Analyzing Hydrogen Production and Distribution Infrastructures Using Agent-Based Modeling and Simulation

Transitioning to a hydrogen-based energy infrastructure will require making critical decisions in a complex global environment that includes macroeconomic forces, energy markets, regulations, environmental considerations, and countless decision-making units (players) in many industries. Complicating this picture is that all of the factors and players involved will continuously interact with and exert influence on one another, causing each to adapt and exhibit new behaviors.

The immense complexity and variability of such a market system makes it virtually impossible to accurately predict outcomes by using conventional types of market models, which typically employ standard game theory techniques, such as repeated auctions, fluctuating supply and demand, profit and loss, etc. The number of variables and possible outcomes for a complex adaptive system (CAS) like the hydrogen economy is too large for models that rely only on such predetermined, overarching rules. This is where agent-based modeling and simulation (ABMS) enters the picture.

Argonne National Laboratory, with its world-class expertise in the study of complex, interdependent systems, was chosen by the U.S. Department of Energy to use ABMS to analyze the development of our nation's hydrogen infrastructure system. Key goals include:

Determining how the hydrogen CAS might progress
Understanding what the intermediate transition stages might be
Identifying factors that could potentially destabilize infrastructure development, and actions or events that could aid system development
Determining how far the system could swing in one direction or another before it corrects itself or possibly collapses

Hydrogen Infrastructure Analysis

ABMS consists of a set of self-directed agents (each representing a player) that have specific traits. These agents operate within a framework that simulates the players' decisions and interactions and facilitates studying the macroscale consequences of these interactions. The ability to deal with the nonlinear nature of the hydrogen market evolution, where a small change in initial conditions or an accumulation of small effects will produce dramatically different end results, helps set ABMS apart from other modeling approaches.

The results of Argonne's ABMS analysis will be used to help identify what actions might affect the rate and stability of the hydrogen market development, and how or when technological or cost-reducing innovations might aid or detract from market growth.

January 26, 2005

ABMS Agents

ABMS agents perform diverse tasks using their own internal decision rules. A special feature of these agents is that they can learn about and adapt to the market and actions of other agents by using "learning" that is either observation-based or exploration-based. The first type of learning uses a structured process in which each agent can:

Look Back – evaluate past performance
Look Ahead – project the future state of the market
Look Sideways – determine what other agents have done

As a result of these observations, each agent can choose to

Mmaintain its current strategy,
Adjust its current strategy, or
Switch to a new strategy.

The second type of learning involves the agent exploring various market strategies and then observing the results of each. After finding a strategy that performs well, the agent uses and fine-tunes the strategy until dramatic market changes occur and the strategy begins to fail. Periodically, or in the event of failure, agents will explore and evaluate new strategies to see if one can be found that works better.

Privacy & Security Notice | Contact Us | Site Index