NSGA2遺傳算法改進研究及其在微電網(wǎng)配置中的應用
發(fā)布時間：2019-06-02 12:09
【摘要】：進化算法是一種極具適用性的多目標優(yōu)化方法,對全局尋優(yōu)極具優(yōu)勢,算法思想是自然界生物進化原則和優(yōu)勝劣汰法則。實際工程應用領域的優(yōu)化問題通常以多場景、多時段、多影響因素等為特征,并且附帶各種性質的約束限制條件,這為問題的解決加大了難度。優(yōu)化問題的約束處理有多種方法可以實現(xiàn),其中罰函數(shù)法受到很多學者的廣泛關注和研究,但是該方法存在固有缺陷即罰因子的設置問題。快速非支配排序遺傳算法(Non-dominated Sorting Genetic Algorithm 2,NSGA2)是一種典型多目標遺傳算法。本文以經(jīng)典NSGA2算法為基礎加以改進,提出改進型INSGA2算法(Improved Non-dominated Sorting Genetic Algorithm 2),用以解決含約束條件的多目標優(yōu)化問題。改進型INSGA2算法處理約束多目標優(yōu)化問題時,將約束條件轉化為待優(yōu)化目標之一,又因NSGA2算法在解決三目標以上優(yōu)化問題時,算法性能明顯下降,故本文只研究帶約束兩目標優(yōu)化問題。在INSGA2算法中,對不可行域內(nèi)性能較好的個體加以利用,將可行解和不可行解執(zhí)行遺傳操作以促使搜索加快向可行域方向靠攏,并自適應調(diào)整執(zhí)行遺傳操作的進化代數(shù)以減少進化后期低效冗余的遺傳操作;并為算法在可行域內(nèi)的搜索設置存活條件,即允許保留的個體必須滿足一定限制條件,此操作設計可以強化進化進程中的選擇壓力,防止進化出現(xiàn)停滯甚至退化現(xiàn)象,使進化趨優(yōu)趨勢更為明顯;在種群進化后期,性狀相似的個體過度擁擠過度重疊可能引發(fā)搜索的局部收斂,針對此問題,提出在種群進化后期執(zhí)行邊際變異操作。在算例驗證分析中,選取約束優(yōu)化基準測試函數(shù)和多模態(tài)優(yōu)化基準測試函數(shù)進行兩種算法的對比實驗驗證,實驗結果表明改進型算法是具有一定優(yōu)勢的。傳統(tǒng)供配電網(wǎng)絡遠距離大范圍互聯(lián)互通,發(fā)配電集中操作與控制,此種運作模式的缺陷日益顯現(xiàn)。逐漸受到重視的分布式發(fā)電(Distributed Generating Power,DG)和微電網(wǎng)(Microgrid,以下簡稱微網(wǎng))應用很大程度上彌補了大規(guī)模集中式供電的不足,提高了供電可靠性,加快電網(wǎng)智能化進程。然而DG的不當并網(wǎng)會給基于線路損耗、電能質量、經(jīng)濟因素、環(huán)境因素等考量的前期規(guī)劃產(chǎn)生干擾和沖擊,因此需對DG的定址定容進行規(guī)劃優(yōu)化。為使系統(tǒng)更加安全可靠和高效運行,本文從供電質量、經(jīng)濟成本、環(huán)境效益等角度出發(fā),研究分析DG并入微網(wǎng)的配置問題,以線路損耗、電壓偏移、初期經(jīng)濟成本、壽命周期碳排量為目標,并將不同目標進行兩兩組合,考慮系統(tǒng)正常運行的各項約束限制條件,以IEEE33節(jié)點配電網(wǎng)系統(tǒng)為實驗對象,進行基于NSGA2算法與INSGA2算法的微網(wǎng)規(guī)劃實例仿真驗證,實驗結果表明算法和模型是合理有效的。
[Abstract]:Evolutionary algorithm is a very applicable multi-objective optimization method, which has great advantages in global optimization. The idea of the algorithm is the principle of biological evolution in nature and the rule of survival of the fittest. The optimization problems in the field of practical engineering application are usually characterized by multi-scene, multi-time period, multi-influencing factors and so on, and attach various constraints and constraints, which makes it more difficult to solve the problem. There are many methods to deal with the constraints of optimization problems, among which the penalty function method has been widely concerned and studied by many scholars, but this method has inherent defects, that is, the setting of penalty factors. Fast undominated sorting genetic algorithm (Non-dominated Sorting Genetic Algorithm 2, NSGA 2) is a typical multi-objective genetic algorithm. In this paper, based on the classical NSGA2 algorithm, an improved INSGA2 algorithm (Improved Non-dominated Sorting Genetic Algorithm 2) is proposed to solve the multi-objective optimization problem with constraints. When the improved INSGA2 algorithm deals with constrained multi-objective optimization problem, the constraint condition is transformed into one of the objectives to be optimized, and the performance of NSGA2 algorithm is obviously degraded when solving the optimization problem of more than three objectives. Therefore, this paper only studies the two-objective optimization problem with constraints. In INSGA2 algorithm, individuals with good performance in infeasible domain are used to perform genetic operation of feasible solution and infeasible solution to promote the search to move closer to feasible domain. The evolutionary algebra that performs genetic operation is adaptively adjusted to reduce the inefficient redundant genetic operation in the later stage of evolution. The survival conditions are set for the search in the feasible domain, that is, the reserved individuals must meet certain constraints. This operation design can strengthen the selection pressure in the process and prevent the stagnation or even degradation of evolution. It makes the trend of evolution more