In a new development doing the rounds, the carbon capture right from the power plants can very well be climate-friendly and also consume less energy, all thanks to AI modeling, as per a study from the University of Surrey.
Apparently, the researchers there tweaked a capture system that was based on a real coal-fired power station. Their objective was to self-optimize the CO2 capture procedure within a renewable energy system by way of an enhanced weathering of calcite along with fresh water within a packed bubble column- PBC reactor.
By way of that process, CO2 gets captured by bubbling of the flue gas through fresh water that has limestone in the reactor, thereby converting the CO2 into bicarbonate as well as making sure to store it in the ocean.
However, since all this takes energy to pump the water as well as the CO2, the capture system happened to have its own wind turbine, but when the wind did not blow, the energy was taken from the grid. By way of using AI, researchers went on to teach a model system to anticipate what would happen so it could pump less water when there happened to be less CO2 to capture or when there was less renewable energy which was available.
Especially, two deep learning models happened to be considered to capture the dynamics of the PBC reactor: one, a long short-term memory network- LSTM and the other, a two-stage multilayer perceptron network- MLP. Data-driven models, which happened to be based on LSTM, were developed to anticipate wind energy- R2: 0.908 as well as the inlet flue gas CO2 concentration- R2: 0.981 by way of using publicly available datasets.
Interestingly, a multi-objective NSGA-II genetic algorithm happened to get applied that made use of the inlet flue gas CO2 concentration as well as wind energy predictions in order to pre-emptively self-optimize reactor process conditions, i.e., the superficial liquid flow rate as well as the superficial gas flow rate, in order to maximize the CO2 capture rate and at the same time minimize the consumption of non-renewable energy.
Researchers went on to remark that the model could go on to capture 16.7% more carbon dioxide in a one-month operation by way of using 36.3% less energy from the National Grid.
