煤气化是煤化工的核心，而煤质直接影响气化效率、合成气质量和气化炉稳定运行。为了从长远战略考虑，为宁东煤化工用煤提供稳定可靠的气化煤源，有必要对宁东矿区煤质及其变化规律进行预测研究。本文利用克里金插值法，对宁东矿区梅花井矿煤层煤质及变化规律进行了预测研究；利用多元线性回归及径向基函数神经网络法，对开采工作面煤质变化规律进行预测；针对宁东矿区四种气化炉对煤质的要求，建立了煤化工用煤的煤质评价系统。主要研究成果如下： 在分析宁东矿区梅花井矿煤质资料的基础上，采用GS+9.0软件中的统计模块，得出了4#、6#、10#三个主采煤层煤质参数（水分、灰分、挥发分、全硫、发热量、流动温度）的变异函数，根据变异函数模型，通过克里金插值法，对梅花井矿三个主采煤层的煤质及其分布规律进行了预测分析。结果表明，三个主采煤层自上而下，水分（Mad）呈略微降低趋势，大部分区域煤的Mad含量为6.00～10.00%，主要为中低水分煤；灰分（Aad）变化较小，各可采煤层Aad为6.00～22.00%，大部分区域为特低灰和低灰煤，局部为中灰煤；挥发分（Vdaf）变化较小，各可采煤层Vdaf为31.57～35.50%，属于中高挥发分煤；全硫（St,d）平均含量有下降的趋势，各可采煤层St,d为0.28～1.57%，水平方向上，本区大部分区域为低硫和特低硫煤，主要为中低硫煤；原煤发热量（Qnet,ar）变化不大，各可采煤层Qnet,ar为25.56～27.86MJ/kg，水平方向上，本区大部分区域为高热值煤，局部为中热值煤；灰熔融流动温度（FT）变化不大，各可采煤层FT为1183～1385 ℃，三个煤层都存在FT自西向东逐渐增加，西部主要为低流动温度，东部主要为中等流动温度。 工作面煤质预测以梅花井矿2010~2014年6#煤层90组月度工作面毛煤灰分数据为基础数据，采用多元线性回归和径向基函数神经网络预测方法对其灰分变化进行了预测研究。结果表明，煤层工作面的灰分变化可用多元线性回归预测模型表示：Y=－0.5239＋0.552X1－0.199X2＋0.960X3（其中Y：预采灰分，X1：预采煤层煤样灰分，X2：上月实际灰分，X3上月煤层煤样灰分）。但是多元线性回归预测结果对于极值点比较敏感，容易陷入局部最小，RBF神经网络预测月度煤质得到结果的误差更小，与实际情况更加吻合。灰分预测模型更加趋向于非线性结果。 根据宁东煤化工基地四种气化炉对煤质的要求，结合上述煤质预测结果，建立了宁煤气化用煤的煤质评价系统，通过该系统可预测未来宁东煤化工基地煤气化用煤的煤源。

Coal gasification is the core of the coal chemical industry, and the coal quality directly affects the properties of coal gasification efficiency and the quality of syngas, stable operation of the gasification stove. For the long term strategic thinking, in order to providing stable and reliable gasification raw coal for Ningdong coal chemical industry, it is necessary to carry out prediction research for Ningdong coal mining area and its change rule. This paper used Kriging interpolation method to prediction the quality and distribution of Ningdong coal mining. By using multiple linear regression and RBF neural network, it forecast the change rule of the mining working face coal quality; to the need of Ningdong four kinds of gasifier for coal quality, it is established the coal quality evaluation system in coal chemical industry. The research results are as follows: Based on the analysis of the coal quality data of MeiHuajing mine of Ningdong area, using geostatistics module in GS+9.0 software, obtained the variation function of 4#,6#,10# coal quality seam(Mad, Aad, Vdaf, St,d, Qnet,ar, FT), according to the model of variation function, by Kriging interpolation method, predicted the coal quality and its distribution law of three main coal seams in MeiHuajing mine. The results show that, from top to bottom of the three main mining coal seams, Mad slightly decreases trend, most of the coal Mad content is 6~10%, mainly for medium and low water coal; Aad small changes, the Aad of the mining coal seam is 6.00～22.00%, most of the region is the ultra low ash, and low ash coal, for the partial is medium coal ash; Vdaf changes little, the Vdaf of the mining coal seam is 31.57~35.50%, which belongs to medium and high volatile coal; St,d has a trend of decrease, the St,d of the mining coal seam is 0.28~1.57%, on the horizontal direction, most of this area is low and ultra low sulfur coal, mainly is low sulfur coal; Qnet,ar with little change, the Qnet,ar of the mining coal seam is 25.56~27.86 MJ/kg, on the horizontal direction, most of the area is high calorific value coal, partial is medium calorific value; FT changes little, the FT of the mining coal seam is 1183~1385 ℃, FT of the three coal seam are gradually increased from west to east, the western region is dominated by low FT and the east is middle FT. The coal quality prediction of the working face is based on the data of 90 sets of monthly working face of the 6 coal seam from 2010 to 2014 of Mei Huajing mine, the change of ash content was predicted by the method of multiple linear regression and radial basis function neural network prediction. The results show that：the coal ash content can be used multiple linear regression prediction model representation: Y=－0.5239＋0.552X1－0.199X2＋0.960X3(Y:pre ash, X1: pre ash of coal seam, X2:last month ash, X3：last month ash of coal seam). The result of multiple linear regression prediction is sensitive to the extreme points, easy to fall into local minimum. RBF neural network forecasting the monthly quality of the results of the error is smaller, more consistent with the actual situation. Ash prediction model tend to be more nonlinear results. According to the requirement of the Ningdong coal chemical industry base four gasifier of coal quality, combined with the coal quality prediction results, established a coal quality evaluation system for coal gasification in Ningdong coal gasification, the system can predict the future coal source of coal gasification for Ningdong coal chemical base.