obvious. In the late stage of population evolution, overcrowding and overoverlap of individuals with similar traits may lead to local convergence of search. In order to solve this problem, it is proposed to perform marginal variation operation at the later stage of population evolution. In the verification analysis of an example, the constrained optimization benchmark function and the multimodal optimization benchmark function are selected to verify the comparison between the two algorithms. The experimental results show that the improved algorithm has certain advantages. The defects of the traditional power supply and distribution network are becoming more and more obvious because of the long distance and large range interconnection and centralized operation and control of the power supply and distribution network. The application of distributed generation (Distributed Generating Power,DG and microgrid (Microgrid,), which has been paid more and more attention, makes up for the deficiency of large-scale centralized power supply to a great extent, improves the reliability of power supply and speeds up the process of intelligence of power grid. However, the improper grid connection of DG will interfere and impact the preliminary planning based on line loss, power quality, economic factors, environmental factors and so on, so it is necessary to optimize the location and capacity of DG. In order to make the system more safe, reliable and efficient, this paper studies and analyzes the configuration of DG integrated into microgrid from the aspects of power supply quality, economic cost and environmental benefit, with line loss, voltage offset and initial economic cost. The life cycle carbon emission is taken as the goal, and the different targets are combined in pairs. considering the constraints and limitations of the normal operation of the system, the IEEE33 node distribution network system is taken as the experimental object. The simulation results of microgrid planning based on NSGA2 algorithm and INSGA2 algorithm show that the algorithm and model are reasonable and effective.
【學位授予單位】：蘭州理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP18;TM727
[Abstract]:Evolutionary algorithm is a very applicable multi-objective optimization method, which has great advantages in global optimization. The idea of the algorithm is the principle of biological evolution in nature and the rule of survival of the fittest. The optimization problems in the field of practical engineering application are usually characterized by multi-scene, multi-time period, multi-influencing factors and so on, and attach various constraints and constraints, which makes it more difficult to solve the problem. There are many methods to deal with the constraints of optimization problems, among which the penalty function method has been widely concerned and studied by many scholars, but this method has inherent defects, that is, the setting of penalty factors. Fast undominated sorting genetic algorithm (Non-dominated Sorting Genetic Algorithm 2, NSGA 2) is a typical multi-objective genetic algorithm. In this paper, based on the classical NSGA2 algorithm, an improved INSGA2 algorithm (Improved Non-dominated Sorting Genetic Algorithm 2) is proposed to solve the multi-objective optimization problem with constraints. When the improved INSGA2 algorithm deals with constrained multi-objective optimization problem, the constraint condition is transformed into one of the objectives to be optimized, and the performance of NSGA2 algorithm is obviously degraded when solving the optimization problem of more than three objectives. Therefore, this paper only studies the two-objective optimization problem with constraints. In INSGA2 algorithm, individuals with