Research On Improving Large Scale Power Flow Calculation Convergenceusing Levenberge-marquardt Method

The power system is the most complicated industrial system. It is characterizedas nonlinearity and high dimensions. In this system power can be transmitted frommechanical energy or other kind of form into electrical energy. Lots ofmodernsystems should be driven by electrical energy highlight the importance ofpower system.Through decades of rapid development of power system it forms apower network with large scale high voltage and much more complex items.In orderto ensure the normal operation of power system power system analysis is urgentlyneeded. With the popularization of computers and more research work on powersystem, load flow calculation short circuit calculation and stability analysis form thefundamental parts of power system analysis. The load flow calculation is developedas the basic tool for system controlling and operation. Results from power flowcalculation can help operators to understand the actual operation situation of powergrid and initiate the later analysis such as stability analysis. With the enlarging scaleof the power grid and variations of operation situations, dispatchers can get the powerflow distribution which satisfiesthe practical requirements after they meet lots ofill-conditioned load flow that conventional newton method may diverge. If the newtonmethod oscillates or even diverges it is difficult for operators and dispatchers to obtainthe effective information to adjust the power flow. Under that situation we can adjustthe load flow only relying onartificial experiment. This kind of method is lowefficiency and difficult to obtain the condition of power grid close to reality. Insummary, research work on algorithms which enhance the load flow convergence hastheoretical value and practical significance.In this thesis the history of algorithms for enhancing power flow convergence isfirstly introduced. Secondly the wildly used optimal multiplier method and theemerging tensor method for improving load flow convergence are introducedtheoretically. Other kind of method based on Optimization algorithmis also introducedand summarized. Then Levenberg-Marquardt(LM) method is introduced based on theleast square model of power flow equations. Combining the latest self-adaptive LMmethod achievementwe get a new kind of load flow calculation method. In this thesisthe characteristics of self-adaptive LM method is studied and summarized. This kindof method has the ability to steer clear of the singular area of Jaccobian matrix to enhance the convergence of power flow calculation. According tosome relative mathematical theory the self-adaptive LM method has the ability to getthe least square solution of power flow equations quadratically. A new kind of sparsetechnique is developed to calculate the iteration step of LM method. This newtechnique simplifies the program interface between LM method and conventionalnewton-Raphson method.The calculation of domesticate large-scale test system verifythat the self-adaptive LM method with sparse technique is a prospective method inengineering applications. Finally the self-adaptive LM method the optimal multipliermethod and the tensor method are compared in terms of computational cost andconvergence properties under different situations. The comparison of differentmethods testifies that the self-adaptive LM method is a general method to improve theconvergence of power flow calculation. At the end of this thesis the self-adaptive LMmethod is applied into small or media scale of AC/DC hybrid system. This exampleillustrates the self-adaptive method has potential to enhance the power flowconvergence of AC/DC hybrid system.