Optimized Neural Network Architectures for Accurate Forecasting of Thermal Power Plant Output

Neural networks are widely recognized for their ability to process vast datasets and perform complex computations, making them particularly valuable for modeling combined cycle power plants. These facilities involve intricate, non-linear interactions among variables such as temperature, pressure, fuel consumption, and turbine efficiency, which challenge traditional analytical methods. This paper presents a novel artificial neural network topology designed to enhance the modeling of combined-cycle power plants by improving predictive accuracy and computational efficiency. The proposed architecture more effectively captures complex interdependencies, enabling better performance monitoring, predictive maintenance, and operational optimization. This approach contributes to more reliable, efficient, and sustainable power generation by anticipating potential issues, enhancing fuel efficiency, and minimizing downtime. Unlike conventional, artificial neural network models, this study introduces an optimized design that reduces computational complexity while improving accuracy.