on Sep. 27, 2022.
Claims priority of application No. 202210633645.3 (CN), filed on Jun. 6, 2022.
Prior Publication US 2023/0393571 A1, Dec. 7, 2023
Int. Cl. G06F 30/20 (2020.01); G05B 23/02 (2006.01); G06F 119/02 (2020.01)
CPC G05B 23/0283 (2013.01) [G06F 30/20 (2020.01); G06F 2119/02 (2020.01)] 6 Claims
OG exemplary drawing
 
1. An effective fault early-warning method applied to a gas turbine unit on a basis of Kriging primary functions, comprising:
calculating, by means of a mechanism model, predicted data of prediction parameters in a gas turbine unit, and performing data comparison on the predicted data and real measurement data of the prediction parameters, so as to obtain an error matrix;
constructing several Kriging primary functions according to the mechanism model;
screening an optimal Kriging primary function from the several Kriging primary functions according to the error matrix, wherein screening the optimal Kriging primary function further include
respectively fitting the error matrix by means of the several Kriging primary functions, and screening the Kriging primary functions according to several fitting results, so as to obtain a first primary function set;
respectively calculating the reliability indexes of the Kriging primary functions in the first primary function set by means of a monte carlo algorithm, a particle swarm optimization algorithm, a response surface algorithm and a unitary quadratic matrix algorithm; and screening, according to the reliability indexes, a second primary function set with a convergence rate being higher than a preset threshold, and then screening the optimal Kriging primary function from the second primary function set;
performing error compensation on the mechanism model by using the optimal Kriging primary function to reduce false alarm rate of the mechanism model for fault early warning; and
performing fault early warning on the gas turbine unit by means of the mechanism model after error compensation.