小波分解重建算法提高近红外煤质水分检测精度

水分是煤质经济价值的基本指标，会造成煤质利用率下降和能源浪费。因此，近红外光谱法实现煤质水分快速、无损、低成本检测具有重要意义。针对目前近红外煤质水分检测精度普遍在1%，且鲜有关于精度提高系统性研究的现状，从光谱预处理及波长选择进行比较，提出了提高煤质水分检测精度的可行性方法。实验结果表明，采用小波分解重建算法对1 300～2 400 nm光谱区间进行降噪及去基线预处理，对重建光谱进行偏最小二乘建模，其建模结果明显优于其他方法，RMSEC达到0.06%，RMSEP达到0.27%，相关性系数R-square达到0.995，且证明模型稳定性较好，显著提高了近红外煤质水分的检测精度。

Improving Precision in Coal Moisture Detection Using Wavelet Transform

Moisture, as a core determination of the economic value of coal, can result in the utilization and energy inefficiency. Near-infrared (NIR) spectroscopy, with advantages of high accuracy and low cost, provides significant solution to the quick and non-invasive detection of coal moisture. In the present paper, the improvement of the coal moisture analysis was conducted based on the precision of 1% and insufficient comparisons in recent experiments, and aspects of spectrum pretreatment and wavelength selection were mainly discussed. The optimized result with R-square of 0.995, RMSEC of 0.06% and RMSEP of 0.27% indicates the priority of wavelet decomposition and reconstruction, compared with other methods, in the noise reduction and baseline removing of original spectra (1 300～2 400 nm) before PLS modeling, and the stability experiment validates its robust potential in improving precision of coal moisture detection based on the NIR spectroscopy.