基于量子免疫遗传算法的煤与瓦斯突出神经网络预测

为了用BP神经网络更准确地预测煤与瓦斯突出危险性,将免疫算法中基于繁殖概率的抗体多样性保持机制引入量子遗传算法(QGA),提出量子免疫遗传算法(QIGA)优化神经网络模型QIGA-BP。模型采用QIGA分别对神经网络的隐含层和连接权值进行全局寻优,以此提高BP网络的搜索效率和泛化能力。以平均影响值分析法筛选的煤与瓦斯突出显著变量作为BP网络的最佳输入参数,分别用QIGA-BP,QGA-BP,免疫遗传优化BP模型和传统BP模型对突出煤层工作面的实例数据进行预测。结果表明,QIGA在BP网络优化过程中具有更好的优化性能,用QIGA-BP模型预测工作面突出危险性具有更好的预测能力和更高的预测准确率。

Optimized Neural Network Method for Predicting Coal and Gas Outburst Base on Quantum Immune Genetic Algorithm

For an accurate prediction of working face coal and gas outburst fatalness through neural network, antibody selection mechanism based on reproduction probability was introduced into the quantum genetic algorithm, and a modeling method based on the BP network optimized by quantum immune genetic algorithm （QIGA-BP） was put forward. The QIGA was used respectively to optimize the neural network hidden layers and weights to improve the searching efficiency and generalization ability of BP network. Advantage factors of coal and gas outburst selected based on mean impact value method were used as the input parameters of forecast model. Contrast experiments between QIGA BP, QGA BP, BP model optimized by immune genetic algorithm and traditional BP model were con- ducted on typical engineering practical data. The results show that QIGA can yield a superior optimization perfor mance to other algorithms, and QIGA BP model has better forecast ability and more accuracy for predicting the out burst fatalness.