Thermal process data detection method based on equivalent change rate calculation

Abstract

The invention discloses a thermal process data detection method based on equivalent change rate calculation. The method includes the following steps: selecting historical data under normal working conditions through a real-time / historical database; then, building process data abnormal value detection models for corresponding working condition regions respectively; calculating values of characteristic statistics as per the time series sample during on-line detection; confirming the data to be normal working condition data if the values do not exceed the threshold value, and storing the data in the real-time / historical database, and using the updated data for follow-up modeling to update the model; calculating out the equivalent change rate of the values of characteristic statistics to detect out the abnormal value in the values of characteristic statistics if the values exceed the threshold value, and completing the abnormal value detection process through the exclusion manner or other treatment manners. According to the invention, the effect cannot be affected by factors of system change, equipment aging, maintenance and the like as the model updating is adopted, and the calculation of equivalent change rate for values of characteristic statistics is simple and concise, and the entire real-time data detection and processing can be well completed based on the calculation.

Description

technical field [0001] The invention discloses a thermal process data detection method based on equivalent change rate calculation, and relates to the technical field of data detection and fault diagnosis. Background technique [0002] The supply of electric energy plays an extremely critical role in modern society. The safe and economical operation of thermal power units has always been the core content of power companies. Research on thermal power unit operation monitoring and energy-saving optimization has always been the focus of scholars. With the continuous improvement of power station automation and informatization level, the distributed control system used for unit monitoring is equipped with tens of thousands of measuring points, forming a complex measuring network, and a large number of measured values ​​form a form of data flow. It becomes the basis of application systems such as power station monitoring and optimization. [0003] The measurement data is the basi...

AI Technical Summary

Problems solved by technology

However, the important bottleneck of KPCA data modeling is that its original phase does not exist, and the location of fault variables cannot be completed through high-dimensional space

Although some scholars have proposed the method of partially deriving the contribution value of dummy variables to solve the problem of fault variable detection, this process involves a large number of complex calculations, and at the same time lacks a certain processing strategy for the detected outliers. Applications

Images