EAGER: Renewables: Collaborative Proposal on Stochastic Unit Commitment with Topology Control Recourse for Networks with High Penetration of Distributed Renewable Resources

由于风能资源的不确定性和可变性，将风能资源大规模整合到电力基础设施的发电组合中，给系统运营商带来了新的挑战。风能和其他可再生能源的间歇性，以及当前电力系统存储的局限性，对将可再生资源整合到电网中，同时保持可接受的服务可靠性提出了严峻的挑战。传统和灵活资源的有效部署需要明确考虑到前一天单位承诺的不确定性的新方法。本项目提出使用拓扑控制作为一种追索机制，通过输电线路切换调动电网的灵活性，根据当前的可再生能源发电条件重新定向潮流，从而克服因间歇性而导致的可靠性损失。然而，这种拓扑控制将需要新的算法创新，这将允许系统操作员通过将拓扑控制作为一种追索行动来计划他们的前一天单元承诺。该项目追求一种新的输电线路切换问题公式，以响应变量生成，有望在计算上可行。希望该项目的成果将有助于促进可再生能源发电的增长，同时保持电网的效率和可靠性。由于电网的拓扑结构是由一组离散的传输链路组成的，因此选择使用哪些链路就成了一个组合优化问题。此外，可再生能源发电存在间歇性，要求输电网络适应所观察到的具体发电情景。这导致了一个两阶段随机优化模型，其中第一阶段选择与慢斜坡发电机有关，而第二阶段模型包括快速斜坡发电机和间歇性发电机。这种设置中的求助行为导致第二阶段的混合整数规划（MIP），这违反了在电力系统运行中非常成功的常见分解算法的凸性要求。这一建议表明，一个新的模型，即两阶段随机MIP (SMIP)，具有非常特殊的结构，可以利用它来解决现实的随机单元承诺问题，即使已知的两阶段SMIP模型的一般类别是极其困难的。该提案概述了一种组合的并行-串行近似策略，该策略看起来很有希望，并且可以改变普遍持有的观念，即可再生能源的高度渗透将对可靠性产生不利影响。该项目是一个高风险、高回报的项目，在两个层面上：a)它可能导致一个真正可持续的方法来实现可靠的可再生能源集成，b)通过分解成更小的部分来解决非常大规模的SMIP的算法进步，而不会牺牲最优性，这将为解决这些非常具有挑战性的优化问题提供重要的一步。

The massive integration of wind resources into the generation mix of the electric power infrastructure poses new challenges to system operators due to the uncertainty and variability of these resources. The intermittent nature of wind and other renewable energy resources, together with limitations of storage in current power systems, poses serious challenges for integrating renewable resources into the power grid, while maintaining acceptable service reliability. Efficient deployment of conventional and flexible resources requires new methods that explicitly account for uncertainty in day-ahead unit commitment. This project proposes to use topology control as a recourse mechanism which mobilizes flexibility of the grid through transmission line switching to redirect power flow in response to prevailing renewable generation conditions, thus overcoming loss of reliability due to intermittency. However, such topology control will require new algorithmic innovations which will allow system operators to plan their day-ahead unit-commitment by incorporating topology control as a recourse action. The project pursues a novel problem formulation for transmission line switching in response to variable generation that holds promise for being computationally feasible. It is hoped that the outcomes of the project will help to facilitate growth of renewable generation while maintaining grid efficiency and reliability.Because the topology of the grid consists of a discrete set of transmission links, the choice of which links to use becomes a combinatorial optimization problem. Moreover, the presence of intermittency in renewable generation requires the transmission network to adapt to the specific generation scenario being observed. This leads to a two-stage stochastic optimization model in which the first stage choices are associated with slow-ramping generators, while the second stage model includes the fast-ramping as well as intermittent generators. The recourse action in this setup leads to a mixed-integer program (MIP) in the second stage, which violates the convexity requirements of common decomposition algorithms which have been highly successful in power system operations. This proposal shows that a new model, which is a two stage stochastic MIP (SMIP), possesses a very special structure which can be exploited so that realistic stochastic unit commitment problems can be solved, even though the general class of two-stage SMIP models are known to be extremely difficult. The proposal outlines a combined parallel-serial approximation strategy which appears promising, and could transform the commonly-held notion that high penetration of renewable energy will have an adverse effect on reliability. The project is a high-risk, high return undertaking on two levels: a) it could lead to a truly sustainable approach to reliable renewable integration, and b) the algorithmic advance of solving very large scale SMIP by decomposing into smaller pieces, without sacrificing optimality would provide a major step in the solution of these very challenging optimization problems.