Power Generation Assets

Abstract

The overall topic of this thesis is the valuation of power generation assets under energy and risk constraints. Our focus is on the modeling aspect i.e. to find the right balance between accuracy and computational feasibility. We define a new not yet investigated unit commitment problem that introduces an energy constraint to a thermal power plant. We define a continuous stochastic dynamic program with a nested mixed integer program (MIP). We introduce a fast implementation approach by replacing the MIP with an efficient matrix calculation and use principal component analysis to reduce the number of risk factors. We also provide a fast heuristic valuation approach for comparison. As both models can only provide lower bounds of the asset value, we investigate the theory of upper bounds for a proper validation of our power plant results. We review the primal dual algorithm for swing options by Meinshausen and Hambly and in particular clarify their notation and implementation. Then we provide an extension for swing options with multiple exercises at the same stage that we developed together with Prof. Bender, University of Braunschweig. We outline Prof. Bender’s proof and describe the implementation in detail. Finally we provide a risk analysis for our thermal power plant. In particular we investigate strategies to reduce spot price risk to which power plants are significantly exposed. First, we focus on the measurement of spot price risk and propose three appropriate risk figures (Forward delta as opposed to Futures delta, synthetic spot delta and Earnings-at-Risk) and illustrate their application using a business case. Second we suggest risk mitigation strategies for both periods, before and in delivery. The latter tries to alter the dispatch policy i.e. pick less risky hours and accept a (desirably only slightly) smaller return. We introduce a benchmark that weighs risk versus return and that we will call EaR-efficient option value. We propose a mitigation strategy for this benchmark that is based on quantile regression. It defines a price interval for executing an individual swing right and is therefore very well suited for real world applications. In case of an American option we are able to show EaR-efficiency of our strategy in particular for a changing risk profile of the underlying price (altering volatility). Finally, we investigate hedging strategies before the delivery period as a function of the maximum available energy. In particular, we look at a hedge for the spot price risk of the power plant using a swing option. We propose a heuristic based on our synthetic spot deltas to find the relevant parameters of the swing option (number of swing rights and swing size) for a given upper generation amount.

75 Figures and Tables

Cite this paper

@inproceedings{Enge2010PowerGA, title={Power Generation Assets}, author={T Gabriel Enge and Christoph Markschies and K. Helmes and Andreas Brandt}, year={2010} }