Optimal Dispatch of Photovoltaic Inverters in Residential Distribution Systems

Emiliano Dall’Anese, Sairaj V. Dhople, Georgios B. Giannakis

TL;DR

This paper proposes a novel framework for optimally dispatching photovoltaic (PV) inverters in residential distribution systems to manage power quality and voltage regulation issues arising from increased PV penetration. By formulating the problem as a convex optimization problem using sparsity-promoting regularization and semidefinite relaxation, the method efficiently determines which inverters to control and their optimal active/reactive power setpoints, leading to improved system performance and reduced losses.

What Problem Does This Paper Address?

Traditional low-voltage distribution feeders were designed for unidirectional power flow from the utility to consumers. The widespread adoption of rooftop PV systems is disrupting this paradigm, leading to challenges like overvoltages when PV generation exceeds local demand and voltage sags during rapid irradiance changes. These issues can compromise power quality and system reliability. While existing solutions like reactive power control (RPC) and active power curtailment (APC) offer some relief, they often rely on local information, leading to suboptimal system-wide performance, or require pre-selecting which inverters to control. This paper addresses the need for a systematic, network-wide approach to optimize PV inverter operation, enabling them to provide ancillary services for voltage regulation and loss minimization, while also intelligently selecting which inverters should actively participate in these services.

Methods

The core of the paper lies in formulating and solving the Optimal Inverter Dispatch (OID) problem. Initially, the OID problem is cast as a mixed-integer nonlinear program (MINP) due to the binary selection variables for inverters and the nonlinear power-flow equations. To overcome the computational complexity (NP-hard), the authors employ two key techniques:

  1. Sparsity-Promoting Regularization: To address the binary inverter selection, a group-Lasso-type regularization function is introduced. This encourages sparsity in the control actions, effectively allowing the framework to select a minimal subset of PV inverters for control, rather than requiring all inverters to participate. This is crucial for practical implementation and cost-effectiveness.
  2. Semidefinite Relaxation (SDR): The nonconvexity arising from nonlinear power-balance constraints and bilinear terms is tackled using semidefinite relaxation. This technique transforms the original problem into a semidefinite program (SDP), which can be efficiently solved using standard solvers. The authors demonstrate that for radial distribution networks, the SDR often yields a rank-1 solution, meaning the relaxed problem's solution is the globally optimal solution to the original nonconvex problem.

The paper also formulates cost functions that consider power losses, costs associated with curtailing active power, and voltage profile flatness, allowing for a multi-objective optimization.

Datasets

The paper utilizes real-world data for its case studies:

  • PV Generation Data: Generated using the System Advisor Model (SAM) from the National Renewable Energy Laboratory (NREL), based on typical meteorological year (TMY) data for Minneapolis, MN, during July.
  • Load Profile Data: Obtained from the Open Energy Info database, specifically using base load data for downtown Saint Paul, MN, during July. This data is then perturbed with a Gaussian random variable to simulate different load profiles.
  • Network Data: An illustrative low-voltage residential distribution network is adopted from existing literature ([3], [13]), complete with line parameters (resistance, inductance, capacitance) and topology.

No new benchmarks or datasets were introduced; rather, existing, realistic data was employed to validate the proposed method.

Key Insights and Ideas

  • Network-Level Optimality: Unlike local control strategies, the proposed OID framework optimizes the entire distribution feeder, leading to better overall system performance and guaranteed system-level benefits.
  • Intelligent Inverter Selection: The sparsity-promoting regularization is a critical innovation, enabling the framework to identify the most effective PV inverters for ancillary services, reducing the burden on individual homeowners and minimizing control costs.
  • Unified Control Framework: The OID strategy seamlessly integrates RPC and APC, offering greater flexibility than traditional methods by jointly optimizing active and reactive power.
  • Computational Feasibility: The combination of sparsity promotion and SDR transforms a computationally intractable problem into an efficiently solvable SDP, making real-time implementation feasible.
  • Rank-1 Guarantee: For radial networks, the SDR approach is shown to often yield a rank-1 solution, which is the globally optimal solution, a significant advantage over other relaxation techniques.

Key Results

  • The proposed OID strategy effectively enforces voltage regulation, keeping voltage magnitudes within acceptable limits across various scenarios.
  • When minimizing active power losses (with no cost for curtailment), the OID strategy achieved the lowest network losses compared to RPC and APC strategies.
  • When minimizing overall power losses (including curtailed power), the OID strategy also demonstrated superior performance, yielding significantly lower total losses.
  • The sparsity-promoting regularization effectively reduced the number of controlled inverters from 12 to 7 in a specific scenario, demonstrating its capability for inverter selection.
  • Long-term simulations (over a year) showed that the OID strategy leads to substantial economic savings by minimizing both network losses and curtailed PV energy. For instance, in one scenario, the OID strategy resulted in an annual economic loss of $499.46, compared to $1520.45 for APC.

Limitations and Future Work

  • Forecast and Communication Errors: The authors acknowledge that future work should investigate the impact of forecast errors (for PV generation and load) and communication delays on the inverter dispatch task.
  • Experimental Validation: While the approach is demonstrated with real-world data, the paper suggests targeting experimental demonstrations in real-world setups to further validate its effectiveness and communicate its benefits to practicing engineers.
  • Unbalanced Systems: While the paper mentions that extensions for unbalanced systems exist ([21]), the primary focus and case studies are on radial networks. Further exploration of complex unbalanced scenarios might be beneficial.
  • Generalization of SDR: While the SDR approach is shown to often yield a rank-1 solution for radial networks, the paper notes that for general networks, rank reduction techniques might be needed, which could lead to suboptimal solutions.

Summary updated

Abstract

Low-voltage distribution feeders were designed to sustain unidirectional power flows to residential neighborhoods. The increased penetration of roof-top photovoltaic (PV) systems has highlighted pressing needs to address power quality and reliability concerns, especially when PV generation exceeds the household demand. A systematic method for determining the active- and reactive-power set points for PV inverters in residential systems is proposed in this paper, with the objective of optimizing the operation of the distribution feeder and ensuring voltage regulation. Binary PV-inverter selection variables and nonlinear power-flow relations render the optimal inverter dispatch problem nonconvex and NP-hard. Nevertheless, sparsity-promoting regularization approaches and semidefinite relaxation techniques are leveraged to obtain a computationally feasible convex reformulation. The merits of the proposed approach are demonstrated using real-world PV-generation and load-profile data for an illustrative low-voltage residential distribution system.
BibTeX
@article{dallanese2014optimal,
  author = {Emiliano Dall’Anese and Sairaj V. Dhople and Georgios B. Giannakis},
  title = {{Optimal Dispatch of Photovoltaic Inverters in Residential Distribution Systems}},
  journal = {IEEE Transactions on Sustainable Energy},
  year = {2014},
  doi = {10.1109/tste.2013.2292828},
}
Figure