In face of climate change and the need to implement environmentally friendly industrial processes, Carbon Capture and Storage (CCS), has been gaining increasing importance. Particularly in the Oil and Gas industry, CO2 removal or absorption through amines plays a prominent role in combustible gases as it is an efficient solution for the treatment of diluted gas streams and for its attractive cost /benefit ratio. Despite its popularity, the process of removal of CO2 by amines presents several technical challenges, among which stands out the balance between temperature, pressure and amine concentration in order to obtain the maximum absorption efficiency in the process. On the other hand, the Internet of Things era enables oil and gas companies to manage and store operational data in real time and, in this perspective, operational efficiency improvement aided by technology has become a mandatory management line. In this context, the brand new Deep Learning algorithms allows to integrate and to analyse high volumes of data to find patterns that can be used for decision making models providing costs reduction, processes optimizing and overall performance improvement. The present work explores the application of the Long Short-Term Memory (LSTM) Neural Network methodology for CO2 concentrations forecasting. The data used was provided by an Oil and Gas company operating in Brazil and was extracted from the regular operations of one of their amine plants. The developed algorithm is able to forecast the CO2 concentrations of the treated gas in an interval of time ranging from 20 minutes to 3 hours.