METHOD FOR PREDICTING THERMAL AGGLOMERATION POTENTIAL IN CIRCULATING FLUIDIZED BED (CFB) BOILER-BASED GENERATING UNITS

Abstract

This invention concerns a method for predicting agglomeration potential, with a minimum accuracy performance of 0.95. This invention provides a method for predicting the potential for thermal agglomeration in a circulating fluidized bed (CFB) boiler-based generating unit consisting of the following steps: (a) collecting real-time operational data from the generating system through a data acquisition unit; (b) performing computer pre-processing of the data with cleaning, normalization, and feature selection steps on historical operational data; (c) training a computer-based Gradient Boosting Classifier algorithm prediction model using the processed historical data to identify non-linear patterns correlated with thermal agglomeration events; (d) implementing the trained prediction model into a real-time operating system that receives actual operational data from the generating unit;(e) calculate the probability of potential agglomeration, where the calculation is carried out continuously by a computer based on the trained prediction model; (f) produce prediction results and early warning signals by the computer at least 30 minutes before the actual event occurs; (g) display prediction results and early warning signals along with recommendations for operational actions on a dashboard; and (h) store prediction results and corrective actions in a database system for the purposes of performance analysis and model improvement in the future.