Research On Online Identification For Low Frequency Oscillation Modes In Power System Based On Least Mean Square Method

Under the circumstance of our state electric power industry debveloping,powergrids interconnection in the nation、the massive applications of ultrahigh voltage、high-capacity、long-distance transmission and plenty of applications of high-gainexcitation regulator in the generators are the necessary trend,but because of suchreasons in the former,low-frequency oscillation happens in the power system now andagain.And the low-frequency oscillation is a hidden danger to the safe reliability ofpower system.Once it happens,it will disturb the normal operation of power system.Consequently,how to monitor and control the low-frequency oscillation when it beginsor when the range of oscillation is not too obvious has became the hotspot which isstudied in the power syetem nowadays.The development and applications of Wide Area Measurement System(WAMS)based on the Global Positioning System(GPS) has afforded the chance for on-lineanalysiing and controling the low-frequency oscillation. At the first time,this article putsforward the Release of Relativity Least Mean Square algorithm which is one kind ofSelf-adpation algorithm appilicated in power system low-frequency on-line monitoring.Self-adaption filter algorithm takes the power system noise signal output as theresult of approximate steady white noise input,then analyzes the output by theSelf-adaption filter to get the filtering transfer function,and finally according to thetransfer function we can calculate the oscillation modes of power system.By analysing basic Least Mean Square algorithm and through simulation analysison basic LMS algorithm,the defect and the shortage of basic algorithm wasdiscovered.Hereby,Release of Relativity of Least Mean Square(RRLMS) algorithm wasput forward to improve the basic algorithm,and such algorithms was applied to actualpower system,further more,the LMS algorithms were successed to identificate lowfrequency oscillation in ChongQing power company.The key point of RRLMS meansunder the circumstance that the input signals are not mutual indepent,the basic LeastMean Square algorithm’s performance will go down,and the result of power systemidentification would deviate the truth-value,and especially the convergence rate will godown.Therefore,it need to release the relativity of input signal vectors,and try to keepthe statistic independence of input signals. For the first time,this algorithm was applied to actual power system,and further more,it runs steady and safely in the WAMS ofChongQing power company.And finally,according to measurement information of the Wide Area MeasurementSystem,this paper adopts PSAT simulation data and parts of State Grid’s active powerdata of area’s connecting line to identificate low-frequency oscillation,combination withpractical project advances the plans for reaserching on low-frequency oscillation basedon Wide Area Measurement System.The comparition among the identification results ofthe Release of Relativity Least Mean Square algorithm and basic Least Mean Squarealgorithm and basic ARMA algorithm shows that Release of Relativity Least MeanSquare algorithm has better effects in some parts,at the same time,matchs therequirement of practical project,proves that this algorithm has the validity in actualpower system.