【摘要】 针对煤炭发热量传统检测方法周期长、效率低的问题，提出了一种基于近红外光谱技术的全信息竞争自适应高斯回归建模策略。该方法首先采用自扩展全信息优化策略，自动寻找最佳光谱预处理方法组合，然后利用竞争性自适应重加权采样算法筛选与发热量最相关的特征波长，最后基于高斯过程回归建立定量分析模型。实验结果表明，其预测决定系数可达0.920，预测均方根误差为0.909，性能显著优于主成分分析、连续投影算法等传统特征提取方法及支持向量机模型。本研究提出的面向煤炭品质快速无损检测的近红外建模方法，在所研究样本与实验条件下表现出较好的预测精度与稳健性，建模流程完整，可用于火电、冶金等工业场景的煤质快速监测与质量控制。更多还原

【Abstract】 To address the long analytical cycle and low efficiency of conventional calorific value measurements, this study develops a near-infrared spectroscopy modeling framework called Full-Information Competitive Adaptive Gaussian Modeling, or FICAGM. The method first employs a self-extending fullinformation optimization strategy to automatically determine the most suitable preprocessing sequence for the spectra. It then applies competitive adaptive reweighted sampling to extract wavelength variables most relevant to calorific value, followed by a Gaussian process regression model for quantitative calibration. Experiments show that the proposed framework achieves a prediction coefficient of determination of 0.920 and a root-mean-square error of prediction of 0.909, outperforming traditional chemometric approaches such as principal component analysis, successive projection algorithms and support vector machines. Under the tested conditions, the framework exhibits strong predictive accuracy and robustness, and its complete modeling pipeline is well suited for rapid, non-destructive coal quality monitoring in coal-fired power generation, metallurgical processes and other industrial applications.更多还原