Abdulkadir Tanimu

This work introduces a feed-aware modeling framework that integrates ensemble anomaly detection, PCA-based segmentation integrated with feed chemistry, and Bayesian-optimized regression to improve the model performance of a Fluid Catalytic Cracking (FCC) unit using pilot-plant experimental data. Three regression models were trained using 5-fold cross-validation and optimized using Bayesian Optimization: Gaussian Process Regression (GPR), Boosted Trees (BT), and Robust Linear Regression (RLR). Modeling on the full unsegmented data set showed that conversion was a highly predictable target. Both GPR and BT achieved strong performance (mean R² values of approximately 0.72 to 0.73 after optimization), while RLR remained unstable due to its linear assumptions. For the gasoline yield, BT with optimization performed best (mean R² of approximately 0.67), while GPR and RLR with optimization showed limited predictive skill (R² values of approximately 0.25 and -0.51, respectively). These results reflect the nonlinear and distributed nature of yield formation. PCA revealed a clear latent structure that aligned strongly with feedstock type, and K-means clustering confirmed this separation by distinguishing naphtha from VGO samples in a chemically meaningful way. Thus, the data set was segmented into feedstock groups, which yielded substantial improvements, especially for GPR. In the naphtha subset, GPR with optimization achieved mean R² values of 0.956 for conversion and 0.902 for gasoline yield, with very low associated error values. Performance also improved for VGO, although to a lesser extent, while RLR continued to underperform. Overall, segmentation based on PCA and clustering was essential for improving model reliability in FCC prediction tasks. Among all algorithms evaluated, GPR consistently delivered the highest accuracy and generalizability across feed types and targets, supporting its suitability for data-driven optimization of catalytic cracking performance.

Hydrogen (H₂) offers a less carbon-intensive energy production method than natural gas. The potential of utilizing hydrogen at a large scale within the future energy mix to fuel the world opens the door to investigating hydrogen production from heavy and extra-heavy oil reservoirs. Various reaction mechanisms are involved in the in situ combustion gasification of heavy oil to produce sustainable and low carbon intensive hydrogen. In-situ catalytic hydrogen production involves injecting a catalyst into the reservoir or utilizing the in situ reservoir materials to catalyze the various reactions. Clay minerals and formation water were found to serve as in situ catalytic materials and enhance the in situ hydrogen production process. This work presents a comparative review of the catalytic and noncatalytic experimental studies carried out on in situ hydrogen production. The formation damage induced by the in situ combustion and its effect on hydrogen production was highlighted. In addition, the impact of the formation damage induced by the in situ combustion on the hydrogen production process is discussed. This study categorized the experimental studies conducted on the hydrogen production from heavy oil into catalytic and noncatalytic processes to highlight the effect of the synthetic and natural reservoir catalytic materials on in situ hydrogen production.

The prediction of the yields of light olefins in the direct conversion of crude oil to chemicals requires the development of a robust model that represents the crude-to-chemical conversion processes. This study utilizes artificial intelligence (AI) and machine learning algorithms to develop single and ensemble learning models that predict the yields of ethylene and propylene. Four single-model AI techniques and four ensemble paradigms were developed using experimental data derived from the catalytic cracking experiments of various crude oil fractions in the advanced catalyst evaluation reactor unit. The temperature, feed type, feed conversion, total gas, dry gas, and coke were used as independent variables. Correlation matrix analyses were conducted to filter the input combinations into three different classes (M1, M2, and M3) based on the relationship between dependent and independent variables, and three performance metrics comprising the coefficient of determination (R²), Pearson correlation coefficient (PCC), and mean square error (MSE) were used to evaluate the prediction performance of the developed models in both calibration and validations stages. All four single models have very low R² and PCC values (as low as 0.07) and very high MSE values (up to 4.92 wt %) for M1 and M2 in both calibration and validation phases. However, the ensemble ML models show R² and PCC values of 0.99-1 and an MSE value of 0.01 wt % for M1, M2, and M3 input combinations. Therefore, ensemble paradigms improve the performance accuracy of single models by up to 58 and 62% in the calibration and validation phases, respectively. The ensemble paradigms predict with high accuracy the yield of ethylene and propylene in the catalytic cracking of crude oil and its fractions.