基于数据驱动分布鲁棒优化的梯级水光蓄联合优化调度_张帅.pdf

新型电力系统DOI:10.15961/j.jsuese.202200921基于数据驱动分布鲁棒优化的梯级水光蓄联合优化调度张帅1，王子涵1，张蜀程1，胡俊刚1，罗颖1，刘俊勇2(1.国网成都供电公司，四川 成都 610041；2.四川大学 电气工程学院，四川 成都 610065)摘要：多种可再生能源互补联合发电技术因其独特优越性正在成为“双碳”背景下电力系统优选供电方案之一，而其不确定性复杂耦合特性下的互补联合调度问题越来越受到人们关注。针对不确定性优化调度问题，本文引入能较好平衡不确定性及鲁棒性的数据驱动分布鲁棒优化理论（data-driven DRO），提出了一种新的基于数据驱动 DRO梯级水光蓄联合发电系统协同优化调度方法。首先，考虑系统互补经济调度成本建立两阶段调度模型，制定各电站日前出力调度计划；然后，引入综合范数约束限定概率置信区间，并考虑最恶劣分布下的实时运行调整成本，获取日前调度计划的最优调整方案，日调度计划和调度调整方案形成最优调度计划；最后，本方法采用MP-SP框架，引入CCG算法展开两阶段协同求解。为验证所提方法的性能，引入四川示范区实际运行数据，开展了有效性验证、性能对比分析、计算效率仿真验证等。结果表明：本调度方法的有效性在数据规模、置信度水平两个维度得到了验证；对于SO、RO及本方法鲁棒性及经济性等性能指标的对比，本方法可获得高于SO的鲁棒性及高于RO的经济性；将本调度方法与概率性时序生产模拟方法的计算耗时进行对比，该方法实现了相同计算精度的较高计算效率。基于两阶段调度模型及循环迭代求解的DRO梯级水光蓄联合优化调度方法实现了协同调度结果经济性与保守性的均衡，其高效性能得到验证，为多种可再生能源互补协同调度提供了新思路。关键词：梯级水光蓄；联合发电调度；水光互补；数据驱动分布鲁棒优化；CCG算法中图分类号：TM7文献标志码：A文章编号：2096-3246（2023）02-0128-13Data-driven Distributionally Robust Optimization Based Coordinated Dispatching forCascaded Hydro-PV-PSH Combined SystemZHANG Shuai1，WANG Zihan1，ZHANG Shucheng1，HU Jungang1，LUO Ying1，LIU Junyong2(1.State Grid Chengdu Power Supply Co.,Chengdu 610041,China;2.College of Electrical Eng.,Sichuan Univ.,Chengdu 610065,China)Abstract:Due to its unique advantages,the multiple renewable energy complementary combined power generation technology is becoming oneof the preferred power supply scheme under the“carbon peaking and carbon neutrality”context.The optimal dispatching of this complementarycoordinated generation system considering its complex uncertain coupling characteristics has attracted more and more attention.For uncertain op-timal dispatching problems,this paper introduced the data-driven distributionally robust optimization(DRO)theory,which can better balance theuncertainty and the robustness of the problem.A new coordinated optimal dispatching method for the cascaded hydro-PV-pumped storage com-bined system was further proposed based on the data-driven DRO theory.This method established a two-stage DRO dispatch model to formulatethe daily dispatch schedule considering the complementary economic dispatch cost of the system firstly.The comprehensive norm constraint wasintroduced to limit the probability confidence interval.Considering the adjustment cost of the real-time operation under the worst distribution,theoptimal dispatch schedule was formed by the optimized adjustment scheme for the day-ahead dispatch schedule,the daily dispatch schedule andthe adjustive dispatch scheme.The two-stage dispatch model was solved by the CCG algorithm according to the MP-SP framework finally.In or-der to verify the performance of the proposed method,the actual operation data of the demonstration area in Sichuan was taken to carry out thevalidity verification,the performance comparative analysis and the simulation verification of computational efficiency.The results show that,thevalidity of the proposed scheduling method is verified in the data scale and the confidence level dimensions.For the robustness and economy收稿日期:2022 08 31作者简介:张帅（1987），男，工程师，博士.研究方向：可再生能源互补联合发电；主动配电网.E-mail：网络出版时间:2023 03 10 14:03:51 网络出版地址:https:/ http:/http:/ 第 55 卷 第 2 期工 程 科 学 与 技 术Vol.55 No.22023 年 3 月ADVANCED ENGINEERING SCIENCESMar.2023comparison of the SO,the RO and the proposed method,the method proposed in this paper can achieve higher robustness than SO and higher eco-nomy than RO.Comparing the calculation time of this scheduling method with the probabilistic sequential production simulation method,thismethod achieves a higher computational efficiency with the same computational accuracy.Based on the two-stage dispatching model and the iter-ative calculation,the data-driven DRO coordinated dispatching method for the cascaded hydro-PV-pumped storage combined system can achievethe balance between the economy and robustness of the dispatching results.Its efficient performance has been verified and a new way is obtainedfor the complementary coordinated dispatching of multi renewables.Key words:cascaded hydro-PV-PSH system;coordinated dispatch;hydro-PV complementation;data-driven distributionally robust optimization(DRO);CCG algorithm 风、光、水等可再生能源存在随机性、间歇性、波动性等出力特征，大规模风电、光伏并入电网将给电力系统带来调节能力骤降、抗干扰能力缺失和连锁故障风险增大等诸多问题，这已成为电力系统安全稳定持续运行的巨大挑战1。因此，为了适应可再生能源渗透率不断扩大的态势，弱化其大规模入网带来的冲击，对促进可再生能源综合利用的多能互补发电协同优化技术开展研究与探索，具有重大的现实工程应用价值。在此背景下，作为应对大规模可再生能源并网影响的主要措施，多能互补发电技术越来越受到研究者的重视，提出“多能互补协同优化的可再生能源综合开发”的战略任务势在必行。目前，多种可再生能源互补联合发电优化调度方面的研究仍主要以技术性目标（可靠性等）、经济性目标（运行成本等）、环保性目标（减排最大）三大类作为目标函数展开探讨。Ming2、An3等以青海龙羊峡互补电站为例，对水光互补发电计划及光伏出力波动平抑等问题开展研究。Kougias等4对水光互补性能进行了量化分析，其研究成果对多电站互补联合调度具有一定指导意义。Wang5、Wang6等则对多能源发电系统的互补运行原理及协调运行模式与策略进行研究与探讨，以实现新能源最大化消纳及促进绿色减排目标。以上研究对可再生能源联合优化调度研究具有一定借鉴意义，但是，对多可再生能源系统的多维不确定特性及耦合特性聚焦不足，当前诸多研究未能满足可再生能源不确定性联合优化调度的实际应用需求。对于多维不确定特性的风、光、水等可再生能源互补联合调度通常采用随机优化（stochastic optimiza-tion，SO）7或鲁棒优化（robust optimization，RO）8及机会约束方法910等方法处理。SO根据随机变量概率分布抽取场景，并将概率分布离散化处理，生成大量离散样本，将各离散场景作为确定性优化问题分别求解11。SO可用于不确定场景下可再生能源发电调度过程的定量分析，但多维不确定因素概率分布较为复杂，难以精确刻画其变化规律12，通常需要预先设定概率分布类型13，这将导致SO技术可靠性降低。SO方法是基于巨量离散场景开展的，巨量场景会导致求解规模过大而使得求解时间较长、计算效率降低。RO方法无需预先设定随机变量概