针对目前煤矿防突预警技术存在的预警模型有效性不明确、预警临界值误差较大、预警系统对煤矿的适应性较差等问题，分析了煤矿防突信息的特征，应用多源数据驱动思维，设计了多源数据驱动的防突预警系统结构，研究了多源数据驱动的防突预警指标自适应技术，实现预警模型有效性的定量评价智能优选和预警临界值自适应训练，提升预警系统对煤矿的适应能力，提高预警准确率。通过建立预警模型有效性定量评价方法，智能优选预警模型，避免人为主观因素造成的模型选择失误；通过建立预警临界值自适应训练方法，使预警临界值能够与煤矿实际情况相适应，并且预警临界值的精确度随时间的推移逐渐提高，减少了临界值考察的工作量。该技术体系在贵州盘州地区某煤矿应用过程中，从4个代表性预警模型中定量筛选了效果较好的预警模型A和预警模型S，在瓦斯体积分数并未超限的情况下，预警模型A提前捕捉到因煤厚变化引起的瓦斯涌出异常状况，并及时发布预警信息，在采取相应措施后消除了事故隐患。该技术能够使煤矿防突预警具有更高的适应性，提高煤矿防突管理水平，为安全生产决策提供支持。

In view of the problems existing in the current coal mine outburst prevention and early warning technology, such as unclear effectiveness of early warning model, large error of early warning critical value and poor adaptability of early warning system to coal mine, this paper analyzed the characteristics of coal mine outburst prevention information, designed the structure of outburst prevention early warning system driven by multi-source data, studied the adaptive technology of outburst prevention early warning index driven by multi-source data, realized intelligent optimization of quantitative evaluation of early warning model effectiveness and adaptive training of early warning critical value, and improved the adaptability of early warning system to coal mine. By establishing a quantitative evaluation method for the effectiveness of the early warning model, the early warning model is intelligently selected to avoid model selection errors caused by human subjective factors; By establishing the adaptive training method of early warning critical value, the early warning critical value can adapt to the actual situation of coal mine, and the accuracy of early warning critical value is gradually improved with the passage of time, thus reducing the workload of critical value investigation. In the application process of this technology system in a coal mine in Panzhou area, Guizhou Province, the early warning model a and early warning model s with better effects were quantitatively screened from 4 representative early warning models. On August 10th, under the condition that the gas concentration did not exceed the limit, early warning model a caught the abnormal situation of gas emission caused by the change of coal thickness in advance, and issued early warning information in time. After taking corresponding measures, the hidden dangers of accidents were eliminated. This technology can make coal mine outburst prevention early warning have higher adaptability, improve coal mine outburst prevention management level, and provide support for safety production decision.