| Model predictive control(MPC) has become an eastablished control method in petrolchemical industry, it proves the practicality of MPC compared with other advanced control method like neural network, optimal control, its use is currently being used in an increasing wide range of higher bandwidth applications, like ships, planes, cars, etc[1] [4]. The new application demands the controller to be smart, intelligent, quick response and low cost, etc. Because MPC algorithm incorporates multi dimension search, the controller needs to do complex computation.FPGA(Field Programmable Gate Array), as a kind of semi-custom IC has widely used in the area of vehicles, industries, communication, aerospace, etc.The small size of FPGA assures the smart of controller; the re-configuration of FPGA assures the intelligent of controller; compared with PC, FPGA is very cheap, therefore it's valuable to research the MPC controller based on FPGA .This paper focus on the design of a multi-variable MPC controller based on FPGA, this controller aims to control some small-scale industry plant.First, some basic knowledges associate with controller based on FPGA are introduced, including MPC algorithm theory and single/multi variable DMC algorithm, multi variable DMC algorithm with constraints and the basic development technologies about FPGA are presented, including the structure and characteristic of FPGA, SOPC technology, NIOS II processor, digital system design and synthesis based on HDL.The main tasks of this paper are:1.The multi ariable DMC controller was implemented in the Altera FPGA, the controller was implemend with SOPC system. This paper first implemented the DMC algorithm in MATLAB platform, then the algorithm was implemented in ANSI C, then the C program based on double-precision was changed to fixed point, in order to accelerate the speed of DMC algorithm and make the transition to HDL by the Impulse C easier. The implementation of MPC controller based on FPGA divided into hardware and software parts, the tasks of hardware design included NIOS II processor custom made, standard peripheral custom made and self-definition peripheral based on Impulse C custom made; the software design included the implementation of multi variable DMC algorithm with C language and the implementation of FPGA and MATLAB communication based on UART.2.The simulation platform based on FPGA and MATLAB/Simulink was eastablished to simulate the multi variable plant in real time. This paper modeled a double dimension tank in Simulink, the controller samples the output of plant by RS-232, the control policy was caculated by the DMC algorithm in FPGA, the the control policy was passed to Simulink by RS-232. |
