A comparative study of machine learning frameworks for predicting CO2 conversion into light olefins

The use of fossil fuels, especially carbon dioxide (CO2) emissions, is commonly associated with global warming and recent climate change. In recent years, there has been an increasing focus on utilizing CO2 to produce different chemicals, prompted by rising apprehensions about global warming. This study primarily revolves around developing machine learning (ML) frameworks to convert CO2 into light olefins, using specific operational parameters. Through modeling process, a database including 180 data points was created using data from previous literature. To this aim, temperature and space velocity were considered as input parameters. Besides, CO2 conversion, CO selectivity, and light olefin selectivity were the outputs. Robust ML techniques were employed to predict the CO selectivity, CO2 conversion, and olefin selectivity with absolute error values of 1.56 %, 6.70 %, and 3.35 %, respectively. Specifically, when assessing CO selectivity using the Cascade Forward Neural Network (CFNN) model optimized with Levenberg-Marquardt (LM) algorithm, approximately 95.5 % of the data exhibited an absolute percent error below 5.2 %. Regarding CO2 conversion, over 47.7 % of the data exhibits an absolute percent error below 5.2 % for the Generalized Regression Neural Network (GRNN) model. For olefins selectivity, more than 80 % of the data demonstrates an absolute percent error less than 5 % using the CFNN-LM model. In addition, sensitivity analysis showed that temperature is the most significant factor for both CO selectivity and CO2 conversion. The accuracy and reliability of the developed models were verified using the Leverage approach. The main focus of this research is using the robust ML paradigms to concurrently estimate CO2 conversion, CO selectivity, and olefin selectivity. Furthermore, no previous research has utilized artificial neural networks (ANNs) as a valuable tool that yields significant outcomes in the direct conversion of CO2 to light olefins.