Under the coordinated promotion of "double carbon" strategy and digital economy, the supply chain is rapidly moving towards a new paradigm of sustainability and digital intelligence. Aiming at the pain points of traditional carbon financial tools optimization, such as ignoring dynamics, multi-agent game and the difference of digital intelligence level, this paper proposes a multi-objective optimization and decision-making algorithm framework for carbon financial tools oriented to digital intelligence supply chain. Firstly, a "dynamic multi-objective-real-time feedback" dual-loop model is constructed: the outer loop adaptively adjusts the economic, environmental and risk weights based on deep reinforcement learning (DRL) to solve the subjective defects of static weights; The inner loop adopts model predictive control (MPC) engine to realize minute-level rolling optimization to cope with carbon price fluctuation and demand disturbance. Secondly, the level of digital intelligence is included in the endogenous variable to quantify the amplification effect of data quality on the optimization effect. Thirdly, the coupling mechanism of cross-agent non-cooperative game and blockchain intelligent contract is designed to automatically verify carbon emission data and ensure incentive compatibility and information transparency of multiple parties in green credit, carbon credit and low-carbon securities allocation. The simulation experiment based on 30-minute scale and 30-day span shows that compared with the static weight and reactive algorithm, the comprehensive target value of this framework is reduced by 23.3% and 39.9% respectively, and the recovery time is shortened by more than 50% in the sudden interruption scenario. Moreover, the blockchain verification can completely suppress the false reporting behavior, and the actual carbon emissions are reduced by 15%. The research provides an extensible and grounded decision-making theory and technical support for the deep integration of carbon financial market and digital intelligence supply chain.
Chen et al. (Sun,) studied this question.