Differentiating Through Power Flow Solutions for Admittance and Topology Control

Samuel Talkington, Daniel Turizo, Sergio A. Dorado-Rojas, Rahul K. Gupta, Daniel K. Molzahn · 2025

Read Differentiating Through Power Flow Solutions for Admittance and Topology Control by Samuel Talkington, Daniel Turizo, Sergio A. Dorado-Rojas, Rahul K. Gupta, Daniel K. Molzahn (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-17071.txt.

Introduction

Samuel Talkington and four co-authors work out how power grids react to small network changes in this 2025 paper. Grid math links the push of power at each node to node voltage size and angle. Planners need to know how those values shift when lines change. Direct re-solving for each change is slow and hides clear links.

The authors use a classic calculus rule to get direct shift formulas. The formulas cover voltage, line flow, and line current shifts. Mild terms ensure the shifts exist for both spoke and mesh grids. One use predicts node voltage after lines switch, with no fresh solve. Another use tunes grid traits to host more local power sources.

Catalog record

CREATORSamuel Talkington, Daniel Turizo, Sergio A. Dorado-Rojas, Rahul K. Gupta, Daniel K. Molzahn
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-17071

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Differentiating Through Power Flow Solutions for Admittance and Topology Control Samuel Talkington, Daniel Turizo, Sergio A. Dorado-Rojas, Rahul K. Gupta, Daniel K. Molzahn Abstract The power flow equations relate bus voltage phasors to power injections via the network admittance matrix. These equations are central to the key operational and protection functions of power systems (e.g., optimal power flow scheduling and control, state estimation, protection, and fault location, among others). As control, optimization, and estimation of network admittance parameters are central to multiple avenues of research in electric power systems, we propose a linearization of power flow solutions obtained by implicitly differentiating them with respect to the network admittance parameters. This is achieved by utilizing the implicit function theorem, in which we show that such a differentiation is guaranteed to exist under mild conditions and is applicable to generic power systems (radial or meshed). The proposed theory is applied to derive sensitivities of complex voltages, line currents, and power flows. The developed theory of linearizing the power flow equations around changes in the complex network admittance parameters has numerous applications. We demonstrate several of these applications, such as predicting the nodal voltages when the network topology changes without solving the power flow equations. We showcase the application for continuous admittance control, which is used to increase the hosting capacity of a given distribution network. I Introduction The power flow equations are widely used for modeling the relationship between the voltage phasors and the power injections in power systems, and are fundamental for key operational and protection functions of power systems. These are used for a number of applications such as state estimation [ 1 ] , scheduling and control based on the optimal power flow (OPF) [ 2 , 3 ] , fault location (e.g., [ 4 ] ), and protection (e.g., [ 5 ] ), among others. The power flow equations are inherently nonlinear, posing significant computational challenges. This is poignantly true in combinatorial optimization problems, where discrete variables compound with the non-convex AC OPF problem, which can result in an optimization problem that becomes intractable as the size of the network grows. A class of such problems that have attracted considerable interest in the power system literature—and network science more broadly—is termed as topology control problems. Examples of problems in this class include line switching problems for power shut-offs [ 6 ] , loss reduction [ 7 , 8 , 9 ] , distribution network reconfiguration [ 10 , 11 , 12 ] , congestion management via bus splitting [ 13 , 14 , 15 ] , etc. Moreover, a conceptually similar topic—flexible AC transmission systems (FACTS)—has been of interest at the transmission scale for a significant period where continuous control of shunt reactors is exercised [ 16 , 17 , 18 ] . Existing literature (e.g., [ 19 , 20 , 21 ] ) suggests the use of series line reactors for active power flow control in power systems. Recent work has developed continuous models [ 22 , 15 ] . These approaches, belonging to the continuous admittance framework, are inclusive of models for transmission line switches, node-breaker substations, and other technologies such as smart wires and unified power flow controllers (UPFCs). This framework has seen application in combinatorial switching problems [ 22 ] and current congestion management [ 15 ] . Overall, both the topology control and the continuous admittance problems in power systems are nonlinear and non-convex, often resulting in intractable formulations. Related work: In the existing literature, different techniques have been investigated to tackle topology control problems, such as convex relaxation techniques (e.g., [ 23 , 24 ] ), linearized approximations (e.g., [ 25 , 26 , 27 ] ), machine learning techniques [ 28 , 29 ] , and other heuristics [ 30 , 31 ] . These methods are valuable because of their computational efficiency—in contrast, direct methods for topology control problems typically comprise mixed-integer non-linear programs, which are NP-hard in general and potentially pathological. Several simplification techniques are based on distribution factors [ 25 , 31 , 32 , 33 ] , which represent the sensitivity of the power system state to changes in the decision variables. In a typical OPF problem, the power system state refers to the bus voltage phasors, and the decision variables are typically the generators’ set-points (or power injections in general). For such applications, the voltage sensitivity coefficients are derived as in [ 34 ] . In other cases, this theory also has been applied to obtain power losses and current-flow sensitivity factors, which are used to minimize losses and tackle line-current congestion problems [ 35 , 36 , 37 ] . Linearization techniques have also been applied to line-switching and topology control problems, such as power transfer distribution factors (PTDFs), line outage distribution factors (LODFs) and outage transfer distribution factors (OTDFs) that are widely used in transmission system applications [ 31 , 32 , 33 ] . PTDFs are used for the expressing incremental changes in real power flow that occur on transmission lines due to real power transfers between two regions; these factors are frequently used for pre-contingency analysis by the system operators and can be used to optimize the generator setpoints with the objective to minimize congestion. On the other hand, LODFs are used to determine the impacts of line outages on power flows. Both of these distribution factors often use the DC power flow approximation which is reasonable only for transmission systems. Recent work in AC line outage distribution factors represents line flow changes with respect to outages while incorporating the full AC power flow equations. The authors of [ 38 ] developed a method based on the holomorphic embedding load flow (HELF) method. The authors of [ 32 ] extended the traditional line outage distribution factor computation method to handle an arbitrary number of discrete line outages. The focus of previous work in the literature has been on the impact of changes in power flows due to incremental changes in the power injections. Previous and recent studies have extended these ideas to understanding the sensitivity of power flow solutions to changes in line admittances [ 25 , 30 , 33 ] , due to their potential for use in various applications in topology control and continuous admittance control. However, the current literature has not yet achieved a formal and complete treatment of this topic, which motivates our own. Our paper presents a general framework for distribution factors that express the sensitivities of the power system state to changes in the network admittance parameters. These coefficients are computed by implicit differentiation of the power flow equations with respect to the network admittance parameters. The proposed coefficients can be used to produce linearized optimization problems for a number of applications in power transmission and distribution networks alike, including topology control for congestion management and continuous admittance control for voltage regulation. Contributions: This work presents a generalized technique to analytically differentiate power flow solutions with respect to network admittance parameters. Earlier investigations of this idea, such as the work of [ 25 , 30 , 31 ] , show that this technique enables several admittance control applications to be made more efficient. Our work contributes

Read the complete text