基于FPGA的神经网络PID控制器设计与实现

Design and Implementation of Neural Network PID Controller Based on FPGA

由于实际工业生产过程中被控对象非线性、时变不确定性等特性,应用常规的PID控制器难以达到理想的控制效果,而采用神经网络在线自适应整定PID参数的方法,能够适应不同特性的被控对象,但是由于算法较为复杂,目前常用的单片机等控制器无法满足对实时性、可靠性要求较高的工业系统。基于大规模可编程逻辑门阵列(FPGA),采用多层神经元自适应PID控制器的FPGA实现方法,设计完成了基于BP神经网络自整定的PID控制器,通过仿真实验得以验证,并深入分析比较了定点和浮点神经网络PID控制器的性能,为其在工程实际中的应用奠定了基础。

In the actual industrial production process, the method of adaptively tuning the PID parameters online by the neural network can adapt to different characteristics of different controlled objects better than the controller with conventional PID. But the commonly used single-chip and other controllers cannot meet the real-time, high reliability requirements of the industrial system because of its complex algorithm. The FPGA based on large-scale programmable logic gate array is used to study the multi-layer neuron adaptive PID controller. And the BP neural network auto-tuning PID controller is designed and implemented. The performances of fixed-point and floating-point neural network PID controller are validated and compared, which lays the foundation for its application in engineering prac- tice.