Page - 85 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

4.1. AdaptiveControl 4.1. AdaptiveControl In general, the control methods that adjust their control actions based on the varying parameters of the controlled systems are called adap- tive control [KV86]. They can be divided into two categories, indirect and direct adaptive control. The indirect control scheme contains on- linesystemidentificationtodescribethesystemdynamics,usingmea- sured system data. In direct adaptive control, no online system iden- tification is needed and the control action relies on direct feedbacks from the plant, such as the most well-known PID controller. Indirect controlschemesaremoresuitableforsystemsmodeledwithunknown constants and slow-varying parameters, while direct control schemes are mainly used to deal with systems with quickly varying dynamics. InHEPHAISTOS,bothcontrol schemeshavebeenapplied inthe tem- peraturecontrolsystem: modelpredictivecontrol(MPC)asanindirect adaptive control scheme and neural network based control (NNC) as a direct/semi-direct adaptive control. Details of MPC and NNC are introducedin this section. 4.1.1. ModelPredictiveControl Model predictive control (MPC), oriented from the receding horizon control (RHC) [MHL99], is an advanced and well-developed feed- back control method. It was firstly implemented in the oil industry [RRTP78]andthenspreadtootherfieldslikechemistryandaerospace [QB03]. In this section, the fundamentals of MPC are introduced at first, and then details of MPC implementations regarding two former derived models (equation 3.41 and 3.46) are presented (linear MPC andnonlinearMPC). No matter which type of models MPC is applied on, the control steps are thesame,whichareaccordingto [CA13] • Predict the future system output based on the given or estimated systemparameters. • Calculate a control sequence that minimizes the predefined objec- tive function. 85