System Identification and Long-range Predictive Control of Multi-rate Systems

Authors

Dongguang Li, S. L. Shah, Tongwen Chen, R. Patwardhan

Year

1999

Conference Paper

Algorithm

Lifted Multirate Predictive Control

Source

American Control Conference (ACC)

Identifiers

DOI 10.1109/ACC.1999.782796

Abstract¶

Discusses system identification and model-based predictive control of multi-rate systems. In particular, the practically useful case of systems with fast manipulative variable sampling and slow output sampling is considered. The lifting method is used to analyze the multi-rate system in a state-space framework. A subspace identification algorithm, is first used to identify the lifted multi-rate system. The single-rate model at the faster sampling rate is then extracted from this estimated system and subsequently used for long-range predictive control. The fast-sampling rate model is used to estimate the outputs at inter-sample instants, using a minimum variance predictor. The estimated output is then used for model-based predictive control at the faster sampling rate. The performance of this controller is compared with the performance of the slow sample-rate scheme, and it is shown that the fast-rate controller has a significantly better performance.

Summary¶

Combines lifted-system identification with long-range predictive control for multirate processes where inputs are sampled faster than outputs. The method identifies a lifted state-space model, extracts a fast-rate model, and uses minimum-variance intersample output estimates to improve predictive-control performance.