Managing commodity and energy conversion assets is difficult because it involves managing operational activities in the face of notoriously volatile commodity and energy prices. These activities include the production, processing, refining, transportation, storage, distribution, and physical trading of commodities and energy. The management of these activities is naturally modeled as the exercise of complex real options on commodity and energy prices. This gives rise to computationally intractable stochastic dynamic programming models. The goal of this award is to develop and analyze novel real options models of fundamental commodity and energy conversion assets, and to design effective and efficient novel approximate dynamic programming methods to compute near optimal operating policies and optimality bounds on their performance. This research will leverage in novel ways operations management models and analyses of operating policies, emerging operations research approaches to approximate dynamic programming based on math programming and Monte Carlo simulation techniques, and financial engineering models of the stochastic evolution of commodity and energy prices.
If successful, this research will generate quantitative models and methods to support managers in maximizing the market value of commodity and energy infrastructure. This has the potential to benefit society at large because commodities and energy are used in virtually every manufacturing and service process, and commodity and energy conversion assets are typically managed heuristically in practice. This research will also benefit operations management, operations research, and financial engineering education by generating novel material for potential adoption in current and/or future courses on real options, providing an operations perspective on real options.
, such as power plants or natural gas storage facilities, as real options gives rise to intractable stochastic and dynamic optimization models. These models thus must be solved heuristically. This research has developed a new approach to obtain near optimal heuristic policies and bound on their estimated market values. This technique is based on the known approximate linear programming approach, and consolidates various existing approximate dynamic programming methods, such as smoothed approximate linear programming, iterated approximate linear programming, and least-squares Monte Carlo. Applications, among others, include the management of natural gas storage assets, both in isolation and in a network that also includes transport assets (such as pipelines). This work has also generated a simpler approach to the estimation of upper bounds on the market value of commodity storage assets, as well as improved versions of the least-squares Monte Carlo approach and of a decision-rule approach commonly used in practice to manage such assets. Other results involve the characterization of the optimal structure of operating policies in wind energy generation, electricity storage, and power procurement, as well as the analysis of the impact of futures-term structure model error on the real options management of commodity storage. These findings have relevance for the development of heuristic operating policies, potentially based on approximate dynamic programming. This work has also introduced the concept of merchant operations: The management of commodity and energy conversion assets as real options. This research thus advances the state of the art in operations management, operations research, and financial engineering, also providing efficient and effective methods to practitioners involved in the management of commodity and energy infrastructure.
