Dictionary learning-based graded warning of water wall expansion of utility boilers

Abstract

As coal-fired power transitions from a primary energy source to a fundamental and system-regulating source, utility boilers are increasingly required to operate under more flexible conditions. This shift inevitably introduces risks of expansion failure in water walls. This paper proposes a dictionary learning (DL)-based expansion warning model to address this issue. To mitigate the impact of the randomly selected initial dictionary, kernel principal component analysis (KPCA) is integrated into DL to create a deterministic approach for optimizing the initial dictionary. Additionally, a three-grade warning framework is established using multiple reconstruction error thresholds to improve the accuracy of expansion warning. A 300 MW power unit is used as a study case for warning model validation. The model is built with 25 operational variables, including three expansions in the X, Y, and Z directions and other operational parameters of the utility boiler, respectively. The results demonstrate that the warning rates of the three-grade warning model align well with actual offset values, confirming the effectiveness of the DL-based graded expansion warning model.