AI-Driven Smart Heat Exchanger Network for Combined Cycle Efficiency Optimization Using Machine Learning and Reinforcement Learning for Adaptive Flow Control and Waste Heat Recovery in Gas-Steam Turbine Cycles

Krishna Chadha, Chaitanya Singla, Satyarth Pandey

Abstract: This study presents an AI-Driven Heat Exchanger Network (AI-HEN) for improving waste heat recovery and thermal efficiency in Brayton?Rankine combined-cycle power plants. The proposed framework combines an Artificial Neural Network (ANN) for predicting the overall heat transfer coefficient and outlet temperature with a Reinforcement Learning (RL) agent that adaptively adjusts coolant flow rates under varying operating conditions. The ANN was trained on physics-based simulated data and achieved test-set R2 values of 0.87 for heat transfer coefficient prediction and 0.97 for outlet temperature prediction. The RL controller uses these predictions to optimize heat recovery without repeatedly solving the complete thermodynamic model. Thermodynamic analyses indicate improvements in steam turbine inlet conditions and combined-cycle efficiency compared with conventional fixed-flow operation. The study demonstrates the feasibility of integrating lightweight AI models into heat exchanger networks and discusses limitations and directions for future experimental validation.

Keywords: AI-Driven Heat Exchanger Network, Combined Cycle Power Plant, Artificial Neural Network, Reinforcement Learning, Heat Recovery Steam Generator, Waste Heat Recovery, Intelligent Process Control, Thermal Efficiency Optimization