基于GNN和蒙特卡羅法的電力采購(gòu)系統(tǒng)圍標(biāo)串標(biāo)行為辨識(shí)算法

打開(kāi)文本圖片集

中圖分類(lèi)號(hào)：TM73 文獻(xiàn)標(biāo)志碼：B 文章編號(hào)：1671-5276(2025)03-0301-05

Identification Algorithm of Bidding Behavior in Power Procurement System Based on GNN and Monte Carlo Method

WU Yong，PAN Xiaohua，YANG Yijun，YE Xuefeng，LIU Fuquan（China Southern PowerGrid Digital Platform Technology（Guangdong）Co.，Ltd.，Shenzhen518102,China）

Abstract:Influenced bythe diversityof biddng methodsand thecomplexityofnoderelationshipin power procurement system, theidentificationaccuracyisusualldissatisfedduetoneglecting theconsiderationofbidingpaternregularitywhenidentifying thebehaviorofbiddingcolusion.Tosolvethisproblem，anidenticationalgorithmof bidding behaviorinpowerprocurement systembasedon GNNand Monte Carlomethodisproposed.The network diagram isconstructed bycombining theentityand cooperationrelationshipwithinthepowerprocurementsystem，andtheatentionmchanismaggregationfunction is introduced. The embedding layer isused to transform theembedded representationof node/edge feature vectors，capturing thestructural atribute informationintheoriginalpowerprocurementsystem.Astatevectorisassignedtoeachbidderandthestatetransition probabilityiscalculated.With MonteCarlomethod,thebid-winningbehaviorissimulatedthroughrandomsampling.Taking the bid price diffrence,，bid winingfrequencyandbidding modeasthekeycharacteristicsofbiddingcollsion，therisk indexis calculated，andcombinedwiththejudgmentresultofidentifcationreshold，theiddingcollusionbehaviorisidentifidfrom thepowerprocurementtransactioprocess，ndeidentificationaccuracyoftheproposedmethodistested.Thefinaltstrsuls showthatthelogarithmicloss functionvalueof theidentficationresultislowandtheidentificationaccuracyisobviouslyhigh. Keywords:electric power procurement；string bidding；behavior identification；bidding；GNNalgorithm；Monte Carlo method

0 引言

電力采購(gòu)系統(tǒng)中，圍標(biāo)串標(biāo)行為往往具有隱蔽性、復(fù)雜性和動(dòng)態(tài)性等特點(diǎn)，傳統(tǒng)的監(jiān)管手段難以有效應(yīng)對(duì)。（剩余8617字）