USING FUZZY LOGIC TO TUNE OPTIMIZATION PARAMETERS IN NEURAL MODEL BASED PREDICTIVE CONTROL TECHNIQUE

摘要

The underlying idea of Model Based Predictive control can be summarized as follows: A plant model is used in order to predict its future behavior. Thus, Neural Networks (NN) are used in this work to build the target plant dynamic model. The control signal is calculated by means of an optimization problem, solved at each sampling period. Selecting "good" optimization parameters ("weights") is a very important issue. However, it is not easy to select these parameters so that some optimization requirements are achieved. In this work, we propose an algorithm for weights fuzzy self-tuning. Optimization requirements are translated in terms of fuzzy rules. As an application, this design technique was applied to solve the perturbation rejection problem for a multivariable benchmark "Continuous Stirred Tank Reactor" (CSTR).