Particle filter-based robust state and parameter estimation for nonlinear process systems with variable parameters
Zhiliang Zhu, Zhiqiang Meng, Tingting Cao, Zhengjiang Zhang, Yuxing Dai · Measurement Science and Technology · 2017
Abstract State and parameter estimation (SPE) plays an important role in process monitoring, online optimization, and process control. The estimation of states and parameters is generally solved simultaneously in the SPE problem, where the parameters to be estimated are specified as augmented states. When state and/or measurement equations are highly nonlinear and the posterior probability of the state is non-Gaussian, particle filter (PF) is commonly used for SPE. However, when the parameters switch with the operating conditions, the change of parameters cannot be detected and tracked by the conventional SPE method. This paper proposes a PF-based robust SPE method for a nonlinear process system with variable parameters. The measurement test criterion based on observation error is introduced to indirectly identify whether the parameters are changed. Based on the result of identification, the variances of the particles are modified adaptively for the tracking of the changed parameters. Finally, reliable SPE can be derived through iterative particles. The proposed PF-based robust SPE method is applied to two nonlinear process systems. The results demonstrate the effectiveness and robustness of the proposed method.