Geometric critical point analysis of lossless power system models

Abstract

Electric utility analysts today face an increasingly difficult task of formulating both long and short term operating plans which will provide at the same time efficient and economical operation while delivering reliable and uninterrupted service to electricity users. One of the key ingredients in this planning is a set of large scale simulations of the steady-state network performance under various anticipated operating conditions. Central to these analyses are the classical "load flow" equations which are the equilibrium equations for the "swing equations" which are a physically based model of the dynamic operation of an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</tex> -node power system. Despite the long standing and widespread use of these equations, there remain a number of very basic open questions: What are the number and nature of the equilibria of the swing equations? How many stable equilibrium operating points are there in an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</tex> -node electric power grid? In this paper some powerful analytical tools from topology and geometry are used to answer certain of these questions. It is well documented that the load flow equations comprise a formidable large scale system but what is interesting, and perhaps surprising, is that even for a small number of buses, these equations possess a rather rich and intricate qualitative behavior which has heretofor been only partially understood. Indeed, until now there was no complete statement in the literature concerning the number of load flows in a general three-bus network. In Theorem 2.7, we state that for the "generic" three-bus network there are, for sufficiently small power injections, either four or six real load flows and that, in either case, exactly one of those load flows is stable. This is a special case of the results we derive for a general <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</tex> -bus powergrid. Our method consists in first transforming the load flow equations for a lossless electric power network by trigonometric substitutions into algebraic equations. This makes it possible to apply some deep and powerful results from algebraic geometry and intersection theory to study these equations. An obvious tool for determining the number of solutions is provided by the classical theorem of Bezout, but it is shown that for systems describing an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</tex> -machine network with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n \geq 4</tex> , this result cannot be directly applied because the solutions contain solution components of positive dimension "at infinity." A major result in this paper is a modified Bezout technique which allows us to compute the number of complex (and a fortiori an upper bound on the number of real) solutions to the load flow equations. Combining this with the classical Morse inequalities, we obtain very explicit results regarding the number of stable load flows for a given network topology and set of power injections. The cases of three and four machine networks are considered in detail.