Study On Dynamic Equivalence Method And Related Technology For Large Scale Power Systems

With power system development, modern power system is gradually grown up towardsinterconnection of large scale so as to enhance the quality of electric energy and the reliabilityof power supply. Thus, how to guarantee safe and stable operation for large scale powersystems becomes a significant and urgent issue. Major challenges include the difficulty toanalyze transient stability for such large scale power systems with numerous elements modeledin detail on one hand, and the limited professional software capability on the other hand, suchas Real Time Digital Simulator (RTDS), Electro Magnetic Transient in DC System (EMTDC),etc. Using dynamic equivalence technology to reduce power systems is considered to be aneffective measure to solve this problem. In order to improve dynamic equivalence method andits related technology for large scale power systems, some research has been done in thedissertation, and main key points as follows:1. An algorithm marked as PSO-FCM, namely, the Fuzzy c-means Clustering (FCM)based on the Particle Swarm Optimization (PSO), is presented to solve the partition problem ofcoherent generators. In this algorithm, the cluster center is taken as a particle for encoding, andthe parallel property and the global searching capacity of the PSO algorithm are applied to theoptimization through continuously renewing the velocity and position of the particles. Thus,the dependency of the FCM algorithm upon the initial value and the risk of falling into localoptimum are largely reduced. Moreover, in order to evaluate the clustering validity, threevalidity functions are constructed. Simulated results on IEEE10-generator39-bus system showthat the PSO-FCM algorithm is suitable for the partition of coherent generators of powersystems in various operation modes because it is of simplicity, easy realization, high accuracyand high speed.2. An algorithm marked as modified PSO-FCM is presented to solve the partitionproblem of coherent generators. In order to improve PSO-FCM convergence performance, twoways are presented: convergence factor and disturbance introduced to the formula ofPSO velocity by randomly selecting optimal position of particle, which can increase diversityof particle swarm and avoid premature on account of aggregation of particle. The modifiedalgorithm can easily converge to global optimum. Simulated results show that the MPSO-FCM algorithm is suitable for the partition of coherent generators due to the improved convergenceperformance the original advantages of PSO-FCM.3. The dynamic equivalent based on coherency is studied on Hainan Power Grid by thePower System Dynamic Equivalence Program (PSDEP) and large simulation tool PSD-BPAfor power systems. First, the overall principle of dynamic equivalence is presented. Second, inorder to remedy the deficiency of Powell optimization algorithm in PSDEP which is used toaggregate equivalent generators, weighting method is presented as supplement to aggregateparameter of coherent generators, and then Differential Evolution (DE) algorithm which hasexcellent property is presented to optimize parameter of equivalent generators acquired byweighting method. Self-adaptive strategy is adopted for zoom factor and crossover probabilityfactor in DE. Finally, simulated results show that the data scale of Hainan Power Grid isreduced six times approximately, and the power flow of equivalent system is almost identicalwith original system, and the main dynamic characteristics of original system retain well. Thealgorithm convergence and precision of equivalent generators’ parameters are improved byself-adaptive strategy in DE. The coherency-based dynamic equivalent method by modifiedweighting method is suitable for reducing large scale power systems.4. Based on small signal analysis procedure SSAP, correlation factor of eigenvalueanalysis for systems is selected as characteristic quantity, and a method combining k-meansclustering and hierarchical clustering is proposed to identify generators with large-scalestatistical software-Statistical Package for Social Sciences (SPSS). On the basis of result ofcoherent generators’ identification, it can aggregate coherent generators to simplify powersystems, and also serve as selecting principle of generators with PSS installed. It proposes thatPSS is installed in strongly correlated generators, and Chaotic Particle Swarm Optimization(CPSO) algorithm for PSS parameters optimization is presented. The disadvantage of PSO,leading into local optimum, and that of chaotic search, sensitive to initial value, are overcomein CPSO. CPSO is independent on initial value, high speed and good ergodic property. Bytesting on Hainan Power Grid based on SSAP, it is concluded that the damp of system isimproved after coordinated optimization of PSS parameters based on coherent identificationresult of generators. The method is effective to solve the problem of identification coherentgenerators in small signal analysis, and also presents a solution of reasonable allocation of multi-generator’ PSS. The small signal stability is enhanced for power systems with themethod, which has a better prospect in engineering application.5. A new method of estimation dynamic equivalence used for online is presented. First,the model of external equivalent system which comprised equivalent generator and compositeload is given. In equivalent system model, refined six-order model of generator is adopted, andstatic state ZIP load combining three-order dynamic inductive motor are included in compositeload model. Then, it is presented the identification strategy based on Differential Evolution(DE) algorithm. In order to solve premature problem of DE, double differential evolutionswarms with different variation are formed, and considering double swarms parallel evolution,and exchange message at regular time. The aim is to increase diversity of swarms and improveconvergence of DE. Simulated result shows that the method is effective to parameteridentification problem of equivalent system. The algorithm is simple and rapid, and it is higheraccuracy and better robustness for identified parameter. Equivalent system model obtained isbetter accord with practical power system, and equivalent system can preserve main dynamiccharacteristics of external system well. The new estimation equivalent method can be used toreduce external system of large scale power systems online.6. A method for identification and aggregation coherent inductive motor loads is proposed.Firstly, it is presented the algorithm of Fuzzy c-means Clustering by Chaotic Particle SwarmOptimization (CPSO-FCM) to identify coherent motors, on the basis of the characteristicswhich reflect the dynamic performance of motor. The merits of CPSO algorithm areconvergence to optimal value rapidly, independence initial value, and parallel search so as toovercome FCM’ disadvantage which is upon the initial value and has the risk of falling intolocal optimum. A validity function is constructed so as to evaluate the clustering validity.Secondly, identification strategy is employed to aggregate coherent inductive motor loads, andthe parameter of equivalent motor load is acquired by using modified double swarms DE. Bytesting on IEEE10-generators39-bus system, the conclusion is presented that CPSO-FCMalgorithm is rapid convergence and robust. It has high accuracy of identified load parameter,and equivalent system can preserve the main dynamic characteristics of original system well.The method can be used to simplify mass inductive motor loads in online estimationequivalence. This dissertation is supported by the National High Technology Research andDevelopment Program of China (863Program)(2011AA05A102).