A Design of ZYNQ-Based Intermediate Frequency Stabilization System by Using BP Neural Network PID

In this paper, an algorithm that combines Back Propagation (BP) neural network and Proportional-Integral-Derivative (PID) is deployed on the Zynq development board and applied to the Polarimetry-Interferometry (POINT) system on Experimental Advanced Superconducting Tokamak (EAST). The data acquisition, frequency calculation, and data output are completed by the PL part of Zynq connected to Analog-to-Digital converter (ADC) and Digital-to-Analog converter (DAC), while the control quantity calculation based on BP neural network PID is completed by PS part. PL and PS communicate and exchange data through the AXI protocol. At present, the program has been implemented and the abovementioned functions have been realized. Afterward, simulation results were also provided, which showed that the algorithm can adjust PID parameters (K<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</inf>, K<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i</inf>, K<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</inf>) online for different conditions, demonstrating good adaptability and robustness. In the future, the frequency stabilization system will be applied to the POINT system on EAST.