Prediction of coal wettability using machine learning for the application of CO2 sequestration

Carbon capture, utilization, and storage (CCUS) is an essential greenhouse gas-reducing technology that can be employed throughout the energy system. Carbon dioxide (CO2) sequestration in underground stratas is one of the effecient ways of reducing carbon emissions. CO2 sequestration in coal formations can be used to improve the methane recovery from coal formations (ECBM). The efficiency of this process highly depend on the wettability of the coal in contact with CO2. Different experimental methods including contact angle (CA) measurments can be used to estimate the wettability. However, the experimental techniques are expensive, incosistant, and timeconsuming. Therefore, this study introduces the application of artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) to estimate the CA in coal-water-CO2 system. ANN and ANFIS techniques were built using 250 point dataset to calculate the contact angle of coal formation. The input parameters were the coal properties, operating pressure, and temperature. 70% of the data set was used to train the model, while 30% of the data was used for the testing process. The models were then validated with a set of unseen data. The results showed that ANN and ANFIS models accurately predicted the contact angle in the coal-water-CO2 system as a function of coal properties and the operating conditions. The correlation coefficient (R) and the average absolute percent error (AAPE) between the actual and estimated contact angle were used as indicators for the model performance. ANN and ANFIS models predicted the contact angle with R values higher than 0.96 for the different datasets. AAPE was less than 7% in both models for the training and testing datasets. An empirical equation was built using the weight and biases from the developed ANN model. The new equation was validated with the unseen data set and the R-value was found to be higher than 0.96 with an AAPE less than 6%.these results confirm the reliability of the proposed models to get the contact angle in the coal formation without laboratory work or complex calculations. These models can be used to screen the coal formation targets for carbon storage.