Ash Fouling Prediction of Individual Heating Surface of Boiler for Ultra-Supercritical Coal Power Plant based on Regression Techniques

Abstract

The ash fouling inside the boiler has a detrimental effect on its performance, leading to suboptimal efficiency. Soot blower systems are commonly used during power plant operations to mitigate fouling. Most soot blower operations are scheduled and fixed without considering the actual degree of fouling inside the boiler, often resulting in either insufficient or excessive blowing. The former reduces heat transfer efficiency, while the latter leads to the wastage of high-pressure steam and shortens the lifespan of boiler pipes. This study aims to predict the fouling conditions of six individual heating surfaces within the boiler: primary, secondary, and final superheaters; primary and final reheaters; and the economizer, utilizing indirect and data-driven methods. Direct methods, such as sensor installation, are impractical due to the extreme conditions within the boiler. The study initially establishes the relative cleanliness level as an indicator of fouling degree by comparing current heat absorption values with reference values derived from statistical analysis during periods of stable boiler conditions. Data cleaning methods are then applied before employing regression techniques for ash fouling prediction. Gaussian Process Regression (GPR), a nonparametric kernel-based probabilistic model, and Support Vector Machine (SVM) with different kernels are experimented with for comparison. Experimental analysis demonstrates high accuracy, ranging from 91.4% to 98.2% for GPR and 89.1% to 98.1% for SVM on the case study data. The implementation of the prediction model in this study is expected to enhance soot blowing operations, ultimately optimizing boiler performance. This improvement will lead to higher energy efficiency and a reduction in detrimental emissions.