基于多主体博弈的跨流域供水水库群联合调度研究

Inter-basin water transfer is an important measure to solve the natural uneven regional distribution of water resources，which with reasonable water demand scheme from water users can improve the water supply efficiency both of the inter-basin water transfer project and the receiving reservoirs. Water price is taken as the basic market signal of water resources allocation to quantifies the economic benefits and costs of each agent in the water transfer system, so that the decision-making mechanism of how to determine the water demand from receiving reservoirs to apply for water transfer in the preparation of the operation plan of the water transfer system is revealed. Based on this, the pricing strategy of the water in different seasons from the manager of the water transfer project (water sources reservoir and transfer channels) is put forward in order to guide the receiving reservoirs to divert and storage enough water in advance. Then the optimization model of inter-basin water transfer project operation based on multi-agent game with the manager of the whole system as the only upper decision-maker while receiving reservoirs as the multi following decision-maker to simulate the influence of bi-level hierarchical relationship and competitive game behavior among multi-agent on operation decision-making. Furthermore, an intelligent algorithm based on Nash game and reinforcement learning is proposed to solve the high-dimensional multi-agent problem so the reasonable water transfer rule and the water supply rule of the receiving reservoirs are extracted. The Inter-basin Water Transfer Project in North Region of Hubei Province is taken as a case study to prove the validity and rationality of the model which the research results can provide theoretical basis and technical support for the operation of the inter-basin water transfer system.

跨流域调水是解决水资源区域分布不均、供需不平衡的重要途径，合理制定受水区引水计划能够充分发挥受水水库群的调蓄作用，提高调水工程的供水效益。项目以鄂北地区水资源配置工程为研究对象，利用水价作为水资源市场供需的基本信号和水资源配置的重要手段，量化调水系统中各利益主体的经济效益与成本，揭示受水水库在调度计划编制中如何确定引水需求的决策机制，提出调水工程管理者调控丰枯分季水价的策略以引导受水水库群合理提前引水充蓄。并基于调水定价策略和受水水库群的引水规则，建立两方决策者动态博弈的一主多从决策模型，探讨多利益主体间的层次结构关系和竞争博弈行为对引水决策的影响。并针对这一高维多主体决策问题，提出基于Nash博弈和强化学习的智能算法以提炼合理的引水规则和供水规则。研究成果可为跨流域供水水库群调度计划的编制提供理论依据与技术支撑。

跨流域调水是解决水资源区域分布不均、供需不平衡的重要途径，其调度规则的合理性直接决定了调水工程供水效益的发挥，而现有的调度建模理论对多主体互馈作用和结构关系考虑得不充分，研究基于多主体博弈的水库联合调度问题具有重要理论意义与实践价值。.（1）针对跨流域调水系统水库调度问题，将调水工程管理者和受水水库视作博弈的多决策主体，考虑受水水库的经济效益与引水成本，建立跨流域供水水库联合调度的两阶段决策模型。根据最优决策原理，以KKT条件对受水水库的两阶段决策模型进行求解，解析引水成本与调度策略间的关系，分析了引水成本对受水水库引水规则的影响机制。.（2）考虑水库供水系统中多主体博弈关系，基于二层规划理论将水库不同调度目标职能部门视作博弈的多决策主体，考虑不同决策者之间的领导服从关系，建立了上层以供水量最大为目标、以供水规模和供水调度线为决策变量，下层以发电量最大为目标、以发电调度线为决策变量的基于多目标层次关系水库调度模型。实例研究结果表明，与传统多目标模型相比，二层规划模型供水保证率更优，但增大了弃水概率，减少了发电效益，反映了供水和发电两个决策者间的博弈约束对综合调度效益的影响。.（3）针对跨流域供水水库群调度高维度、非线性、强约束的特性，将合作型协同进化思想与粒子群算法相结合，提出一种基于种群停滞搜索技术的协同进化粒子群算法。实例研究结果表明，相对于传统粒子群算法，本算法计算结果在累积缺水总量和供水保证率等指标上均更优，且收敛速度更快，并在一定程度上避免了陷入局部最优。