随着我国煤炭开采强度的不断增大，开采过程中瓦斯涌出量也不断增大，严重影响煤矿安全、高效的生产。瓦斯涌出量影响因素以及瓦斯涌出预测的研究，是瓦斯防治工作的基础，对矿井的安全生产、防止瓦斯事故的发生具有重要作用。本文以沙吉海煤矿有限责任公司煤矿为研究对象，由于神经网络其自组织、自适应、并行化处理等特性，基于神经网络的瓦斯预测方法开始逐渐被使用。随着深度开采，越来越多的技术与设备投入到矿井生产过程中，对矿井瓦斯涌出量的精准预测的需求，也在这过程中不断的增大。但是由于矿井瓦斯涌出量其自身所具有的特性，影响因素多而繁杂，会影响神经网络的规模，所以要通过其他方法降低神经网络其自身的泛化能力。本文根据主成分分析算法、粒子群算法、神经网络算法为基础，对算计进行优化，提升模型的整体性能。

本文提出基于优化的PCA-DO-PSOA-BPNN神经网络预测模型，由于矿井瓦斯涌出量的影响因素自身特性，首先使用主成分分析算法，对繁杂的影响因素进行降维，选取主成分，之后通过改进的双优化PSO算法，对神经网络进行优化，最后将数据输入到BPNN神经网络中进行学习与预测。结合沙吉海煤矿矿井瓦斯涌出特性，设计实验将原始数据分为训练子集和测试子集。将基于优化的PCA-DO-PSOA-BPNN神经网络矿井瓦斯涌出量预测模型与主流的BP神经网络模型等模型进行对比实验，基于MATLAB设计矿井瓦斯涌出量预测软件，证明了其在预测精度和时间优良性方面表现良好。

With the increasing intensity of coal mining in China, the amount of gas emission in the mining process is also increasing, which seriously affects the safe and efficient production of coal mines. The research on the influencing factors of gas emission and the prediction of gas emission is the basis of gas prevention and control work, which plays an important role in the safe production of mines and the prevention of gas accidents. This paper takes the coal mine of Shajihai Coal Industry Co., Ltd.as the research object. Due to the characteristics of self-organization, self-adaptation and parallel processing of neural network, the gas prediction method based on neural network has been gradually used. With the deep mining, more and more technology and equipment are put into the mine production process, and the demand for accurate prediction of mine gas emission is also increasing in the process. However, due to the characteristics of mine gas emission, there are many and complicated influencing factors, which will affect the scale of neural network. Therefore, other methods should be used to reduce the generalization ability of neural network. Based on principal component analysis algorithm, particle swarm optimization algorithm and neural network algorithm, this paper optimizes the calculation and improves the overall performance of the model.

In this paper, a PCA-DO-PSOA-BPNN neural network prediction model based on optimization is proposed. Due to the characteristics of the influencing factors of mine gas emission, the principal component analysis algorithm is used to reduce the dimension of the complicated influencing factors and select the principal components. Then, the neural network is optimized by the improved double optimization PSO algorithm. Finally, the data is input into the BPNN neural network for learning and prediction. Combined with the gas emission characteristics of Shajihai Coal Mine, the original data is divided into training subset and test subset. The prediction model of mine gas emission based on optimized PCA-DO-PSOA-BPNN neural network is compared with the mainstream BP neural network model. The prediction software of mine gas emission is designed based on MATLAB, which proves that it performs well in prediction accuracy and time superiority.