Auction design for the allocation of carbon emission allowances to supply chains via multi-agent-based model and Q-learning (Q2140815)

Different techniques are known to analyze the behavior of bidders in the carbon auction market, and the most famous of them are optimization models, experimental methods and simulation methods. In solving this extremely important task, the authors impose reasonable constraints on the bid price and volume, with a view to controlling the price and reducing inefficiencies in the carbon auction market. With the constraints thus imposed, the paper presents a successful attempt to compare uniform and discriminatory pricing in the carbon auction with and without the constraints, using a multi-agent-based model consisting of the government and supply chains. The authors' research can be summarized as follows: without the constraints, the carbon price under uniform pricing is less than or equal to the discriminatory pricing method; there are no differences between them in the case of constraints. In general, the auction reduces the profit of supply chains. This negative effect is less uniform than discriminatory pricing in the case without the restrictions. Nevertheless, strong supply chains make huge profits from the auction when mitigation rates are high. An interesting Q-learning and CPLEX solving algorithm is presented to analyze and develop bidding strategies, determine the equilibrium price of the carbon trading market, and the corresponding optimal supply chain solution. The reader can find a relatively complete bibliography on this interesting subject.