روش های رگرسیون برداری برای بازگشت همزمان داده های مصالحه و تشخیص خطا ناخالص در سیستمهای دینامیکی غیر خطی

The quality of process data in a chemical plant significantly affects the performance and benefits gained from activities like performance monitoring, online optimization, and control. Since many chemical processes often show nonlinear dynamics, techniques like extended Kalman filter (EKF) and nonlinear dynamic data reconciliation (NDDR) have been developed to improve the data quality. Recently, the recursive nonlinear dynamic data reconciliation (RNDDR) technique has been proposed, which combines the merits of EKF and NDDR techniques. However, the RNDDR technique cannot handle measurements with gross errors. In this paper, a support vector (SV) regression approach for recursive simultaneous data reconciliation and gross error detection in nonlinear dynamical systems is proposed. SV regression is a compromise between the empirical risk and the model complexity, and for data reconciliation it is robust to random and gross errors. By minimizing the regularized risk instead of the maximum likelihood in the RNDDR, our approach could achieve not only recursive nonlinear dynamic data reconciliation but also gross error detection simultaneously. The nonlinear dynamic system simulation results in this paper show that the proposed approach is robust, efficient, stable, and accurate for simultaneous data reconciliation and gross error detection in nonlinear dynamic systems within a recursive real-time estimation framework. It can also give better performance of control.

An SV regression approach for recursive simultaneous data reconciliation and gross error detection in nonlinear dynamic system was presented in this paper. The SV regression approach is found to be robust and have superior performance. This approach considers the statistical learning theory as the framework of data reconciliation and gross errors detection instead of the empirical risk, so it can detect gross errors and estimate gross error values. At the same time, it is based on a recursive estimation framework, which makes it preferable for online application. Another advantage of our approach is that no linearized system model is needed, which eliminates the adverse effect on the accuracy of the state and the covariance matrix estimates by an approximate linear model of the process. The nonlinear dynamic system simulation results in this paper show that the SV regression approach is robust, stable, and accurate for simultaneous data reconciliation and gross error detection in nonlinear dynamic systems within a recursive real-time estimation framework. It can also give better performance of control.The SV regression approach proposed in this paper was applied to a widely used model defined in (1) which constitutes the constraints l (x) in the SV regression approach. Furthermore, more complex models could be used instead of the model defined in (1) to constitute the constraints l (x), then the SV regression approach proposed in this paper could be extended to more complex models, which shall be addressed in future works.