Data-driven financial transmission right scenario generation and speculation

Abstract

This paper proposes a data-driven framework to solve the financial transmission right (FTR) portfolio construction problem from the perspective of a speculator. FTR speculation is modeled as a stochastic programming problem in which uncertainty comes from the price spread across different pricing nodes over a certain holding period. Since it is difficult to model and forecast the joint distribution of prices for typical electricity markets with thousands of pricing nodes, k-means clustering with network congestion patterns is first used to help focus on important nodes and reduce the problem size. Then, a quantile regression (QR)-based method is proposed to predict the conditional distribution of average nodal prices. A Gaussian copula is further used to construct the joint conditional distribution of average nodal prices. The proposed method is tested on real market data obtained from the southwest power pool (SPP). The results show that the method has a steady performance in both node selection and price scenario generation and outperforms state-of-art methods, including copula-GARCH and truncated skew-t distributions.