<p>To achieve higher precision current output for DC magnet power supplies, this study proposes an integral separation–adaptive feedforward control algorithm. Conventional PI controllers face inherent limitations in accelerator applications, including bandwidth-limited feedback regulation against multi-frequency disturbances and delayed dynamic response due to integral action. This algorithm addresses these challenges by incorporating an adaptive feedforward path alongside the conventional PI feedback loop. By enabling dynamic integral separation and using the separated integral term to iteratively update the feedforward compensation table, the method improves transient performance while achieving high steady-state precision for accelerator magnet power supplies. The algorithm was validated through comprehensive Simulink simulations, which demonstrated significant improvements in dynamic response, robustness against bus voltage variations, and effective suppression of lower-frequency interference. To facilitate practical deployment, a new digital controller, the CSNS Intelligent Power Supply Controller (CIPSC), was developed on the Chinese domestic FMQL series system-on-chip (SoC) platform from Fudan Microelectronics. This controller features a heterogeneous CPU–FPGA architecture optimized for real-time control with modular design and scalability. Experimental validation was conducted on 20 A and 300 A CSNS magnet power supply platforms. Voltage ripple testing on the 20 A platform showed 30% improvement in peak-to-peak values with effective low-frequency suppression. Eight-hour stability tests achieved current deviations of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$\pm 10.69$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo>±</mo> <mn>10.69</mn> </mrow> </math></EquationSource> </InlineEquation> ppm and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$\pm 8.37$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo>±</mo> <mn>8.37</mn> </mrow> </math></EquationSource> </InlineEquation> ppm, respectively, significantly outperforming conventional PI control. These results confirm the algorithm’s effectiveness for high-precision, reliable current regulation in accelerator magnet power supplies.</p>

Integral separation-adaptive feedforward control for high-precision DC magnet power supplies in CSNS

期刊论文
发表日期：

2025年11月29日

查看全文

Ran Li,
Jun Li,
Guo-dong Zhao,
Xiao-ke Zhang,
Wenqing Zhang,
Yuan Huang,
Zhengyong Chai,
Xin Qi

摘要

To achieve higher precision current output for DC magnet power supplies, this study proposes an integral separation–adaptive feedforward control algorithm. Conventional PI controllers face inherent limitations in accelerator applications, including bandwidth-limited feedback regulation against multi-frequency disturbances and delayed dynamic response due to integral action. This algorithm addresses these challenges by incorporating an adaptive feedforward path alongside the conventional PI feedback loop. By enabling dynamic integral separation and using the separated integral term to iteratively update the feedforward compensation table, the method improves transient performance while achieving high steady-state precision for accelerator magnet power supplies. The algorithm was validated through comprehensive Simulink simulations, which demonstrated significant improvements in dynamic response, robustness against bus voltage variations, and effective suppression of lower-frequency interference. To facilitate practical deployment, a new digital controller, the CSNS Intelligent Power Supply Controller (CIPSC), was developed on the Chinese domestic FMQL series system-on-chip (SoC) platform from Fudan Microelectronics. This controller features a heterogeneous CPU–FPGA architecture optimized for real-time control with modular design and scalability. Experimental validation was conducted on 20 A and 300 A CSNS magnet power supply platforms. Voltage ripple testing on the 20 A platform showed 30% improvement in peak-to-peak values with effective low-frequency suppression. Eight-hour stability tests achieved current deviations of $\pm 10.69$ $\pm 10.69$ ppm and $\pm 8.37$ $\pm 8.37$ ppm, respectively, significantly outperforming conventional PI control. These results confirm the algorithm’s effectiveness for high-precision, reliable current regulation in accelerator magnet power supplies.