湖南安化抽水蓄能电站围岩氡析出主控因素及其预测OA

The issue of radon exhalation from rock masses has become increasingly promi-nent during the construction of underground powerhouse facilities in pumped storage power stations,posing a long-term health threat to construction and operational personnel.Meas-ured data from the exploratory adit of the Anhua Pumped Storage Power Station in Hunan Province indicate that deep radon concentrations significantly exceed allowable limits,ne-cessitating effective prevention and control measures.This study developed a prediction model for radon exhalation rates through experimental investigations on samples collected from the adit combined with machine learning methods.The results show that:(1)Lithol-ogy,temperature,humidity,atmospheric pressure,and water content significantly influ-ence the radon exhalation rate.Lithology determines the order of magnitude of the radon exhalation rate.Increasing temperature and humidity promote radon exhalation,while a high-pressure environment has an inhibitory effect.Under saturated water content condi-tions,the radon exhalation rate increases significantly,confirming the"radon carrier"effect of water molecules.(2)Grey relational analysis indicates that lithology and water content are the dominant factors controlling the radon exhalation rate.(3)Prediction mod-els were constructed based on Support Vector Machine(SVM)and Back Propagation(BP)neural networks,and were optimized using the Particle Swarm Optimization(PSO)algorithm.Among these,the SVM model exhibited the best performance,achieving a co-efficient of determination(R2)of 0.964 46 and a root mean square error(ERMSE)of 1.533 1×10-3 on the training set.After optimization,the ERMSE of the PSO-BP model was reduced by 9.8%compared to the BP model.The research findings can provide theoret-ical support and a technical basis for radon prevention and control in pumped storage power stations.