基于高斯过程回归和BP神经网络的油储地罐容积表标定研究  

Research on the volume gauge calibration of oil storage tank based on Gaussian process regression and BP neural network

作　　者：王彩玲[1] 程叶 许欣黎 倪庆旭 WANG Cailing;CHENG Ye;XU Xinli;NI Qingxu(Xi'an Shiyou University)

机构地区：[1]西安石油大学,陕西省西安市710065

出　　处：《石油石化节能与计量》2025年第2期26-30,35,共6页Energy Conservation and Measurement in Petroleum & Petrochemical Industry

摘　　要：石油作为中国重要的能源资源之一,广泛应用于发电、运输、工业生产等各个领域。准确的油储地罐容积表标定对于确保各类石油产品储存、运输和交易的精确计量至关重要。传统的标定方法通常高度依赖于静态测量和经验公式,易受时间、环境条件及人为因素的影响。为了解决这一问题,提出了一种基于高斯过程回归(GPR)和反向传播神经网络(BPNN)的标定验证方法。在真实加油站数据构建的数据集上进行实验,结果显示,高斯过程回归模型和BP神经网络模型的平均均方根误差RMSE分别为3.435、8.409,模型的预测效果相对较好,研究结果可为容积表的标定工作提供有价值的参考。As one of the important energy resources in China,petroleum is widely used in power generation,transportation,industrial production and other fields.The accurate tank volume gauge calibration is essential to ensure accurate measurement of the storage,transportation and trading of all types of petroleum products.The traditional calibration methods often rely highly on static measurements and empirical formulas,and it are susceptible to time,environmental conditions,and human factors.In order to solve this problem,the calibration verification method based on Gaussian process regression(GPR)and Back Propagation neural network(BPNN)are proposed.The experiments on the data set constructed from real gas station data show that the average RMSE of the Gaussian process regression model and the BP neural network model are 3.435 and 8.409 respectively,and the prediction effect of the model is relatively good,which makes research results can provide a valuable reference for the calibration of the volume.

关 键 词：容积表标定 BP神经网络 高斯过程回归 数据挖掘 误差预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TE972[自动化与计算机技术—控制科学与工程]