Robust observer and long short-term memory-based maximum H₂/CH₄ value predictive intelligent tracking control for...

Two-stage anaerobic digestion (TSAD) is a promising technology for organic waste-to-energy conversion, yet optimizing biogas yield remains challenging due to process nonlinearity and unmeasurable states. To address this challenge, this paper proposes a robust observer and long short-term memory (LSTM) network-based maximum H₂/CH₄ value predictive intelligent tracking controller. The proposed dual-driven framework integrates mechanistic modeling with process data, utilizing a state observer to estimate unmeasurable states, thereby augmenting LSTM input features. Learning from historical data, the LSTM predicts future trajectories of maximum biogas yield. A time-delay estimation-based intelligent proportional-derivative input constraint control algorithm, combined with an auxiliary system, is developed to track these trajectories while ensuring input feasibility. Simulation results demonstrate an average 15% enhancement in biogas production compared to extremum-seeking control methods, validating the effectiveness of the proposed strategy.