Mean-Variance Optimization of Power Generation Portfolios Under Uncertainty in the Merit Order
University of Duisburg-Essen
Chair for Management Sciences and Energy Economics
October 1, 2011
In this article we discuss welfare‐optimal capacity allocation of different electricity generation technologies available for serving system demand. While the classical peak load pricing theory derives the efficient portfolio structure from a deterministic marginal production cost curve ("merit order"), we investigate in particular the implications of possible reversals in the merit order — sometimes also referred to as fuel switch risks — induced by uncertain operating costs. We propose a static, non‐convex optimization model combining the classic peak load pricing model with elements of mean‐variance portfolio (MVP) theory and analytically discuss possible solution cases and important optimality properties. We examine the approach in a case study on the efficient structure of generation portfolios consisting of CCGT and hard coal technologies in Germany.
With special emphasis, we study the emergence of overcapacities (exceeding maximal demand) in efficient portfolios and show that diversification is not beneficial per‐se. The results show that the efficient technology mix may be significantly impacted by a risk for reversals in the merit order. Therefore, our findings support the importance of considering this risk factor especially with long‐term investment horizons.
The model is applicable to various investment problems related to production of nonstorable goods under price uncertainty of input factors. Similar problems can e.g. be found in transportation systems or in the process industry.
Number of Pages in PDF File: 50
Keywords: power plant investments, peak load pricing, mean‐variance portfolio theory, fuel mix diversification
JEL Classification: G11, L94, Q43, C44working papers series
Date posted: October 12, 2011
© 2013 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo7 in 0.391 seconds