Deep Real Options: Valuation of Real Options on Green Energy using Deep Learning Methods

Abstract

This paper investigates using deep neural networks to value real options in green energy context. We look at testing the capability of neural networks to value American options first and then apply them to real options. Two projects are compared, where one incorporates stochastic variable cost, a gas powered plant, and the other with fixed costs, a wind powered plant. The Least Squared Monte Carlo valuation method for American options was used as a benchmark. A simple fully connected network was tested on benchmark simulated data and showed promising results. In essence, the model learns and then predicts continuation values in discrete time intervals and compares them to the option payoff. The model however did not extend training accuracy when dealing with testing data of real options. We conclude that neural networks have potential in valuing real options though may suffer from over- fitting to its trained data.

Date of Award	Aug 2023
Original language	American English
Supervisor	Yerkin Kitapbayev (Supervisor)