Abstract
This study aims to enhance the operational efficiency of NAWEC thermal power stations through a comprehensive approach involving stochastic modeling, analysis, and forecasting of energy generation (EG) and fuel consumption (FC). The historical power plant monitoring data underwent rigorous examination using both the bivariate autoregressive moving average model BARMA and artificial neural networks (ANNs). The investigation revealed that KT1ENG8 demonstrated exceptional performance in univariate analysis, as evidenced by the autoregressive moving average model, which provides minimized Akaike information criterion (AIC) and Kulback information criterion (KIC) values. Consequently, the effectiveness of the bivariate ARMA (p,q) model extends the top prediction of fuel consumption and energy generation from thermal power plants, providing valuable insights for thermal health and maintenance scheduling. To help the reader grasp the inherent dynamic of each of these strategies and their performances, a comparative synopsis of the outcomes is provided. The ANN results further enrich the study by presenting R2 values for different engines, offering ranked perspectives on their respective predictive capabilities. BK1ENG1 and KT1ENG8 have higher r2 values of 99.6% and 97.3%, respectively, than do the other power plant engines.