2025-11-15T12:46:11.588359

Optimal Voltage Control Using Online Exponential Barrier Method

Zhang, Zhang
This paper address the optimal voltage control problem of distribution systems with high penetration of inverter-based renewable energy resources, under inaccurate model information. We propose the online exponential barrier method that explicitly leverages the online feedback from grids to enhance the robustness to model inaccuracy and incorporates the voltage constraints to maintain the safety requirements. We provide analytical results on the optimal barrier parameter selection and sufficient conditions for the safety guarantee of converged voltages. We also establish theoretical results on the exponential convergence rate with proper step-size. The effectiveness of the proposed framework is validated on a 56-bus radial network, where we significantly improve the robustness against model inaccuracy compared to existing methods.
academic

Optimal Voltage Control Using Online Exponential Barrier Method

Basic Information

  • Paper ID: 2506.10247
  • Title: Optimal Voltage Control Using Online Exponential Barrier Method
  • Authors: Peng Zhang (University of Wisconsin–Madison), Baosen Zhang (University of Washington)
  • Classification: math.OC cs.SY eess.SY
  • Publication Date: 2025 (arXiv preprint)
  • Paper Link: https://arxiv.org/abs/2506.10247

Abstract

This paper investigates the optimal voltage control problem in distribution systems with high penetration of inverter-based renewable energy resources under inaccurate model information. An online exponential barrier method is proposed that explicitly leverages online feedback from the power grid to enhance robustness against model inaccuracies while incorporating voltage constraints to maintain safety requirements. The paper provides analytical results for optimal barrier parameter selection and sufficient conditions for convergent voltage safety guarantees, as well as theoretical results establishing exponential convergence rates under appropriate step sizes. The effectiveness of the proposed framework is validated on a 56-node radial network, demonstrating significantly improved robustness against model inaccuracies compared to existing methods.

Research Background and Motivation

Problem Background

  1. Importance of Voltage Control: With large-scale integration of distributed energy resources (DERs), distribution systems face greater voltage fluctuations. Traditional mechanical devices (such as on-load tap-changing transformers and shunt capacitors) respond slowly and struggle to meet fast voltage control requirements.
  2. Model Uncertainty Challenge: Traditional optimization methods require detailed parameter and topology information, but in practical applications this information is often inaccurate or unavailable.
  3. Safety Requirements: Voltage control must satisfy voltage constraints during intermediate iterations to prevent equipment damage and system instability.

Limitations of Existing Methods

  1. Two-Stage Approaches: First learn the system model, then perform model-based control. However, it remains unclear how to control the system during the data collection phase and how to handle system changes.
  2. Online Primal-Dual Methods: While achieving feasibility at convergence, they cannot guarantee constraint satisfaction during intermediate iterations.
  3. Logarithmic Barrier Methods: Can enforce constraint satisfaction, but barrier weight selection is heuristic, leading to slow convergence and optimality gaps.

Research Motivation

This paper aims to develop an online voltage control method that guarantees both optimality and safety, particularly when distribution network models have significant estimation errors.

Core Contributions

  1. Proposes Online Exponential Barrier Method: Integrates voltage constraints as exponential barriers into the objective function, utilizing real-time feedback from the physical power grid for real-time optimization.
  2. Theoretical Analysis:
    • Provides analytical results for optimal barrier parameter selection
    • Establishes sufficient conditions for convergent voltage safety guarantees
    • Proves exponential convergence rates under appropriate step sizes
  3. Safety Guarantees: Significantly reduces voltage constraint violations during intermediate iterations compared to existing methods.
  4. Robustness Enhancement: Validation on a 56-node radial network demonstrates significantly improved robustness against model inaccuracies compared to existing methods.

