Page - 183 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

6. SummaryandConclusion This dissertation gives a detailed introduction of microwave heating systems with spatially distributed heating sources. The innovative temperature control methods that are implemented in HEPHAISTOS achieved promising performance with respect to the temperature ho- mogeneityoftheheat load. Thebasicprinciplesofmicrowaveheating and microwave heating systems are explained. In order to accurately describethedynamicsofthethemicrowaveheatingprocess, threedif- ferentheatingmodelshavebeendeveloped: the linearheatingmodel, the nonlinear heating model and the neural network based black-box model. Parameters within each model are estimated online using the correspondingsystemidentificationmethod. Basedontheestimatedparameters,variousadaptivecontrolmethods, including the model predictive control (MPC) method and the neural networkbasedcontrol (NNC)method,are implemented. Bothcontrol methodshavebeenmodifiedandoptimizedaccordingtothepractical situationsandrequirementsofthemicrowaveheatingsystem,inorder to improvecontrolperformance. Besides the adaptive control scheme that focuses on individual mea- sured temperatures, another reinforcement learning based intelligent control (RLC) scheme is built in this dissertation. The dynamics of HEPHAISTOS are described by a Markov decision process. Then the control task is solved using a look-up table basedQ(λ) controller, which is simplified and optimized based on the empirical knowl- edge. The aforementioned heating models and control methods have been tested in practical experiments and the corresponding experimental data are presented. Firstly the validity of different models is verified throughnumerouspreliminarytests. Thendifferentsystemidentifica- tionalgorithmsaretestedandcomparedusingrealexperimentaldata. 183