National security evaluation and early warning system of petroleum resources in China

Amidst the constraints of carbon peaking and carbon neutrality goals (dual carbon) and high import dependence, safeguarding China's petroleum security faces intensified challenges. This study constructs a dynamic weight evaluation framework integrating four dimensions: "Availability", "Affordability", "Reliability", and "Sustainability". A multi-model earlywarning mechanism—combining AHP-based variable weights with eight algorithms (benchmark models: ARIMA, ETS; machine learning: BP neural network, SVM, RF; deep learning: CNN-LSTM, Transformer, N-BEATS)—is further developed. Key findings reveal: (1) Dynamic weighting identifies Availability and Reliability as core dimensions, with geopolitical risks and substitution dependence dominating security outcomes. Weight fluctuations (e.g., reserve-production ratio shifts of 6–7% driven by technology/resource cycles) demonstrate systemic adaptability. (2) The historical assessment (2000–2023) reveals that security levels progressed from an unsafe level (2.7 in 2000) to a moderately safe level (4.4 in 2023), exhibiting a pattern of fragile recovery. Periods of instability (2000–2013, 2021–2022) were associated with supply-demand imbalances, price volatility, and geopolitical shocks (e.g., Middle East turbulence, the COVID-19 pandemic). (3) CNN-LSTM projections (2025–2050) indicate security indices fluctuating between 4.4 and 4.6 under baseline scenarios, while Sustainability emerges as the primary risk source. Without policy intervention, indices under the low scenario (4.6 by 2050) will linger near the moderately safe threshold, facing significant downside risks. (4) Core vulnerabilities include high import dependence, reserve gaps, transport bottlenecks, price volatility, and delayed low-carbon substitution. Consequently, a tiered response policy system—activated by quantified security thresholds—is proposed to enhance petroleum industry resilience and advance dual carbon objectives. • Proposing a dynamic 4A framework for China's petroleum security assessment. • Developing a multi-model early-warning system with superior CNN-LSTM performance. • Revealing a fragile recovery pattern from unsafe to moderately safe. • Identifying sustainability as the future key risk and proposing tiered policies.