Machine Learning Powers the Clean Energy Economy

Scientists at Los Alamos National Laboratory are developing powerful machine learning models (applications of artificial intelligence) to simulate underground hydrogen storage operations under various cushion gas scenarios, which will play a key role in the future low-carbon economy.

“One of the most practical ways to store hydrogen is in deep saline aquifers, or depleted hydrocarbon reservoirs,” says Mohammed Mehana, the team's lead scientist. “But to do this, we first need to inject a cushion gas into the reservoir, which displaces the existing fluid and provides pressure support for hydrogen recovery.”

Scientists have studied the effects of cushion gases (usually methane, carbon dioxide and nitrogen) on such underground hydrogen storage systems, but until now, it has not been fully understood how cushion gases affect the operational performance of underground hydrogen storage.

In a new paper published in the International Journal of Hydrogen Energy, the Los Alamos team was able to explore a comprehensive range of cushion gas scenarios, providing important insights into the impact of different cushion gases on underground hydrogen storage performance.

Complex solutions

Expanding the hydrogen economy is a key part of the country's efforts to decarbonize. Like gasoline, hydrogen gas needs to be produced and stored locally to power heavy-duty clean-energy trucks, generate direct electricity, and provide durability for solar farms during winter months.

To reach this scale, the country would need to tap a wide range of underground reservoirs. Previous studies have focused on a single geological and operational condition. But to mimic a real-world scenario, the Los Alamos team's model considered multiple geological conditions, the presence of water, and the operational effects of multiple cushion gases.

“Underground hydrogen storage is complex due to hydrogen's unique properties and complex operating conditions,” said Xiaowen Mao, a postdoctoral researcher on the Los Alamos team. “We need to maximize hydrogen recovery and purity during the extraction stage while mitigating the risk of water production. Understanding these and other factors is essential to making underground hydrogen storage economically viable.”

To achieve this, the Los Alamos team used deep neural network machine learning models to analyze a combination of geological and operational parameters to mimic the variability of real-world scenarios. In the paper, some of the team's key findings include:

• Technical feasibility of underground hydrogen storage in porous rocks with improved storage performance over the cycle
• The advantages and disadvantages of underground hydrogen storage in saline aquifers and depleted hydrocarbon reservoirs, and
• The impact of different cushion gas scenarios on hydrogen recoverability, purity, water production risk, and well injectivity into porous rocks.

Years of research

The paper builds on years of hydrogen storage research at Los Alamos, one of the first institutions to study the technology from multiple angles.

Los Alamos scientists have investigated hydrogen flow and transport behavior in a subsurface environment, which will help understand the impact of cushion gas on underground hydrogen storage performance.

Another phase of the study, all underway, explores potential hydrogen storage sites in the Intermountain West region, an effort that combines the physics of underground geological structures with machine learning-enabled simulations.

Yet another research division has been working on developing tools that can assess the reliability, risk, and performance of hydrogen storage under different conditions. This latter work has resulted in OPERATE-H2, an industry-first software that integrates advanced machine learning for optimizing hydrogen storage.

paper: “Effect of cushion gas on hydrogen storage in porous rocks: Insights from reservoir simulation and deep learning.” International Journal of Hydrogen Energy. DOI: 10.1016/j.ijhydene.2024.04.288

Funding: Los Alamos National Laboratory Technology Evaluation and Demonstration Fund and the Laboratory Directed Research and Development Program.