Study On Design And Application Of High Speed Model Predictive Control Algorithm

Model Predictive Control (MPC) has wildly applied in industry for itsbetter performance on coping with constraints and controlling effect, which isin various industries’ good graces. However, the online optimizedcomputational complexity of MPC has become the focus, which restricts theapplication of MPC in high-order, complex and fast system. Starting fromthese points of view, this dissertation studies and designs a high speed modelpredictive control algorithm based on aggregation strategy, and makes arealization and application in3dof helicopter experiment performance andpower inverter system.The main work of this dissertation can be summarized as the following3aspects:1) Study on high speed MPC algorithm based on aggregation strategy.Firstly, the MPC problem based on aggregation strategy will be concluded.The designs of structure and the flow of the algorithm follow after. Analysison feasibility and computational complexity of the algorithm is detailed.Simulation compared with QP solver shows that the solved speed of theproposed algorithm has reached a level of several tens microseconds, whichcan provide reference for applying MPC in higher fast system.2) Study on realization of the proposed algorithm in3dof helicopterexperiment performance. Analysis on modelling3dof helicopter bymechanism and linearization is introduced. Description of the helicoptercontrol problem and simulation follow after. And realization of the fast MPC controller in MATLAB environment is presented. Experiment results showthat it is feasible to realize high speed MPC algorithm in motion controlsystem.3) Study on application of the proposed algorithm in power invertersystem. Modelling the inverter system is described first. The MPC problemand the design of the control scheme of the inverter are provided. And therealization of the fast MPC controller of the inverter system in DSP embeddedperformance is carried out. Experiment of power invert control system showsthat it is feasible to apply MPC in low embedded fast system.