(398an) Evaluation of the Efficiency in a Set of Air Separation Units through Data Envelopment Analysis and Malmquist Productivity Index

Nowadays, the efficient use of energy resources is one of the main principles for industry, and it is one of the more complex challenges faced by companies. The goal treat to solve the trouble between raising living standards by continuing to expand industrial production and lessening the negative environmental impacts of industrial manufacturing activities. One of the approaches to address this challenge is to improve the energy efficiency in the core of industry’s production processes, which accounts for around one-third of the world’s final energy demand. Furthermore, the use of energy by industries is often inefficient, leaving much room for improvement. For this reason, improving industrial energy efficiency will be a critical success factor for companies in order to improve cost competitiveness, brakes emissions of carbon dioxide and saves money.

Cryogenic air separation process is an old energy-intensive technology aimed to produce large amounts of technical gases with high purity standards. These pure gases can be obtained in Air Separation Units (ASUs) by cooling air until it liquefies and distilling it by means of a physical process. In the distillation, the air components (i.e., nitrogen, oxygen and argon) can be separated and recovered since they have different boiling temperatures. This process requires a very tight integration of heat exchangers and separation columns to guarantee good performance, and it uses large electricity amounts (i.e., several tens of megawatts) to cover the high compression and liquefaction needs.

In this context, we need to apply decision-making tools that allow assessing and quantifying the level of efficiency of the ASUs. In this work we propose a method to determine the efficiency in a set of real ASUs based on the non-parametric technique of Data Envelopment Analysis (DEA). The mathematical methodology of DEA has been used to assess the performance of many types of entities in many different contexts (i.e., from production and business firms, to non-profit agencies such as hospitals, universities, armies, countries, etc.) and it uses linear programming (LP) to objectively evaluate the relative efficiency of a set of units under multiple criteria conditions. In DEA, the relative efficiency of a DMU is defined as the maximum ratio of weighted sum of outputs to the weighted sum of inputs, being the efficiency less or equal than 1. The linear optimization is applied for each single unit, also called Decision Making Unit (DMU), and it provides the performance scores representing the degree of efficiency of each evaluated entity. In this work, the capabilities of DEA have been demonstrated by creating an input-oriented variable return to scale (VRS) model which allow identifying qualitatively (i.e., sources) and quantitatively (i.e., amounts) the inefficiencies in each input and output for every entity. For this, DEA creates an "efficiency frontier" formed by the efficient entities (i.e., those with efficiency scores equal than one) which serves to identify benchmarks against which such inefficient unit can be compared and, also, to define improvement targets to make efficient the inefficient units if achieved. Furthermore, the efficient units are also compared between them by solving the super-efficiency model which ranks the efficient ASUs and lets to observe differences among them.

This work is complemented by measuring the productivity change over time in each ASU by means of the Malmquist Productivity Index (MPI). This approach identifies the ratio between efficiency, as calculated by the DEA, for the same production unit in two different time periods (i.e., years) to identify efficiency gains and losses in each ASU over time. The MPI can be decomposed into two components measuring the change in the technology frontier and in the technical efficiency. In this paper, both components are calculated to obtain more information from the individual Malmquist components to identify patterns of efficiency change.

The methodology of this paper is demonstrated by applying it in a set of 33 existing Air Separation Units from a gas company operating plants around the world. The MPI is illustrated to the set of ASUs during the last four years. The results provide insightful information into how the unit’s performance has evolved over time, as well as how to improve the units to make the air separation technology more efficient.