good performance in infeasible domain are used to perform genetic operation of feasible solution and infeasible solution to promote the search to move closer to feasible domain. The evolutionary algebra that performs genetic operation is adaptively adjusted to reduce the inefficient redundant genetic operation in the later stage of evolution. The survival conditions are set for the search in the feasible domain, that is, the reserved individuals must meet certain constraints. This operation design can strengthen the selection pressure in the process and prevent the stagnation or even degradation of evolution. It makes the trend of evolution more obvious. In the late stage of population evolution, overcrowding and overoverlap of individuals with similar traits may lead to local convergence of search. In order to solve this problem, it is proposed to perform marginal variation operation at the later stage of population evolution. In the verification analysis of an example, the constrained optimization benchmark function and the multimodal optimization benchmark function are selected to verify the comparison between the two algorithms. The experimental results show that the improved algorithm has certain advantages. The defects of the traditional power supply and distribution network are becoming more and more obvious because of the long distance and large range interconnection and centralized operation and control of the power supply and distribution network. The application of distributed generation (Distributed Generating Power,DG and microgrid (Microgrid,), which has been paid more and more attention, makes up for the deficiency of large-scale centralized power supply to a great extent, improves the reliability of power supply and speeds up the process of intelligence of power grid. However, the improper grid connection of DG will interfere and impact the preliminary planning based on line loss, power quality, economic factors, environmental factors and so on, so it is necessary to optimize the location and capacity of DG. In order to make the system more safe, reliable and efficient, this paper studies and analyzes the configuration of DG integrated into microgrid from the aspects of power supply quality, economic cost and environmental benefit, with line loss, voltage offset and initial economic cost. The life cycle carbon emission is taken as the goal, and the different targets are combined in pairs. considering the constraints and limitations of the normal operation of the system, the IEEE33 node distribution network system is taken as the experimental object. The simulation results of microgrid planning based on NSGA2 algorithm and INSGA2 algorithm show that the algorithm and model are reasonable and effective.
【學位授予單位】：蘭州理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP18;TM727
【參考文獻】
相關期刊論文 前10條
1 李智勇;黃滔;陳少淼;李仁發(fā);;約束優(yōu)化進化算法綜述[J];軟件學報;2017年06期
2 陳建建;趙彩虹;高星辰;胡駿;陳笑;;配電網(wǎng)中分布式電源的優(yōu)化配置研究[J];電氣技術;2016年05期
3 秦躍進;王有春;范黎;文閃閃;劉欣;;微網(wǎng)中分布式電源的容量優(yōu)化配置方法[J];武漢大學學報(工學版);2014年03期
4 司賀秋;;《分布式發(fā)電管理暫行辦法》解讀[J];中國電力企業(yè)管理;2013年21期
5 楊大為;黃秀瓊;楊建華;張菁;奉斌;;微電網(wǎng)和分布式電源系列標準IEEE 1547述評[J];南方電網(wǎng)技術;2012年05期
6 解翔;袁越;李振杰;;含微電網(wǎng)的新型配電網(wǎng)供電可靠性分析[J];電力系統(tǒng)自動化;2011年09期
7 肖曉偉;肖迪;林錦國;肖玉峰;;多目標優(yōu)化問題的研究概述[J];計算機應用研究;2011年03期
8 劉學平;劉天琪;王劍;;基于小生境的配電網(wǎng)多目標分布式電源規(guī)劃[J];電網(wǎng)技術;2010年10期
9 蘇玲;張建華;王利;苗唯時;吳子平;;微電網(wǎng)相關問題及技術研究[J];電力系統(tǒng)保護與控制;2010年19期
10 梁昔明;龍文;秦浩宇;李山春;閻綱;;基于種群個體可行性的約束優(yōu)化進化算法[J];控制與決策;2010年08期
相關重要報紙文章 前2條
1 薛傳琦;;“十三五”可再生能源發(fā)展規(guī)劃發(fā)布[N];中國電力報;2016年
2 張翼;;可再生能源“十二五”規(guī)劃發(fā)布[N];光明日報;2012年
相關碩士學位論文 前4條
1 陳春泉;配電網(wǎng)分布式電源優(yōu)化配置[D];廣東工業(yè)大學;2013年
2 楊R，
本文編號：2491088
本文鏈接：http://sikaile.net/kejilunwen/dianlidianqilunwen/2491088.html
最近更新
教材專著
