摘要
鉴于含有加性噪声的指数模型描述了一类重要的非线性随机系统。本文给出这类系统的参数递推辨识算法，克服了迭代算法不能在线运行、需反复矩阵求逆的不足。当系统时变时，还采用了虚拟噪声技术来补偿因参数时变引起的建模误差，从而改善动态预报器的性能。应用这种方法，对油田采油井、注水井的套管损坏情况进行了多步动态预报。
The exponential models with additive noise describe a class of important nonlinear stochastic systems. This paper presents a recursive parameter identification algorithm for the system. Compared with the iterative algorithm, it can avoid the matrix inversion and can be operated on-line. When the system is time-varying, a fictitious noise technique is used to compensate the modeling error caused by the time-varying parameters, which can improve the performence of the dynamic predictor. Practical utility of the approach is illustrated by modeling and predicting the casing collapse rate of oil and intake wells for an oil field.The calculation result is satisfactory.
基金
中国石油天然气总公司"九五"攻关课题资助
关键词
非线性系统
参数辨识
套管损坏
动态预报
采油井
nonlinear system
exponential model
additive noise
parameter identification
casing collapse
dynamic prediction
