A modified LQG benchmark for economic performance assessment of model predictive control

摘要

In this paper, an improved LQG benchmark for the economic performance assessment of model predictive control system is presented. The LQG benchmark is obtained by a regression of a series of discrete points calculated through solving LQG problem. The discrete points obtained by traditional method present an unbalanced distribution, thus leading to an unsatisfied regression curve especially at the lower part where the distribution is sparse. However, the lower part of the LQG trade-off curve is more significant for the economic performance assessment as they are closer to the minimum variance. To avoid this distribution phenomenon and get a better regression performance curve, we change the form of the LQG benchmark and propose a mathematic method which uses an exponential increasing weighting factor to build the discrete points. The effectiveness of the proposed approach is illustrated by a simulation example to economic performance assessment of a model predictive control system.