Research On Power System Intelligent Reactive Power Optimization And Quasi Real Time Reactive Power/Voltage Control

Reactive power oPtimization is of great impotae to voltage quality twee, powerloss minimization and thUS security and economic operation of power system. Detaildiscussion and partition of it is given according to the research contellts, methods, tfor sPanand objective function etc. The method aPplied here, Genetic AlgorithIn (GA) for reactivepower planning problem and interior point method (IPM ) for quasi real time voltage/reactivepower contrOl are separate surnmarized fully and comprehensively Hence, twO aspects ofreactive power operating optimiZation, i.e. optimal reactive power compensation planning andquasi-real time voltage/reactive power control are studied in detail in this thesis withintelligent method, such as fully mathematic and expert system etc., integrated. The researchincludes following contents'l. According to the characteristics of reactive power planning optimization, a fullyscheme of refined GA was proposed in this thesis. A tWo-stage hybrid scheme is proposedbased on the criteria that decreasing mode nUInber reasonably and different attention shouldbe given to different class control variables, i.e., at the begirming stage, a larger granu1arity isused to make GA searching the area contaning optimal sollltion with less generation. Thehybrid scheme of binary and integer code distinguished by voltage level is of clear levelconception thus can search the high voltage control variables more carefully. Crossover andmutation of GA are also refined on the basis of above scheme. And a hybrid crossover methodand a tail prevail mutaion principIe are proposed. Furthermore, the production method of firstpopulation according to initial power flow mode and the simplification of control vedablesbased the conception of power plantistation are discussed also. Aimed at conquering the local-searching shortcoming in GA, after establishing fuzzy regulation scale of normal voltage, a socalled on-site regulation expert Anowledge based on the plantrstation concept in power systemis implemented to eliminate voltage /reactive power overlimit, to amer decrease investInentand to minimiting active power loss.2. GA has the common shortcomings While aPplied in large-scale netWork, such asconsuming huge calculation time, easy to be traPped by local fninimize and not so goodability of searching local optimal solution. The Pseudo-Parallel GA technology and adaPtivechanging population size are used to generate ndtial poPulaton, thus elevating the quality ofinitial population and furthermore improving optimal solution, voltage/reactive is of localcharacteristic in actual power system, and this fetch oul decomposing sub-area GA accordingto network natural management region. A noveI aggregation in selection process is used togenerate new individuaI betWeen individuals. To improve calculation speed, a conventionalIoptimal method-successive quadratic programming is combined with GA to compose a hybrid GA which can greatly decrease calculation time and improving the precision of optimal solution.3. Reactive power optimization based on interior point method is studied in this thesis as that GA's calculation speed can not satisfy the requirement of real time voltage control. A full variable quadratic programming model is established for reactive power optimization problem. And a new methodology algorithm based on primal-dual IPM is deduced for such a non-standard quadratic programming model. Furthermore, over limit correction is conducted based on IPM, and selection of the ready to be corrected offending constraints and the corrective control variables are also discussed, as well as how to decrease control times. It is also noticed that some voltages be too close to its upper boundary and this may cause reactive power control equipment to act too often thus decreasing network security. Fuzzy security constraints, such as voltage and generator's reactive power output, are introduced in QP model, and a new IPM with fuzzy constraints is deduced in detail, being the critical program in quasi real-time voltage/reactive control.