为了提高导水断裂带发育高度预测的准确性，通过收集以往地质采矿条件相似地区导高实测数据，综合分析取采高、埋深、倾角及工作面斜长4个因素作为算法模型的导高主要影响因子，利用工程探测、机器学习和数值模拟相结合的方法，对采矿覆岩导水断裂的发育特征进行分析研究。通过钻孔试验以及数控摄像技术对浅埋煤层综采条件下的导水断裂带发育高度进行实测；通过构建基于自适应粒子群算法优化最小二乘支持向量机回归算法（APSO-LSSVR）和UDEC数值模拟的导高预测模型，结合实测导高数据确定华宁煤矿覆岩破坏高度计算方法。结果表明：研究区实测导水断裂带发育高度处于60.3～90.6 m；基于APSO-LSSVR算法的导高模型的预测结果与真实值的拟合优度为94.79%，均方根误差为1.652 3，预测的准确性较高，将其应用到研究区不同工作面导水断裂带高度的预测中，该模型与实测数据相比平均相对误差为1.36%，相较于传统UDEC数值模拟预测方法，其准确性相对提高了9.03%。可以看出经自适应粒子群算法优化后的支持向量机模型对于数据特征较小的数据集合处理性能较高，能够较好反映导水断裂的发育情况，可满足实际开采需求；将基于APSO-LSSVR的导水断裂带发育高度预测模型应用于研究区待开采的22109、22110工作面，得出22109工作面导水裂隙带为62.7 m,22110工作面导水裂隙带为67.3 m。

In order to improve the accuracy of the prediction of the development height of the water-conducting fissure zone, by collectingthe measured data of the lead height in the areas with similar geological and mining conditions in the past, the four factors of miningheight, burial depth, inclination angle and working face slope length are comprehensively analyzed as the lead height of the algorithmmodel. The main influencing factors are to analyze and study the development characteristics of water-conducting fissures in the miningoverburden by using a combination of engineering detection, machine learning and numerical simulation. Through drilling experimentsand numerical control camera technology, the development height of the water-conducting fracture zone under the conditions of fullymechanizedmining in shallow coal seams was measured; by constructing an adaptive particle swarm optimization algorithm based on optimizedleast squares support vector machine regression algorithm (APSO-LSSVR) and UDEC The numerical simulation of the lead heightprediction model, combined with the actual measured lead height data, determines the calculation method of the damage height of theoverlying strata in Huaning Coal Mine. The results show that the measured development height of the water-conducting fracture zone inthe study area is between 60.3~90.6 m; the goodness of fit between the predicted result of the lead height model based on the APSO-LSS-VR algorithm and the true value is 94.79%, and the root mean square error is 1.652 3 , The prediction accuracy is high, and it is applied tothe prediction of the height of the water-conducting fissure zone of different working faces in the study area. Compared with the measureddata, the average relative error of this model is 1.36%. Compared with the traditional UDEC numerical simulation prediction method, Itsaccuracy is relatively improved by 9.03%. It can be seen that the support vector machine model optimized by the adaptive particle swarmalgorithm has higher processing performance for data collections with smaller data characteristics, can better reflect the development ofwater-conducting fissures, and can meet actual mining needs; it will be based on APSO -LSSVR's water-conducting fracture zone developmentheight prediction model is applied to the 22109 and 22110 working faces to be mined in the study area, and it is concluded that theguided height of the 22109 working face is 62.7 m, and the guided height of the 22110 working face is 67.3 m.

0 引言
1 研究区概况
2 导水裂隙带高度影响因素分析
3 导水裂隙带发育高度预测模型
3.1 APSO-LSSVR回归预测原理
3.2 模型的构建
3.3 数据的预处理
3.4 模型参数寻优
3.5 回归模型检测
3.6 模型的检验
4 基于UDEC的导水裂隙带高度预测
4.1 模型建立及参数选取
4.2 数值模拟预测结果分析
5 预测结果对比及应用
5.1 预测结果检验
5.2 预测方法应用
6 结论