School of Communication and Information Engineering, Xi’an University of Science and Technology
School of Safety Science and Engineering, Xi’an University of Science and Technology

为对煤矿瓦斯爆炸事故风险等级进行精准预测，以符合实际情景的特征向量为前提，基于改进粒子群算法优化概率神经网络(RWPSO-PNN)实现瓦斯爆炸风险等级预测模型。首先利用中文分词提取煤矿瓦斯爆炸事故致因，以灰色关联分析(GRA)选取模型的输入特征向量；并针对概率神经网络(PNN)中平滑因子易引起网络识别率低的问题，提出了RWPSO-PNN，实现平滑因子的自适应调整；最后对RWPSO-PNN进行了实例分析，并与极限学习机算法、BP神经网络和支持向量机算法进行对比。结果表明：RWPSO-PNN预测准确率为90%，平均绝对误差为0.133，明显优于对比算法。

In order to accurately predict the risk level of coal mine gas explosion accidents, based on the feature vector in line with the actual situation, the particle swarm optimization probabilistic neural network (RWPSO-PNN) is improved to realize the prediction model of gas explosion risk level. Firstly, the cause of coal mine gas explosion accident is extracted by Chinese word segmentation, and the input feature vector of the model is selected by grey correlation analysis (GRA). Aiming at the problem of low recognition rate caused by smoothing factor in probabilistic neural network (PNN), RWPSO-PNN is proposed to adjust the smoothing factor adaptively. Finally, RWPSO-PNN is analyzed and compared with extreme learning machine algorithm, BP neural network and support vector machine algorithm. The results show that the prediction accuracy of RWPSO-PNN is 90 %, and the average absolute error is 0.133, which is obviously better than the comparison algorithm.

国家重点研发计划资助项目（2018YFC0808301）

郝秦霞，尚海涛. 煤矿瓦斯爆炸事故致因选取与风险等级预测[J]. 煤矿安全，2023，54（10）：243−249.

HAO Qinxia, SHANG Haitao. Causes selection and risk level prediction of coal mine gas explosion accident[J]. Safety in Coal Mines, 2023, 54（10）: 243−249.