The transition to a circular bioeconomy in agriculture demands precise, real-time optimization of organic waste valorization, with anaerobic digestion (AD) being a central process. However, the inherent non-linearity, time-varying dynamics, and complex microbial interactions in large-scale agricultural AD reactors pose significant challenges to traditional kinetic models and human operators. This study proposes a novel data-driven hybrid CNN-LSTM-attention model to predict and optimize biogas yield and carbon-nitrogen (C/N) ratios using high-frequency multi-sensor data. By integrating real-time sensor feeds of pH, volatile fatty acids (VFAs), total solids (TS), and historical biogas producti... More >