基于组合权和改进灰色关联理论，针对潘谢矿区4个含水层中提取的35个学习样本，建立了矿井突水水源识别模型，并利用该模型对7个检验样本进行了水源识别。结果表明：相同含水层的学习样本和检验样本中Na++K+,Ca2+,Mg2+,Cl-,SO2-4和HCO-3等6项化学指标值的含量变化趋势更为接近，符合灰色关联理论。组合权重综合考虑了主客观权重，避免人为因素的干扰，同时考虑了识别指标的实际情况。组合权方法计算的6项识别因子中，Ca2+,Mg2+,HCO-3的权重分别为0.231,0.383,0.203，且3者的权重值相加占总值的81.7%，说明3项指标在矿井突水水源识别中起主要作用。采用建立的组合权-改进灰色关联度模型对7个检验水样进行识别，除1个水样外，其余均与实际结果一致，识别准确率达到86%，表明该模型在矿井水源识别中具有一定的适用性。

Based on combination weights and improved grey relational theory, a model for identifying water sources of mine water inrush was established for 35 learning samples extracted from 4 aquifers in Panxie Mining Area, and the model was used to identify water sources for 7 test samples. The results show the content changes of six chemical index values, such as Na++K+, Ca2+, Mg2+, Cl-, SO2-4 and HCO-3 in the learning samples and test samples of the same aquifer are more similar, which conforms to the grey relational theory. The combined weight comprehensively considers the subjective and objective weights, avoids the interference of human factors, and considers the actual situation of the identification indicators. Among the six identification factors calculated by the combined weight method, the weights of Ca2+, Mg2+ and HCO-3 are 0.231, 0.383 and 0.203, respectively, and the combined weight values of the three factors account for 81.7% of the total value, indicating that these three indicators have a great impact on the identification result of mine inrush water source. The established combination weight-improved grey relational degree model was used to identify the test water samples of 7 different aquifers. Except for one water sample, the others were consistent with the actual results, and the recognition accuracy rate reached 86%, indicating that the model has certain accuracy and applicability in mine water source identification.