摘要
基于机器学习的时间序列预测方法能够挖掘时序数据本身的规律,可提高大坝变形预测的精度。以江坪河水电站面板堆石坝为研究对象,对监测资料进行分析,采用Prophet机器学习模型建立堆石坝变形监测数据的时间序列预测模型,并采用贝叶斯优化Prophet模型的超参数。预测结果表明,利用机器学习模型预测堆石坝变形的精度是可以接受的,且实施过程方便快捷,无需太多的人工干预,对建立面板堆石坝的长期变形的实时动态预测模型与高堆石坝全生命周期的性状评估与隐患及时预警具有一定的实用价值。
The time series prediction method based on machine learning can mine the rule of time series data itself and improve the accuracy of dam deformation prediction. Taking Jiangpinghe face rockfill dam as the research object, the monitoring data is analyzed. The Prophet model is used to establish the time series prediction model of the rockfill dam deformation monitoring data, and the Bayesian optimization is used to adjust the hyperparameters of Prophet model. The prediction results of the Prophet model shows that the accuracy of predicting the deformation of rockfill dam using the machine learning model is acceptable, and the implementation process is convenient and fast without much manual intervention. This has certain practical value for establishing the real-time dynamic prediction model of long-term deformation of face rockfill dam and the performance evaluation of high rockfill dam in whole life cycle and the timely warning of hidden dangers.
作者
冷天培
马刚
殷彦高
谭瀛
周伟
LENG Tianpei;MA Gang;YIN Yangao;TAN Ying;ZHOU Wei(State Key Laboratory of Water Resources and Hydropower Engineering,Wuhan University,Wuhan 430072,Hubei,China;PowerChina Zhongnan Engineering Corporation Limited,Changsha 410014,Hunan,China)
出处
《水力发电》
北大核心
2020年第6期29-34,共6页
Water Power
关键词
面板堆石坝
变形预测
Prophet模型
贝叶斯优化
江坪河水电站
face rockfill dam
deformation prediction
Prophet model
Bayesian optimization
Jiangpinghe Hydropower Station
