An analysis of electricity congestion price patterns in North America

Highlights

• Electricity congestion price data allows detecting transmission congestion patterns.

• Principal component analysis (PCA) is the main statistical tool applied.

• Outputs of the PCA convey information that is easily interpretable.

• PCA scores exhibit seasonality, spikes and auto-correlation.

• A time series model for scores based on their stylized facts is proposed.

Abstract

The present paper illustrates the use of principal component analysis (PCA) on the congestion component of local (i.e. zonal) electricity price data to detect the most salient congestion patterns in electricity transmission grids managed by either a Regional Transmission Organization (RTO) or an Independent System Operator (ISO). Outputs from the PCA along with some data visualization tools are shown to make the identification of such patterns seamless and straightforward. An empirical analysis is conducted for three North American power systems, namely NYISO, ISO New England and PJM. Finally, a simple time series model representing the evolution of PCA scores is proposed.

Introduction

The present paper aims to study the dynamics of differences between electricity prices at multiple nodes of a power transmission grid. Indeed, at a given point in time, power prices differ over the various locations due to limitations in transmission capacity which can generate transmission bottlenecks. When binding, such constraints force the dispatching of more expensive power generation units located in areas in which the capacity to import power is curtailed. This phenomenon, referred to as congestion, leads to additional charges for the consumption of power in the concerned areas, and thus to more expensive local electricity prices.

In centralized markets of North America,¹ the operation of the electricity flow is conducted by an independent not-for-profit entity referred to as either a Regional Transmission Organization (RTO) or an Independent System Operator (ISO). The dispatching of power generating units is determined by the latter entities through an optimization process attempting to make power prices as low as possible and reduce congestion charges while ensuring the integrity of the power system. The electricity price at a given time period (e.g. an hour) and any given node of the electricity grid stemming from such an optimization is typically determined through a so-called locational marginal approach which is meant to reflect the marginal cost of the consumption of an additional megawatt hour (MWh) of electricity at the given node for the period. When reported by RTOs or ISOs, nodal prices, referred to as locational marginal prices (LMP) or analogous names depending on the market, are typically decomposed into three components: an energy cost component which has the same value for all nodes of the grid, a congestion component reflecting price disparities due to bottlenecks, and an energy loss component meant to reflect thermal losses occurring during the transmission process.

The presence of the congestion and loss components are the cause of price differentials (called spreads) among the various nodes of a grid. Electricity locational spreads directly impact revenues of many market participants such as power generators, electricity retailers and power merchants; understanding the dynamics of the congestion and loss components of power prices which drive such spreads is therefore of paramount practical importance. Furthermore, in practice, the congestion component is much more impactful than the loss component to explain price disparities, see for instance Hadsell and Shawky (2006) who discuss this point for the particular case of the New York ISO. As such, the current paper focuses exclusively on the dynamics of the congestion component. Another good reason to study the congestion component in isolation is that the payoff of financial derivatives allowing to hedge locational spreads (e.g. Financial Transmission Rights (FTRs) in the New England ISO, the Midcontinent Independent System Operator (MISO) and the PJM RTO, or Transmission Congestion Contracts (TCC) in the New York ISO) is often tied to the difference between the congestion component for two nodes rather than to the total LMP differential. A sound valuation of electricity transmission financial derivatives therefore should rely on a proper understanding of the congestion component dynamics.

The current study mainly lies within the realm of data visualization rather than full fledged stochastic modeling. Its main contribution consists in proposing an approach relying on simple and well-known statistical methods and data visualization tools to perform a quick and seamless identification of the most salient congestion patterns within a given electricity grid. More precisely, this study relies on the celebrated principal component analysis (PCA) method described in more details subsequently, which allows expressing the (potentially large) set of prices across all considered nodes as a simple and well-interpretable vector of a lower dimension. The screening of congestion patterns is performed by looking exclusively at the history of the congestion component of electricity prices at various locations on the grid. Our approach can be applied to obtain a quick grasp of the power flow dynamics on the grid through an explanatory analysis before running a potential full-blown stochastic analysis based on the development of sophisticated statistical models. Nevertheless, the current work can also prove useful to market participants wishing to develop a more in-depth and technical analysis of congestion dynamics; indeed, outputs of the current analysis (i.e. principal component scores time series) can serve as a first building block for more advanced models of multivariate congestion prices. This study takes a step in that direction by proposing a simple seasonal stochastic model based on regime-switching dynamics for the representation of scores time series. The current work could be of interest to multiple market stakeholders such as regulators, ISOs and RTOs, power traders, generators, retailers, congestion derivatives traders, and transmission infrastructure owners and builders.

The paper is subdivided as follows. Section 2 presents a quick survey of the literature on electricity price modeling. The PCA theoretical framework and its application to the analysis of electricity congestion prices is discussed in Section 3. Section 4 illustrates empirical results of the application of a PCA on hourly zonal day-ahead congestion prices for the NYISO power market, while analogous analyses for ISO New England and PJM are presented in Appendix A and Appendix B respectively. In Section 5, the dynamics of PCA scores for NYISO are analyzed and an associated time series model is proposed for its representation. Section 6 concludes.