基于PSO训练的NN-PID控制器设计及其FPGA实现

Design of NN-PID Controller Based on PSO and Its FPGA Implementation

提出了一种基于PSO学习、VHDL描述和FPGA实现的NN-PID控制器设计方法。首先借助MATLAB系统仿真工具,在闭环控制系统中通过PSO优化算法训练前馈网络,得到优化的NN-PID控制器参数;然后在FPGA集成开发环境下进行控制器的VHDL层次化设计,重点研究单个神经元和前馈网络的结构以及实现方式;最后对该控制器进行了闭环时序测试,并在一个具体的FPGA器件上实现。研究结果表明,PSO用于NN-PID控制器训练速度快,VHDL描述和FPGA实现该控制器时序验证方便,而且控制器具有较好的鲁棒性。

In this paper, a design method of NN-PID controller based on PSO learning, VHDL description and FPGA implementation is proposed. At first, with the simulink of MATLAB, the feedforward neural networks went through training based on PSO in the closed-loop control system, and the optimized parameters of NN-PID are obtained. Then under FPGA＇ s development toolkit, a hierarchical design of the controller based on VHDL is carried out, emphasize on the research of single neuron as well as the structure and the realization methods of feedforward neural networks. At last the closed-loop test of the controller, implemented on a FPGA chip, is carried out. The research results indicated that NN-PID controller＇s training speed based on PSO is quick, the timing validation of the controller based on VHDL description and FPGA implementation is convenient and quality of controller is robust.