[Paper Review] A Geometric Analysis of Power System Loadability Regions

Yang Weng, Ram Rajagopal+1 more|arXiv (Cornell University)|Jun 22, 2017

Power System Optimization and Stability20 references3 citations

TL;DR

This paper proposes a geometric framework for power system loadability analysis using rectangular voltage coordinates to model active and reactive power flows. It formulates linear programming (LP) methods to verify boundary points, compute loadability margins, and locate boundary points efficiently—offering faster, more scalable results than traditional nonlinear or eigenvalue-based approaches.

ABSTRACT

Understanding the feasible power flow region is of central importance to power system analysis. In this paper, we propose a geometric view of the power system loadability problem. By using rectangular coordinates for complex voltages, we provide an integrated geometric understanding of active and reactive power flow equations on loadability boundaries. Based on such an understanding, we develop a linear programming framework to 1) verify if an operating point is on the loadability boundary, 2) compute the margin of an operating point to the loadability boundary, and 3) calculate a loadability boundary point of any direction. The proposed method is computationally more efficient than existing methods since it does not require solving nonlinear optimization problems or calculating the eigenvalues of the power flow Jacobian. Standard IEEE test cases demonstrate the capability of the new method compared to the current state-of-the-art methods.

Motivation & Objective

• Address the lack of geometric intuition for loadability in systems larger than two buses, especially when including reactive power.
• Overcome computational limitations of existing methods that rely on nonlinear optimization or eigenvalue analysis of the power flow Jacobian.
• Enable directional exploration of the loadability boundary, moving beyond uniform load increases across all buses.
• Develop a computationally efficient method to verify if an operating point lies on the loadability boundary, compute its margin, and locate boundary points in any direction.

Proposed method

• Represent complex voltages in rectangular coordinates to linearize the power flow equations and Jacobian matrix.
• Formulate a linear programming (LP) problem to verify if an operating point lies on the loadability boundary using the condition that no feasible direction exists to increase active power without violating constraints.
• Use LP to compute the loadability margin by finding the maximum scaling factor of active power injections that keeps the point within the feasible region.
• Define a directional search via LP to locate the boundary point in any specified direction by solving for the intersection of the Pareto-front with a ray from the operating point.
• Leverage the convexity of the feasible region and the linearity of the Jacobian in rectangular coordinates to ensure computational efficiency.
• Validate the method on IEEE test systems (14, 300, 13659-bus transmission and 8, 123-bus distribution networks) using CVX in MATLAB.

Experimental results

Research questions

• RQ1 How can the geometric intuition of the PV curve in a two-bus system be extended to larger systems with both active and reactive power flows?
• RQ2 Can a linear programming formulation replace nonlinear optimization or eigenvalue computation for loadability boundary verification and margin calculation?
• RQ3 To what extent does the proposed method improve computational scalability compared to existing methods for large power systems?
• RQ4 How does the inclusion of reactive power limits affect the loadability margin, and can the method detect binding constraints?
• RQ5 Can the method efficiently locate boundary points in arbitrary directions, enabling comprehensive boundary exploration?

Key findings

• The proposed LP-based method verifies loadability boundary membership in under 10 seconds for large systems (e.g., 13,659-bus) on a standard laptop, demonstrating high computational efficiency.
• Loadability margins computed via the proposed method are significantly larger than those from Thevenin-equivalent methods—e.g., 8.6 vs. 1.1 margin in the IEEE 118-bus system—indicating better robustness assessment.
• Reactive power limits were found to be binding: adding a -50 to 50 MVar limit at bus 69 reduced the margin from 8.6 to 6.1, confirming the method’s sensitivity to operational constraints.
• The method successfully located boundary points matching the true Pareto-front in a 3-bus network, validating accuracy against exhaustive computation.
• Computational time scales well with system size, with margin computation taking less than 10 seconds for systems with thousands of buses.
• The method outperforms eigenvalue-based approaches, as the latter are necessary but insufficient for boundary detection—highlighting the need for a more robust geometric criterion.