One of the main causes of climate change on our planet is an excess of carbon dioxide emissions. It occurs mainly in electricity production and industrial processes, including the manufacture of steel and cement. Currently, engineers and chemists are looking for ways to capture carbon that could kidnap and store carbon dioxide, preventing him from being released in the atmosphere.
The concept of carbon capture is an effective way to reduce greenhouse gas emissions. Special carbon capture factories operate on the basis of amine technology, using chemical compounds that can dissolve carbon dioxide. Amines are also used in many industries, such as pharmaceutical products, the production of epoxy resins and dyes.
The problem is that amines can be potentially harmful to the environment as well as human health, it is therefore essential to alleviate their impact. When you use amines in carbon capture factories, emissions must also be checked, which is a challenge due to the technological difference between plants.
A team of scientists from the EPFL Basic Sciences School and the Research Center for Carbon Solutions of Heriot-Watt University has developed A new solution based on automatic learning To predict amine programs from carbon capture plants. The solution was tested experimentally in a real plant in Germany led by a team with Professor Berend Smit of the School of Basic Sciences of the EPFL and Professor Susana Garcia of the Research Center for Carbon Solutions of the Heriot-Watt University in Scotland as research chiefs.
The experiments were carried out in the largest coal -in -law center in Germany, where the new generation amine solution has been tested in a pilot factory for more than a year. However, it has been found that amines can be released with combustion gases. This presented a problem because amine emissions must be checked.
Professor Susana Garcia, with the owner of the plant, the RWE and the TNO in the Netherlands, developed a stress test to study amine emissions in various process conditions. Professor Garcia described that they had designed an experimental campaign to understand how and when amine shows would be generated. But some of the experiences also required the intervention of factory operators to ensure its safe operation.
These interventions have led to the question of how to interpret the data. Are Amine's frees the result of stress tests themselves, or did the operator interference indirectly affect emissions? It was still complicated by a general misunderstanding of scientific mechanisms underlying the versions of Amine.
“In short, we have had an expensive and successful campaign that has shown that Amine emissions can be a problem, but no tools to analyze the data more,” reports Smit.
It really looked like an insoluble problem. All measures have been taken every five minutes and many data has been collected. And then Kevin Maik Jablonka decided to use automatic learning for the recognition of models to predict future Amine emissions according to plant data. Using a new amine emission prediction model, the researchers were able to separate the emissions caused by the operator's intervention from emissions caused by the stress test.
The model has shown that certain interventions had the opposite effect on the emissions of solvent components. Thus, the attenuation strategies required for capture installations operating on a single solvent (for example, monoethanolamine) must be examined if they are operated using a mixture of amines.
“I am very enthusiastic about the idea of the potential impact of this work; it is a whole new way of looking at a complex chemical process,” said Smit. “This type of forecast is not something that can be done with conventional approaches, so it can change the way we operate chemical factories.”
