One of the main causes of climate change on our planet is excess carbon dioxide emissions. It mainly occurs in the production of electricity and industrial processes, including steel and cement production. Currently, engineers and chemists are looking for ways to capture carbon that could sequate and store carbon dioxide, thereby preventing it from freeing it into the atmosphere.
The concept of carbon capture is an effective way to reduce greenhouse gas emissions. Special carbon capture plants operate on the basis of amine technology, using amines – chemical compounds that can dissolve carbon dioxide. Amins are also used in many industries, such as pharmaceuticals, production of epoxy resins and dyes.
The problem is that amines can be potentially harmful to the environment, as well as human health, which is why it is very important to alleviate their influence. When using amines in carbon capture plants, emissions should also be controlled, which is a challenge due to the technological difference between plants.
A team of EPFL School of Basic Sciences and the Carbon Solutions of Heriot-Watt University research center has developed New solution based on machine learning To predict amine emissions from carbon capture plants. The solution has been experimentally tested at a real factory in Germany led by the team with Professor Berend Smit from the School of Basic Sciences and Professor Susana Garcia from the Carbon of Heriot-Watt University Solutions Research Center in Scotland as heads of research.
Experiments were carried out at the largest coal power plant in Germany, where the new generation amina solution has been tested at the pilot power plant for over a year. However, it was found that the amines can be released with internal combustion gas. This was a problem because amina emissions must be controlled.
Professor Susana Garcia, together with the owner of the plant, RWE and TNO in the Netherlands, developed a test of extreme conditions in order to examine amine emissions in various procedural conditions. Professor Garcia described that they designed an experimental campaign to understand how and when the amine emissions will be generated. But some experiments also required the intervention of installation operators to ensure its safe operation.
These interventions led to the question of how to interpret data. Are the release of the amina to the same tests of extreme conditions or the disturbance of the operator indirectly influenced the emissions? This was additionally complicated by the general misunderstanding of scientific mechanisms underlying Amina publications.
“In short, we had an expensive and successful campaign that showed that amina emissions can be a problem, but without tools for further data analysis,” says Smit.
It really looked like an unsolvable problem. All measurements were performed every five minutes and a lot of data was collected. And then Kevin Maik Jablonka decided to use machine learning to recognize patterns to predict future amina emissions based on plants. Using the new amina emission forecasting model, scientists were able to separate the emissions caused by the intervention of the operator from the emission caused by the test of extreme conditions.
The model showed that some interventions had the opposite impact on the emission of solvent components. Therefore, relief strategies required to capture objects operating on one solvent component (e.g. monoetanoloamine) should be viewed if they are operated using the Amin mixture.
“I am very enthusiastic about the potential impact of this work; this is a completely new way to look at the complex chemical process,” says Smit. “This type of forecasting is not something that can be done with any of the conventional approaches, so it can change the method of handling chemical plants.”
