基于量子粒子群优化支持向量回归算法计算煤层气体含量的预测

Prediction of Gas Content in Coal Stratum Based on Support Vector-Based Regression Algorithm Trained by Quantum Particle Swarm Optimization

煤层气是近一二十年在国际上崛起的洁净、优质能源和化石能源的新矿种,其地质研究对能源的补充意义重大.该文对煤层气的含量提出基于量子粒子群优化的支持向量回归算法进行预测.支持向量回归算法是一种非线性的基于内核的回归方法,它可以采用良好的函数逼近,并具有泛化能力.由于支持向量回归算法的参数对预测性能影响很大,量子粒子群优化在本研究中可用于选择支持向量回归算法参数.本文选用基岩深度,煤层的厚度,断层间的水平距离,煤的挥发分作为煤层气含量的预测模型的输入向量,经过比较量子粒子群优化的支持向量回归算法和支持向量回归算法之间的煤层气体含量的预测误差表明,量子粒子群优化得到的煤层气体含量的预测精度均高于支持向量回归算法的精度.

Coalbed methane is a kind of clean, high quality fossil energy that is widely used in recent years. The great significance of energy supplement draws the interests of its geological study. In this paper, based on support vector regression trained by quantum particle swarm optimization, the prediction of gas content in coal stratum is proposed. Support vector regression is a nonlinear kernel-based regression method, which can employ good function approximation and has generalization capabilities. As the parameters of support vector regression are important influence on its forecasting performance, quantum particle swarm optimization can be used to select the parameters of support vector regression in this study. In the paper, the bedrock depth, the thickness of coal seam, horizontal distance from faultage, and volatile coal are used as the input vector of prediction model of gas content in coal stratum. The comparison of the prediction error of gas content in coal stratum between support vector regression trained by quantum particle swarm optimization and support vector regression indicates that the prediction accuracies for gas content in coal stratum of support vector regression trained by quantum particle swarm optimization are higher than those of support vector regression.