Scholarlink

This paper proposes RBFRMS(an adaptive sliding mode control strategy integrating Radial Basis Function neural networks with Root Mean Square Propagation optimization) to address nonlinearity, uncertainty, and time-varying parameter challenges in wind turbine pitch systems under dynamic wind conditions. The RBF network approximates system dynamics in real time, while RMSProp dynamically adjusts weight learning rates and suppresses jitter via gradient square moving-average attenuation. Validated on a SIMPACK/Simulink co-simulation platform with a rigid-flexible coupled wind turbine model, RBFRMS achieves:Under step winds: Rotor speed fluctuations reduced by 72.1–78.4% versus PID; Blade root bending moment means lowered by 4.7–5.2%; Power overshoot maintained 2.0–3.3% below PID (stable at 12.83 ± 0.04 MW). Power standard deviation decreased by 66.5–78.3%, enabling ± 1.2% power stability.In random turbulence: Power output variance reduced by 64.1–67.2%; Rotor speed deviation decreased by 63.8%; Grid frequency fluctuations cut by ~ 50%. However, RBFRMS increases blade root bending moment stdev by 39.9% and yields 4.9–6.0% lower pitch control versus PID, indicating mechanical fatigue trade-offs.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

RBF Adaptive Sliding Mode Pitch Control Optimized by RMSprop Algorithm

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Wenxuan Wang,
  • Caichao Zhu,
  • Chengwu Li,
  • Shuyi Yang,
  • Ye Zhou

摘要

This paper proposes RBFRMS(an adaptive sliding mode control strategy integrating Radial Basis Function neural networks with Root Mean Square Propagation optimization) to address nonlinearity, uncertainty, and time-varying parameter challenges in wind turbine pitch systems under dynamic wind conditions. The RBF network approximates system dynamics in real time, while RMSProp dynamically adjusts weight learning rates and suppresses jitter via gradient square moving-average attenuation. Validated on a SIMPACK/Simulink co-simulation platform with a rigid-flexible coupled wind turbine model, RBFRMS achieves:Under step winds: Rotor speed fluctuations reduced by 72.1–78.4% versus PID; Blade root bending moment means lowered by 4.7–5.2%; Power overshoot maintained 2.0–3.3% below PID (stable at 12.83 ± 0.04 MW). Power standard deviation decreased by 66.5–78.3%, enabling ± 1.2% power stability.In random turbulence: Power output variance reduced by 64.1–67.2%; Rotor speed deviation decreased by 63.8%; Grid frequency fluctuations cut by ~ 50%. However, RBFRMS increases blade root bending moment stdev by 39.9% and yields 4.9–6.0% lower pitch control versus PID, indicating mechanical fatigue trade-offs.