基于神经网络的测斜仪井斜角误差补偿研究

Research on Error Compensation of Angle of Deviation Based on Neural Network

介绍了陕北石油基地普遍使用的磁性电子测斜仪的测量原理,分析了其可能存在的误差和来源。对井眼姿态测量中主要测量参数之一的井斜角,分别采用反向传播(BP)神经网络和径向基函数(RBF)神经网络算法,建立了以测量井斜角为输入、补偿后的井斜角为输出的单入单出神经网络模型,并用实际测斜仪器的测量数据进行现场测试。实验数据表明,采用BP和RBF神经网络补偿算法,可将井斜角的实际测量精度分别从±0.77°提高至±0.16°和±0.19°,可见两种算法均可提高测斜仪井斜角的测量精度,BP补偿算法精度较高,但RBF补偿算法的收敛速度更快。

The structure and measurement principle of common magnetic dip meter in North Shaanxi oil base are introduced,whose possible error and its source are analyzed. For the angle of deviation,one of the main parameters of borehole attitude measurement,a neural network is established based on Error Back Propagation（ BP） and Radial Basis Function（ RBF） compensation algorithm respectively,which input is measured deviation angle and output is compensated deviation angle,and the sampling data of the magnetic dip meter are used to test. The experiment results show that using the BP and RBF compensate algorithm,the precision of deviation angle can be improved from± 0.77° to ± 0.16°and ± 0.19° or better respectively. The precision of BP compensate algorithm is higher and the convergence rate of RBF compensate algorithm is faster.