Robust trajectory-constrained frequency control for microgrids considering model linearization error

Yichen Zhang, Chen Chen, Tianqi Hong, Bai Cui, Zhe Xu, Feng QiuApplied Energy 2022 · DOI · arXiv

Abstract

This paper presents a robust frequency control framework for renewable-rich microgrids, specifically addressing the performance gap caused by model linearization errors. By integrating zonotope-based reachability analysis with a two-level control architecture—comprising centralized scheduling and decentralized triggering—the proposed method ensures frequency stability under significant disturbances. The framework formulates a robust mixed-integer linear program (MILP) to pre-calculate optimal power injection signals from wind turbine generators. Simulation results on a 33-node microgrid demonstrate that the robust controller maintains frequency within safety limits where standard linear-optimal controllers fail, highlighting the necessity of accounting for nonlinear dynamics in microgrid stability.

Introduction

As power systems transition toward high penetrations of renewable energy, microgrids are losing the natural mechanical inertia traditionally provided by heavy synchronous generators. Wind turbine generators (WTGs), interfaced via power electronics, do not inherently respond to frequency deviations. While "Grid Supportive" (GS) modes can force these turbines to inject power during disturbances, a significant challenge remains: the Linearization Gap.

To facilitate real-time optimization, researchers typically linearize the complex, nonlinear differential equations governing WTGs. However, this simplification introduces errors. If a controller is designed based on an overly optimistic linear model, the actual frequency nadir in the physical system may drop below safe thresholds, potentially triggering emergency load shedding or equipment damage.

Methodology

The proposed solution utilizes a two-level control paradigm designed to bridge the gap between linear modeling and nonlinear reality.

  1. Scheduling Level (Centralized): A central controller solves a Numerical Optimal Control (NOC) problem to pre-calculate the optimal utilization of stored kinetic energy in WTGs.
  2. Triggering Level (Decentralized): Local controllers at each turbine monitor the grid frequency and execute the pre-calculated response immediately upon detecting a disturbance.
The two-level control architecture consisting of a centralized scheduling level and a decentralized triggering level.

Quantifying Linearization Error

The core innovation lies in treating linearization error as an "unknown-but-bounded" uncertainty set. Using Zonotope-based reachability analysis and interval arithmetic (via the CORA toolbox), the authors calculate the worst-case difference between the linear model and the true nonlinear system. This error bound is then embedded into a Robust Mixed-Integer Linear Program (MILP), allowing the controller to make discrete decisions regarding GS mode switching while guaranteeing safety margins.

Simulation and Results

The framework was validated using a high-fidelity simulation of a modified 33-node microgrid featuring a diesel generator and multiple Doubly-Fed Induction Generator (DFIG) wind turbines. The control signals were tested against a full-order three-phase nonlinear model in Simulink.

Key Findings

  • Constraint Satisfaction: During a major disturbance (0.7 MW loss), the "Normal" controller failed to maintain frequency within the 0.5 Hz limit when applied to the nonlinear model. In contrast, the Robust controller successfully kept the frequency within bounds by accounting for model inaccuracies.
  • Control Effort: The robust schedule required a higher control effort (an integer-based metric of 42 vs. 33), demonstrating that the turbines must perform additional work to compensate for the nonlinearities of the system.
Comparison of scheduled inputs for WTG 1 and WTG 2 under robust vs. normal scheduling regimes.
  • Computational Efficiency: Despite the added complexity of robustness, the MILP problem was solved in under 10 seconds using the Gurobi solver, making it suitable for near real-time grid updates.

Conclusion

This research demonstrates that controllers optimized for linear models are often insufficient for ensuring the stability of nonlinear microgrid systems. By explicitly modeling linearization error as a bounded uncertainty, the proposed robust framework provides a reliable buffer against frequency instability. Future work will look to extend this methodology to solar PV and battery storage systems, as well as refining triggering thresholds to minimize false alarms while maintaining rapid response times.

Summary updated

BibTeX
@inproceedings{zhang2022robust,
  author = {Yichen Zhang and Chen Chen and Tianqi Hong and Bai Cui and Zhe Xu and Bo Chen and Feng Qiu},
  title = {{Robust trajectory-constrained frequency control for microgrids considering model linearization error}},
  booktitle = {Applied Energy},
  year = {2022},
  doi = {10.1016/j.apenergy.2022.120559},
}
Figure