煤炭的发热量是评价煤炭品质的重要指标。 首先对比分析了平滑处理、 微分处理、 多元散射校正(MSC)以及标准归一化(SNV)等光谱与处理方法在改善煤粉近红外漫反射光谱信噪比的效果, 然后利用偏最小二乘法(PLS)和主成分分析方法(PCR)分别对采用每种预处理方法处理后的光谱建立煤粉发热量模型, 发现采用5点平滑处理、 多元散射校正和标准归一化处理可使模型的性能有较明显的改观, 5点平滑效果最好, 相关系数、 校正标准差和预测标准差分别为: 0.989 9, 0.0004 9和0.0005 2, 采用25点平滑处理产生了过平滑现象, 导致模型的性能变坏, 采用微分预处理后的光谱建立的模型没有明显变化, 对模型的性能影响不大。

The calorific value of coal ash is an important indicator to evaluate the coal quality. In the experiment, the effect of spectrum and processing methods such as smoothing, differential processing, multiplicative scatter correction (MSC) and standard normal variate (SNV) in improving the near-infrared diffuse reflection spectrum signal-noise ratio was analyzed first, then partial least squares (PLS) and principal component analysis (PCR) were used to establish the calorific value model of coal ash for the spectrums processed with each preprocessing method respectively. It was found that the model performance can be obviously improved with 5-point smoothing processing, MSC and SNV, in which 5-point smoothing processing has the best effect, the coefficient of association, correction standard deviation and forecast standard deviation are respectively 0.989 9, 0.000 49 and 0.000 52, and when 25-point smoothing processing is adopted, over-smoothing occurs, which worsens the model performance, while the model established with the spectrum after differential preprocessing has no obvious change and the influence on the model is not large.