Mitigating furnace pressure fluctuations under rapid load ramping using a wavelet-LSTM-PPO based intelligent control framework

Rapid load ramping in coal-fired power plants with high renewable energy integration often induces severe furnace pressure fluctuations, threatening combustion stability and operational safety. To address this challenge, we propose a predictive and adaptive control framework that integrates wavelet transform, long short-term memory (LSTM) neural networks, and proximal policy optimization (PPO) reinforcement learning. Wavelet-based multi-resolution decomposition is employed to extract key features from pressure signals, while an LSTM model forecasts short-term pressure dynamics. Based on predictive feedback, a PPO agent learns an optimal control strategy to regulate secondary air and fuel inputs in real time. Validation on a 600 MW supercritical boiler unit demonstrates a 42.2% reduction in the standard deviation of furnace pressure fluctuations, improved stability under variable load conditions, and smoother actuator response compared with conventional control schemes. These results highlight the potential of combining deep learning and reinforcement learning techniques to enhance combustion stability and support secure, flexible operation of coal-fired power plants under high renewable energy penetration.