Photovoltaic inverter controllers seeking AC optimal power flow solutions

Emiliano Dall’Anese, Sairaj V. Dhople, Georgios B. GiannakisIEEE Transactions on Power Systems 2015 · arXiv

TL;DR

This paper introduces a novel feedback control strategy for photovoltaic (PV) inverters that enables them to dynamically track optimal power flow (OPF) solutions in real-time. By bridging the gap between slow, system-wide optimization and fast inverter control, this approach promises to enhance grid stability and efficiency in increasingly PV-dominated distribution networks.

What Problem Does This Paper Address?

Modern distribution grids are seeing a surge in inverter-interfaced renewable energy sources, particularly PV systems. While these systems are excellent at maximizing energy harvest, their uncoordinated operation can lead to grid instability, voltage issues, and inefficiencies. Traditional approaches to optimizing grid operation, like Optimal Power Flow (OPF), are typically solved offline, resulting in setpoints that are outdated by the time they are implemented by fast-acting inverters. This creates a temporal mismatch. Furthermore, existing distributed OPF methods often require full convergence before updating inverter setpoints, making them slow to react to changing grid conditions. This paper aims to bridge this gap by developing a distributed feedback controller that allows PV inverters to continuously pursue and implement OPF-optimal setpoints in real-time, even under asynchronous updates and faster-than-settling-time reference signals.

Methods

The core of the proposed solution lies in a dual $\epsilon$-subgradient method. This approach is used to solve a convex relaxation of the AC Optimal Power Flow (OPF) problem, specifically using semidefinite programming (SDP) relaxations to handle the inherent non-convexity of AC OPF.

Here's a breakdown of the key methodological components:

  • OPF Formulation: The paper starts by formulating a standard AC OPF problem for distribution systems, considering inverter operational constraints and power balance equations.
  • SDP Relaxation: To overcome the non-convexity of the AC OPF, the problem is reformulated as a semidefinite program (SDP) by dropping the rank-1 constraint on the voltage outer-product matrix. This allows for the application of convex optimization techniques.
  • Dual $\epsilon$-Subgradient Method: The authors leverage dual $\epsilon$-subgradient methods to design a feedback controller. This method iteratively updates dual variables (Lagrange multipliers) and primal variables (voltage-related matrices and inverter setpoints).
  • Feedback Controller Design: The controller is designed to continuously update inverter setpoints based on current system outputs and problem parameters. This is achieved by using sampled inverter outputs to update the dual variables, which in turn inform the calculation of new reference setpoints for the inverters.
  • Asynchronous Updates: A significant contribution is the handling of asynchronous updates. The controller is designed to function even when computational limits lead to slower updates of the SDP-relaxed OPF solution (matrix V) compared to the faster updates of inverter setpoints. This is managed by introducing a time-scaling factor (M) and ensuring the step-size rules are diminishing but square-summable.
  • Distributed Implementation: The proposed controller is inherently distributed, with individual PV inverters performing local updates and a central DSO (or equivalent entity) handling the slower, system-wide optimization steps.

Datasets

The paper evaluates the proposed control scheme using modified versions of two standard test feeders:

  • IEEE 37-node test feeder: Used for initial validation and demonstration of the controller's capabilities.
  • IEEE 123-node test feeder: Used to demonstrate scalability and performance on a larger, more complex network.

The authors note that the network topology, line impedances, shunt admittances, and active/reactive loads are adopted from these standard datasets.

Key Insights and Ideas

  • Bridging the Time-Scale Gap: The most crucial insight is the successful development of a control strategy that effectively bridges the temporal gap between slow, system-wide OPF optimization and fast, real-time inverter control. This is achieved by continuously updating inverter setpoints based on evolving dual variables, rather than waiting for full OPF convergence.
  • Real-time OPF Tracking: The controller allows inverters to dynamically track OPF solutions, enabling them to adapt to changing grid conditions (e.g., solar irradiance, load variations) and actively participate in grid support functions like voltage regulation and loss minimization.
  • Handling Asynchronicity and Computational Limits: The framework's ability to handle asynchronous updates and computational limitations in solving the SDP relaxation is a significant practical advancement. This makes the approach more robust and applicable to real-world scenarios where perfect synchronization is not feasible.
  • Privacy Preservation: By exchanging only Lagrange multipliers (dual variables) between inverters and the DSO, the proposed scheme inherently preserves the privacy of individual PV owners' load and generation data.
  • Analytical Convergence Guarantees: The paper provides analytical proofs for the global convergence of inverter output powers to the OPF solution, even under the challenging conditions of asynchronous updates and faster-than-settling-time reference signals.

Key Results

  • Global Convergence: The paper analytically establishes global convergence of PV inverter output powers to the AC OPF solution for both synchronous and asynchronous update scenarios.
  • Numerical Validation: Simulation results on the IEEE 37-node and 123-node test feeders demonstrate that the proposed controllers successfully drive inverter output powers to the OPF solutions.
  • Adaptability to Changing Conditions: The simulations show that the controllers can adapt to varying solar irradiance levels and PV system configurations, maintaining voltage regulation within acceptable limits.
  • Performance on Larger Networks: The tests on the IEEE 123-node feeder indicate that the convergence speed is not significantly degraded with an increased number of controlled PV systems.
  • Voltage Regulation: The scheme effectively contributes to voltage regulation, keeping voltages within specified limits even during peak generation periods.

Limitations and Future Work

  • Convexity Assumption: The current framework relies on SDP relaxation, which assumes the OPF problem can be effectively approximated by a convex problem. The paper acknowledges that the convergence of dual $\epsilon$-subgradient methods is guaranteed for convex problems.
  • Future Work on Non-Convex OPF: The authors explicitly state that developing provably convergent feedback controllers for non-convex OPF formulations will be the subject of future research efforts.
  • Multi-Phase Systems: While the paper briefly discusses the extension to multi-phase systems, the detailed analysis and simulations focus on balanced, single-phase networks.
  • Computational Complexity of SDP: Although the paper mentions techniques to speed up SDP computation, the inherent complexity of SDP solvers can still be a limiting factor for very large-scale systems.
  • Model Simplifications: The paper uses first-order models for inverter dynamics and assumes certain properties of the system (e.g., Lipschitz continuity of certain functions). Real-world inverter dynamics and system behaviors might be more complex.

Summary updated

Abstract

This paper considers future distribution networks featuring inverter-interfaced photovoltaic (PV) systems, and addresses the synthesis of feedback controllers that seek real- and reactive-power inverter setpoints corresponding to AC optimal power flow (OPF) solutions. The objective is to bridge the temporal gap between long-term system optimization and real-time inverter control, and enable seamless PV-owner participation without compromising system efficiency and stability. The design of the controllers is grounded on a dual ϵ-subgradient method, and semidefinite programming relaxations are advocated to bypass the non-convexity of AC OPF formulations. Global convergence of inverter output powers is analytically established for diminishing stepsize rules for cases where: i) computational limits dictate asynchronous updates of the controller signals, and ii) inverter reference inputs may be updated at a faster rate than the power-output settling time.
BibTeX
@article{dallanese2015photovoltaic,
  author = {Emiliano Dall’Anese and Sairaj V. Dhople and Georgios B. Giannakis},
  title = {{Photovoltaic inverter controllers seeking AC optimal power flow solutions}},
  journal = {IEEE Transactions on Power Systems},
  year = {2015},
}
Figure