【作者】 王硕；

【导师】 江晓东；

【作者基本信息】 天津大学 ， 电力系统及其自动化， 2018， 博士

【摘要】 随着智能电网及新能源技术的发展,电力系统运行优化需要兼顾经济性、可靠性等多个相互矛盾的优化目标,多目标问题模型应运而生。针对电力系统多目标优化的高维非线性特点,有效利用运行人员的偏好信息,能够提高多目标优化模型的求解效率,得到满足实际需求的解。为此,本文提出基于用户偏好的多目标优化模型,并从理论、解法、具体算法及应用四个方面开展了研究,主要贡献如下:(1)理论层面:提出了基于用户偏好的多目标优化问题模型,使得各目标函数同时实现一定程度的最优;基于非线性动力系统理论,建立了该模型的偏好可行解与动力系统稳定平衡状态的对应关系;基于理论研究,提出了计算偏好可行解的UPE方法。(2)方法层面:提出了基于共识智能优化算法的三阶段求解方法,提高智能优化方法对Pareto前沿的计算效率。针对电力系统问题的复杂非线性特点,提出了UPE协同智能优化算法的三阶段求解方法,进一步改善了智能优化算法的寻优能力。此外,为满足电力系统在线优化控制需求,本文还提出了一种迭代UPE方法,实现了在有限计算时间内,对一个偏好Pareto最优解的快速精确计算。(3)算法层面:将三阶段求解方法与电力系统具体问题的领域知识相结合,设计高效的计算算法,实现了快速计算满足用户偏好的Pareto最优解(集)。(4)应用层面:将提出的模型与方法分别应用于配电系统、输电系统运行中的多目标优化问题,并针对问题特点设计具体求解算法,包括:配电系统供电恢复问题:提出了考虑恢复供电量及开关操作次数的多目标供电恢复模型。在IEEE标准系统及实际三相不平衡系统测试算例中,求得了多种最优的供电恢复方案,验证了三阶段方法在计算多个偏好Pareto最优解方面有明显优势。考虑风电接入的日前有功备用协调优化问题:以运行成本最低与可靠性水平最高为优化目标,确定了含高比例风电的系统最优旋转备用水平,得到了不同可靠性水平的多种备用配置方案,并保证24小时内系统可靠性水平满足用户偏好。考虑风电接入的日内有功调度优化问题:提出了以运行成本、环境成本与风电消纳比例作为优化目标的有功超前调度(look-ahead power dispatch)模型,求解满足网络传输能力及用户偏好的准确调度方案。该方案可以消纳给定不确定集合中的风电出力偏差,维持调度解的可行性及最优性。更多还原

【Abstract】 With the development of smart grid and increasing penetration of renewable energy sources(RESs),power systems operation involves multiple incompatible objectives,such as economy,reliability and so on.Hence multi-objective optimization(MOO)model is becoming more and more important.Since the high dimension and strong nonlinearity of MOO problems in power systems,the operators’ preference information can significantly improve the solving efficiency,and determines the final solution.To this end,this thesis proposes a user-preference based multi-objective problem formulation,and analyzes the MOO problem in 4 aspects: theoretical analysis,solution methodology,computational algorithms and applications in power systems.The main contributions of the thesis is as follows:(1)Theoretical analysis: a user-preference based multi-objective optimization model is built to achieve different degree of optimality for different objectives.Based on the theoty of nonlinear dynamical systems,a theoretical relationship between the multi-objective optimization problems and a class of non-hyperbolic dynamical systems is built,which leads to the theoretical characterization of the user-preferred feasible region.Based on the theoretical results,a user preference enabling method is proposed to calculate the user preferred-feasible solution of MOO problems.(2)Solution methodology: a three-stage optimization method is proposed based on the consensus-based intelligent optimization algorithms,which improves the search efficiency for Pareto front.Regarding the highly nonlinearility of the power system problem,a UPE guided intelligent optimization method is proposed.This hybrid method improves the search capability of the existing methods.Moreover,in order to meet the requirement of online control and optimization,an iterative UPE method is proposed to calculate an accurate user-preferred Pareto-optimal solution in limited time period.(3)Computational algorithms: highly-efficient computational algorithms are designed combining the three-stage methodology with domain knowledgy of specific applications,which can fast calculate the Pareto-optimal solutions satisfying user preference.(4)Applications: the proposed multi-objective optimization model and methodology are applied into transmission and distribution system operation.Detailed computation algorithms are degined according to the problem features.Three MOO problems in power systems are studied,including:1)Service restoration problem in distribution network: a multi-objective model considering the restored energy amount and switch operation number is proposed.The three-stage method is evaluated on both the IEEE 30-bus test system and a practical unbalanced three-phase distribution system.The case study shows that the proposed method can find diversified restoration schemes with different user preferences.2)Spinning reserve optimization problem with renewable uncertainties: the multiobjective optimization model is established to determine the optimal reserve amount,which minimizes generation cost,reserve cost and maximizes reliability level.The three-stage methodology shows its advantange in generating multiple reserve allocation schemes with different reliability level.3)Look-ahead power dispatch problem with renewable uncertainties: a multiobjective look-ahead power dispatch model is established considering generation cost,emission amount and wind power utilization ration as objectives.The obtained dispatch scheme satisfies network transfer capability and user’s preference over different objectives.It can also accommodate the uncertain set of renewables and maintain a feasible and optimal operation point.更多还原