Energy Risk Management: Stochastic Optimization for Industrial Gas Operations

Abstract

This thesis explores the various elements of uncertainty that businesses face every day when managing their operations. As part of a partnership with Princeton and Air Liquide, a worldwide leader in industrial gases, this thesis incorporates stochastic optimization across a complex network of a modeled section of Air Liquide's business, including production facilities and customers. By analyzing the major areas of opportunities that are present, this thesis presents a model of the operational decisions made by Air Liquide.

Air Liquide's operations involve purchasing large quantities of natural gas and electricity, which introduces significant uncertainty into the cost of production of industrial gas because of the stochasticity of their prices. This paper comprehensively models the network of resources, as well as the current practices that are already in place in order to mathematically analyze and potentially improve the optimal operational policies. While most modeling of a network may involve deterministic linear programming, the introduction of storage capabilities within the company's operations significantly change the feasibility of modeling the system in a deterministic way. This thesis investigates strategies for structuring optimal decision-making policies using physical hedging (in the form of a hydrogen cavern for storage) to reduce volatilities introduced by customer demand and energy price fluctuations.

First, we will focus on the plant operations that run on natural gas, comparing different classes of policies in order to optimize the production of industrial gases. Second, we will look at the importance of the electricity market on Air Liquide's operations, more specifically the market heat rate. This thesis will study the relationship of wind and heat rate in order to understand the impact that wind has on Air Liquide's network, as well as the long-term implications of increased renewable energies being implemented into the electricity grid.