【作者】 翟海保；

【导师】 程浩忠；

【作者基本信息】 上海交通大学 ， 电力系统及其自动化， 2007， 博士

【摘要】 电网灵活规划包括两个方面的内容:1)建立合理的规划模型;2)针对规划模型给出高效的求解算法。本文从算法和模型两方面研究电网灵活规划。针对大规模电网灵活规划计算难以全局收敛的难点,提出采用蚁群算法及其改进算法进行求解;通过对规划方案经济性、可靠性和灵活性的综合考虑,提出计及电能资源充裕度的多目标电网规划模型;通过对多阶段电网规划问题变量维数多、计算量大这一特点的分析,给出基于最小期望悔值的多阶段电网规划模型。本文主要内容如下:在深入分析当前各种主要规划求解算法优缺点的基础上,结合电网规划特性,提出应用蚁群算法求解电网规划问题。针对蚁群算法在求解大规模优化问题时易产生未成熟收敛的缺点,提出多种群并行蚁群算法和基于模式记忆的并行蚁群算法加以改进。多种群并行蚁群算法通过多个蚁群之间的并行搜索,提高收敛速度,同时可减少参数设置和种群规模对算法收敛性的影响;而基于模式记忆的并行蚁群算法则进一步通过模式记忆进行解空间分解,从而将问题规模化整为零,并能有效地识别、记忆和跳出局部最优解,最大程度的减少重复搜索,提高计算效率。算例表明这两种改进算法具有良好的计算效率和优秀的全局收敛性。通过对电网规划中各种不确定信息的分析、归类与合并,结合盲数特性,给出直接影响电网规划的3类4种主要不确定信息(节点注入功率——包括负荷预测和发电出力、系统潮流、线路投资)的盲数建模方法,并通过判断矩阵分析法确定各盲数模型的可信度分布。以确定性可用输电能力(ATC)的直流灵敏度分析法为基础,提出盲数ATC的直流灵敏度算法,并给出盲数运算的简化处理方法。在此基础上,提出计及电能资源充裕度的多目标电网规划模型,给出盲信息的模糊评价方法,通过模糊综合评判法解决多个目标之间的不可公度性和矛盾性。该多目标规划模型以盲数ATC、盲数投资以及柔性“N-1”过负荷概率的综合最优为目标,实际算例分析证明该模型能有效考虑电能资源的充裕度,其规划方案具有良好的经济性、可靠性和适应性。通过分析多阶段电网规划问题的特点和求解难点,指出采用多场景电网规划方法处理多阶段电网规划更加合理。在给出常规多阶段电网规划的数学模型并采用模式记忆并行蚁群算法进行求解的基础上,提出了基于最小期望悔值的多场景规划模型。该模型以规划方案在未来各种可能场景下的投资期望悔值最小为目标,实际算例分析表明此模型可以合理处理多阶段电网规划中的不确定信息,有效降低计算规模,并能使规划方案具有最好的综合适应性。更多还原

【Abstract】 Flexible electric power network planning includes two aspects: one is to build up reasonable planning model; the other is to give an appropriate solving algorithm. In this paper, the planning models and corresponding solving algorithms of electric power network flexible planning are studied in detail. By investigating the difficulty of global convergence in solving flexible planning model, the ant colony algorithm (ACA) and its improved algorithms are introduced to solve it. By synthetically considering the economy, reliability and flexibility of the planning scheme, a multi-object planning model concerning the power resource abundance is proposed. By analyzing the characteristic of high-dimensioned variables in the multi-stage planning problem, a minimal expectant regret concerned electric power network planning model and the corresponding solving method are proposed. The main content is as follows:Considering the character of power network planning, the ACA is proposed to solve single-stage network planning. Because the existence of premature convergence in ACA, two improving strategies are brought forward. One is multi-group parallel ACA (MPACA), and the other is the schema recording parallel ACA (SRPACA). On one hand, MPACA can improve the convergence speed by the parallel search of multi colony. On the other hand, it can also reduce the influence of the parameters’improperly setting effectively. Thus, it can avoid the premature convergence adequately. Besides the advantages which MPACA has, SRPACA can partition the solution space through schema recording, and can identify, record and jump away from the local optimal solution. In this way, the reduplicate search can be reduced furthest, and the computation efficiency is greatly improved. The simulation results of two sample systems show that SRPACA has high computation efficiency and good local & global convergence.By analyzing the characteristics of the four kinds of uncertain information which directly influence network planning (load forecast, generation, power flow and investment of lines) in detail, the blind number models of them are established. Furthermore, the distribution of the blind numbers’reliability can be calculated by means of‘judgment matrix analytical method’. Then, after introducing the direct current sensitivity analysis method of general ATC, the direct current sensitivity analysis method of blind number ATC is studied, and some predigestion tips are also proposed to decrease computational effort. Based on the calculation of blind number ATC, a multi-object planning model concerning the power resource abundance is formulated. By establishing the fuzzy evaluation model of blind information, the fuzzy integrated evaluation method is used to solve the multi-object model. This model can be solved easily by the SRPACA method proposed in Chap. 2. This optimal object of this model is the integration of blind number ATC, blind number investment and the probability of overload under flexible‘N-1’constraint. The simulation results of a practical sample system show that this model can consider the abundance of power resource effectively. The solutions of this model have good character of economy, reliability and flexibility.In multi-stage power network planning problems, the difficulty is that the values of decision variables in each stage are restricted by each other. The computational effort is directly proportional to the stage number. In this paper, a minimal expectant regret concerned multi-stage planning model is proposed and the SRPACA is used to solve this model. The simulation results show that the planned network under this model has the minimal expectant invest regret in all the scenarios, and the curse of dimensionality is also prevented.更多还原