資源描述:
《(最新)基于人工智能的配電網(wǎng)故障恢復(fù)重構(gòu)研究(論文)》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、南京理工大學(xué)碩士學(xué)位論文基于人工智能的配電網(wǎng)故障恢復(fù)重構(gòu)研究姓名:李鵬飛申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):電力系統(tǒng)及其自動(dòng)化指導(dǎo)教師:都洪基20080626碩士論文基于人工智能的配電網(wǎng)故障恢復(fù)重構(gòu)研究摘要網(wǎng)絡(luò)重構(gòu)是配電系統(tǒng)運(yùn)行和控制的重要手段,也是配電管理系統(tǒng)的重要組成部分。隨著智能技術(shù)的發(fā)展,運(yùn)用智能算法重構(gòu)配電網(wǎng)絡(luò)來達(dá)到降低網(wǎng)損的目的已經(jīng)成為一種可能。論文闡述了配電網(wǎng)重構(gòu)的背景、現(xiàn)狀與特點(diǎn);研究了目前較為流行的配電網(wǎng)重構(gòu)算法;分析了各種配電網(wǎng)重構(gòu)的目標(biāo)函數(shù)以及約束條件。在此基礎(chǔ)上,論文針對(duì)支路交換法,研究其算法原理,并對(duì)算法進(jìn)行有效地改進(jìn)。運(yùn)用PSASP電力系統(tǒng)分析綜合程序
2、,對(duì)不同結(jié)構(gòu)的配電網(wǎng)絡(luò)進(jìn)行實(shí)際算例仿真,考察算法的實(shí)用性。BP神經(jīng)網(wǎng)絡(luò)等人工智能算法能避免一般算法反復(fù)進(jìn)行潮流計(jì)算而耗費(fèi)時(shí)間的缺點(diǎn),并且有很好的全局逼近優(yōu)點(diǎn)。論文從BP算法自身著手,針對(duì)其學(xué)習(xí)速度慢和可能局部收斂?jī)蓚€(gè)問題,采取有效的手段分別改進(jìn)。論文還從圖論知識(shí)著手,構(gòu)造三種數(shù)學(xué)模型。用配電網(wǎng)的樹形拓?fù)浼s束改進(jìn)BP算法的誤差函數(shù),并且對(duì)BP神經(jīng)網(wǎng)絡(luò)的輸出結(jié)果做了環(huán)網(wǎng)和孤立頂點(diǎn)約束判斷,很好地解決了BP算法可能收斂到局部最小的缺點(diǎn)。最后以實(shí)際算例對(duì)改進(jìn)算法進(jìn)行仿真測(cè)試,證實(shí)該算法具有輸出結(jié)果速度快、能有效全局逼近的特點(diǎn)。BP神經(jīng)網(wǎng)絡(luò)還有自學(xué)習(xí)——這一最重要也最令人注目的
3、特點(diǎn),非常適合長(zhǎng)期運(yùn)行的配電網(wǎng)絡(luò)。因此,論文所研究的算法有很好的應(yīng)用前景。關(guān)鍵詞:配電網(wǎng)重構(gòu),支路交換法,BP神經(jīng)網(wǎng)絡(luò),圖論Abstract碩士論文Networkreconfigurationisnotonlyallimportantmethodincirculatingandcontrollingofdistributionsystem,butalsoallimportantpartofdistributionmanagementsystem.Withthedevelopmentofintelligenttechnology,itispossibletoreduce
4、thelossofenergyintheelectricnetworkthroughintelligentalgorithm.Thebackground,thesituationandthecharacteristicsofthereconfigurationofthedistributionnetworkarediscussedinthispaper,andthepopularalgorithmsforreconfigurationofdistributionnetworknowadaysareresearched.Furthermore,theobjectivef
5、unctionsandconstraintconditionsofthereconfigurationareanalyzed.Onthisbasis,accordingtothebranch-exchangealgorithm,itsprincipleisinvestigatedandisimprovedefficientlyitself.Andthen,throughthesoftwareofPSASPpowersystem,actualexamplesandsimulationresultsofthedistributionnetworksofdifferents
6、tructuresaregiventoverifythepracticabilityofthismethod.ArtificialintelligentalgorithmCanavoidrepeatedlypowerflowcalculationofordinaryalgorithms,anditisadvantageousinfunctionapproximation.Inallusiontotheproblemsofitslowlearningspeedandlocalconvergence,workhasbeendonetoimprovetheBPalgorit
7、hm.Followingbytheknowledgeofgraph,threemathematicalmodelsareestablished.Throughtheconstraintsoftree—topologyofthenetwork,theerrorfunctionisimproved,anddeterminationofring—webandisolatedvertexaremadetoefficientlysolvetheproblemoflocalconvergencetominimumoftheBPalgorithm.Proceedi