Active Disturbance Rejection Control with Sensor Noise Suppressing Observer for DC-DC Buck Power Converters.

Krzysztof Łakomy, Rafał Madoński, Bin Dai, Jun Yang, Piotr Kicki, Maral Ansari, Shihua LiarXiv 2020 · arXiv

Abstract

The class of active disturbance rejection control (ADRC) algorithms has been shown in the literature to be an interesting alternative to standard control methods in power electronics devices. However, their robustness and stability are often limited in practice by the high-frequency measurement noise, common in industrial applications. In this article, this problem is addressed by replacing the conventional high-gain extended state observer (ESO) with a new cascade observer structure. The presented experimental results, performed on a DC-DC buck power converter system, show that the new cascade ESO design has increased estimation/control performance compared to the standard approach, while effectively suppressing the detrimental effect of sensor noise over-amplification.

Introduction

DC-DC buck converters are fundamental components in modern power electronics, used extensively in renewable energy systems and battery management. Precise voltage regulation is critical, yet challenging due to internal model uncertainties and external load variations, collectively termed "total disturbances."

Active Disturbance Rejection Control (ADRC) has emerged as a robust alternative to traditional PID control because it estimates and cancels these disturbances in real-time using an Extended State Observer (ESO). However, a significant practical hurdle remains: the "noise explosion" phenomenon. Achieving high estimation accuracy typically requires high observer gains, which unfortunately amplify high-frequency sensor noise, leading to jittery control signals and potential hardware degradation. This paper introduces a Cascade Extended State Observer (CESO) to break this trade-off.

The Proposed Cascade ESO (CESO)

To address the noise sensitivity of standard ADRC, the authors propose a $p$-level cascade structure. Instead of one high-gain observer, the CESO decomposes the disturbance estimation process into multiple stages.

Block diagram of the proposed control system featuring the p-th level cascade ESO (CESO) and state selector.

In this architecture, the first level provides a primary estimate while acting as a low-pass filter for sensor noise. Subsequent levels ($p=2, 3, \dots$) then estimate the "residual" disturbance that the previous levels could not capture due to bandwidth constraints. A state selector logic synthesizes these estimates into a final control signal. This multi-level approach allows for high effective bandwidth with significantly lower noise sensitivity compared to a conventional high-gain ESO.

Experimental Validation and Results

The proposed method was validated using real-time hardware-in-the-loop experiments on a physical DC-DC buck converter paired with a dSPACE controller. The system was tasked with following a filtered square-wave voltage trajectory under time-varying external disturbances.

Experimental results comparing standard ESO (p=1) with CESO (p=2, 3). Note the significant reduction in control error and the "cleaner" control signal for higher p.

The results demonstrate a dramatic performance leap. For a 3-level CESO ($p=3$), the tracking accuracy improved nearly six-fold, with the integral of absolute error dropping from 0.2310 to 0.0381. Crucially, the control signal "jitter" (measured by the integral of the absolute derivative of the control signal) plummeted from 315.58 to 29.11. This indicates that the CESO provides a much smoother control action, reducing stress on the converter's physical components without sacrificing tracking speed.

Conclusion

By restructuring the observer into a cascade, this work effectively silences the noise amplification inherent in high-gain ADRC. The CESO offers a practical and mathematically grounded solution for high-performance voltage regulation in noisy industrial environments, paving the way for more reliable and efficient power electronic systems.

Summary updated

Abstract

The class of active disturbance rejection control (ADRC) algorithms has been shown in the literature to be an interesting alternative to standard control methods in power electronics devices. However, their robustness and stability are often limited in practice by the high-frequency measurement noise, common in industrial applications. In this article, this problem is addressed by replacing the conventional high-gain extended state observer (ESO) with a new cascade observer structure. The presented experimental results, performed on a DC-DC buck power converter system, show that the new cascade ESO design has increased estimation/control performance compared to the standard approach, while effectively suppressing the detrimental effect of sensor noise over-amplification.
BibTeX
@misc{akomy2020active,
  author = {Krzysztof Łakomy and Rafał Madoński and Bin Dai and Jun Yang and Piotr Kicki and Maral Ansari and Shihua Li},
  title = {{Active Disturbance Rejection Control with Sensor Noise Suppressing Observer for DC-DC Buck Power Converters.}},
  year = {2020},
  eprint = {2009.02948},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2009.02948},
}
Figure