Seite - 87 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

4.1. AdaptiveControl Due to above reasons, MPC can be used in the temperature control system of HEPHAISTOS. The entire structure of the MPC based tem- peraturecontrol systemisshownas infigure 4.2. Plant (HEPHAISTOS) MPC Controller Model Estimator (System Identification) Target Temperature Yt Measured Temperature Yr Control Input U Estimated System Dynamics [A], [B] (or [Φ]) Figure4.2. Temperaturecontrol systemofHEPHAISTOSusingMPC. Inthis temperaturecontrolsystem,therearetwokeycomponentsthat determine thefinalcontrolperformance. 1. Modelestimator The model estimator is used to capture the dynamics of the HEP- HAISTOS oven and send the estimated model dynamics, e.g. ma- trices [A(k)],[B(k)] or [Φn(k)], to the MPC controller. The accu- racy of the estimated system model directly influences the final control performance of any MPC algorithms, and the form of the estimatedmodelalsodetermineswhichkindofMPCalgorithmsis appropriate for thegivenapplication. In principle, both the grey-box and black-box modeling ap- proachesintroducedinthelastchaptercouldbeusedasthesystem estimator intheabovefigure 4.2. But inpractice, theevaluationof theNNmodeltakesmuchmoretimethanthatofgrey-boxmodels. In order to keep a fast control pace, only the grey-box approaches (linear 3.41 and nonlinear 3.46 models) are implemented in this model estimator. The NN model will be used in the neural net- workcontroller insection 4.1.2. 87