PredBlock: A physics-informed AI and blockchain framework for integrity verification in CO2CCS pipelines optimization

초록

The scaling of Carbon Capture and Storage (CCS) infrastructure is critical for global decarbonization, yet the safety, economic viability, and energy efficiency of carbon dioxide (CO2) pipeline transport remain constrained by the lack of optimized control and transparent monitoring mechanisms. Traditional supervisory systems often fail to optimize compression energy in real-time and lack the immutable audit trails required for rigorous Monitoring, Reporting, and Verification (MRV). To address these challenges, this study introduces PredBlock, unified framework integrating physics-informed Artificial Intelligence (AI), Reinforcement Learning (RL) for energy-aware flow control, and a decentralized blockchain architecture. The system employs a random forest (RF) ensemble enriched with thermodynamic state equations to predict risk indicators, such as leakage, corrosion, and overpressure, arising from impurities like H2S, SO2, O2, and H2O. Simultaneously, a Q-learning agent dynamically regulates flow to minimize power consumption. To ensure regulatory compliance, a custom PureChain Proof-of-Authority and Association (PoA2) blockchain layer guarantees zero-cost, immutable logging of sensor telemetry and model predictions. Evaluated entirely on physics-informed synthetic data, the framework achieves a macro-averaged F1-score of 0.9989 +/- 0.0009 across five random seeds, with AI-inference latency of 33.9 ms and blockchain write latency of similar to 1953 ms - well within the 60-second sensor sampling interval. The RL controller demonstrates proof-of-concept energy savings of up to 4.39% over conventional Proportional-Integral-Derivative (PID) control in the best-converging seed; the mean energy saving across all five seeds is-1.85% +/- 3.70%, indicating that reliable convergence requires extended training (>= 2000 episodes) or coarser state discretization, with consistent convergence identified as a direction for future work. By bridging predictive safety intelligence, energy optimization, and decentralized trust, this work establishes a simulation-validated digital paradigm for autonomous, efficient, and auditable CCS energy infrastructure.

키워드