Method Details

Task Definition

Inputs:

  • Inaccurate model estimate B^\hat{B} of the distribution system
  • Voltage safety limits [x‾,xˉ][\underline{x}, \bar{x}]
  • Inverter capacity limits [u‾,uˉ][\underline{u}, \bar{u}]
  • Online voltage measurements

Outputs:

  • Active and reactive power adjustments for inverters u=(up,uq)u = (u_p, u_q)

Constraints:

  • Voltage safety constraints: x‾≤x≤xˉ\underline{x} \leq x \leq \bar{x}
  • Power capacity constraints: u‾≤u≤uˉ\underline{u} \leq u \leq \bar{u}

Model Architecture

1. System Modeling

Employs linearized AC power flow equations: x=Rup+Xuq+ex = Ru_p + Xu_q + e

where:

  • x∈Rnx \in \mathbb{R}^n: normalized voltage deviation vector
  • R,X∈Rn×nR, X \in \mathbb{R}^{n \times n}: resistance and reactance matrices
  • e∈Rne \in \mathbb{R}^n: voltage drop caused by baseline solar generation and external loads

2. Optimization Problem Formulation

For overvoltage problems, the original problem is reformulated as:

\min_{u \in \mathbb{R}^{2n}} &\quad c(u) \\ \text{s.t.} &\quad Bu + e \leq \bar{x} \\ &\quad \underline{u} \leq u \leq \bar{u} \end{align}$$ where $B = [R \quad X] \in \mathbb{R}^{n \times 2n}$, $c(u) = \frac{1}{2}u^T Q u$. #### 3. Exponential Barrier Method Integrates constraints into the objective function: $$\tilde{c}(u|\hat{\alpha}, \hat{B}) = c(u) + \sum_i \frac{\hat{\alpha}_i}{\beta_i} e^{\beta_i(\hat{b}_i^T u + e_i - \bar{x}_i)}$$ where $\hat{\alpha}_i, \beta_i$ are barrier weight parameters. #### 4. Gradient Update Computes gradients based on latest voltage measurements: $$F(u|\hat{\alpha}) := \nabla_u c(u) + \sum_i \hat{\alpha}_i e^{\beta_i(x_i - \bar{x}_i)} \hat{b}_i$$ ### Technical Innovations 1. **Exponential vs. Logarithmic Barriers**: - Exponential barriers avoid explosion problems near boundaries - Allow direct computation of barrier parameters, avoiding double-loop processes - Enable analysis of voltage dynamics on boundaries 2. **Online Feedback Mechanism**: - Utilizes actual voltage measurements $x_i = b_i^T u + e_i$ rather than model predictions - Automatically corrects cumulative effects of model errors 3. **Adaptive Parameter Adjustment**: - Dynamically adjusts focus to critical nodes (nodes with highest voltages) - Handles weight updates for inverter saturation cases ## Experimental Setup ### Dataset - **Test System**: 56-node single-phase radial distribution network - **Line Parameters**: Standard test data from literature - **PV Inverter Configuration**: Randomly installed at selected nodes, active power range $u_p \in [-p_{av}, 0]$, reactive power range $u_q \in [-0.4p_{av}, 0.4p_{av}]$ ### Evaluation Metrics - **Optimality**: Voltage deviation from known optimal solution with correct model - **Safety**: Degree of constraint violations during intermediate iterations - **Convergence**: Number of iterations required to reach convergence ### Comparison Methods 1. **No Control Baseline** 2. **Direct LCQP Solving** (using correct and incorrect models) 3. **Regularized Online Primal-Dual Method** ### Implementation Details - **Voltage Safety Range**: ±5% of nominal voltage 12kV - **Cost Function Parameters**: $c_p = 3$, $c_q = 1$ - **Model Error**: $\varepsilon_B/\|B\| = 44.1\%$ and $52.8\%$ - **Step Size**: $\eta = 0.01$ ## Experimental Results ### Main Results 1. **Voltage Regulation Performance**: - No Control: Maximum voltage 12.77kV (exceeds safety range) - Direct LCQP Solving (incorrect model): 12.39kV (overly conservative) - Proposed Method: 12.59kV (near safety limit, optimal) 2. **Convergence Speed**: - Proposed Method: Converges in fewer than 20 iterations - Online Primal-Dual Method: Exceeds 100 iterations ### Safety Analysis - **Proposed Method**: Nearly zero constraint violations throughout the entire iteration process - **Online Primal-Dual Method**: Significant constraint violations in intermediate steps ### Barrier Parameter Settings | Inaccurate Model | Safety Factor $\gamma^s$ | Barrier Weight $\alpha^s$ | |------------------|-------------------------|--------------------------| | $\hat{B}_1$ | $2.7 \times 10^{-3}$ | $3.5 \times 10^{-3}$ | | $\hat{B}_2$ | $4.5 \times 10^{-3}$ | $5.4 \times 10^{-3}$ | ### Experimental Findings 1. **Robustness**: Effectively controls voltage even with 52.8% model error 2. **Real-Time Performance**: Fast convergence characteristics make it suitable for real-time applications 3. **Safety**: Significantly reduces intermediate constraint violations compared to existing methods ## Related Work ### Main Research Directions 1. **Optimization-Based Inverter Voltage Control**: Formulates problems as optimal power flow (OPF) 2. **Linearized Models**: Simplifies computational complexity of non-convex OPF problems 3. **Online Learning and Control**: Simultaneously performs learning and control to handle model uncertainty 4. **Barrier Methods**: Interior point methods for constrained optimization ### Advantages of This Work 1. **vs. Two-Stage Methods**: Avoids control issues during data collection phase 2. **vs. Online Primal-Dual Methods**: Guarantees safety during intermediate steps 3. **vs. Logarithmic Barrier Methods**: Provides analytical parameter selection methods ## Conclusions and Discussion ### Main Conclusions 1. **Method Effectiveness**: The online exponential barrier method achieves optimal and safe voltage control under model inaccuracy 2. **Theoretical Guarantees**: Provides complete theoretical analysis of barrier parameter selection, safety guarantees, and convergence 3. **Practical Value**: Method is compatible with existing online system identification techniques, achieving good balance between optimality and safety ### Limitations 1. **Model Assumptions**: Relies on accuracy of linearized power flow equations 2. **Overvoltage Focus**: Primarily addresses overvoltage problems; handling of undervoltage is limited 3. **Static Scenarios**: Current analysis mainly addresses static cases; tracking characteristics for time-varying loads and solar generation require further research ### Future Directions 1. **Global Convergence Analysis**: Extend to more general convergence analysis 2. **Time-Varying Scenarios**: Handle time-varying characteristics of loads and solar generation 3. **Multi-Objective Optimization**: Simultaneously consider economic and technical constraints ## In-Depth Evaluation ### Strengths 1. **Theoretical Completeness**: Provides complete theoretical framework from barrier parameter selection to convergence analysis 2. **Strong Practicality**: Method design considers model uncertainty issues in practical applications 3. **Safety Guarantees**: Significantly improves operational safety compared to existing methods 4. **Computational Efficiency**: Avoids double-loop optimization process, improving computational efficiency ### Weaknesses 1. **Limited Application Scope**: Primarily targets radial distribution networks; applicability to meshed networks requires verification 2. **Parameter Sensitivity**: Barrier parameter selection may require adjustment for different systems 3. **Experimental Scale**: Validation only on 56-node system; lacks verification on large-scale systems ### Impact 1. **Academic Contribution**: Provides new barrier method framework for online optimization control theory 2. **Engineering Application**: Offers practical solution for real-time voltage control in smart grids 3. **Reproducibility**: Detailed algorithm description and complete theoretical analysis ensure good reproducibility ### Applicable Scenarios 1. **Distribution systems with high renewable energy penetration** 2. **Power grids with incomplete or inaccurate model information** 3. **Real-time control scenarios with high safety requirements** 4. **Fast-response voltage regulation applications** ## References The paper cites 26 related references covering multiple relevant fields including distribution system optimization, online learning control, and barrier methods, providing a solid theoretical foundation for this research.