Improved Transient Stability Simulation Based On Bidirectional Module Reduction

The time-domain trajectory simulation is one of the necessary tools to study thepower system behavior. The simulation provides data for many system securityanalyses. With the enlargement in system scale and capacity and the application ofnew facilities, the power system becomes much more complex to analyze. Thus theprogramming of the digital simulation in power system engineering has become morechallenging too: not only in accuracy but also in speed. In addition it should beconvenient to add models of new equipments to simulation programs. The main workof the thesis aims at improving the transient simulation algorithm to meet theserequirements.First, the general procedure of the system simulation is presented, followed by themathematic models of various components and their differential equations accordingto the trapezoidal rule. The basic idea of the Newton method is discussed, in whichthe repetitive reform and LU decomposition of Jacobian matrix impose a heavycomputational burden. This is the bottleneck to enhance the speed of transientsimulation.In Chapter 3, a technique of Bidirectional Module Reduction (BMR) isintroduced. By means of the BMR technique, the simulation programming based onNewton iteration has been modularized and easier to be expanded, with quickconvergence, high precision and good numerical stability. Equations of power gridand generators with their auxiliary controllers are derived for code program namedforward reduction and backward substitution method.To improve the efficiency of system simulation, certain inherent attributes inJacobian matrices of electric network and generator modules are utilized. This buildsthe basis of partial matrix decomposition technique in Chapter 4. During iterations,the decomposition of the Jacobian matrix will be restricted within a small sub-matrix,but retain the same result of decomposition for full Jacobian matrix. Thecomputational burden of LU decomposition is thus greatly reduced, while the fastconvergence of Newton iteration will remain.Finally, a simulation program using the new techniques is developed. Casestudies using the program are carried out on three realistic electric networks inChina—i.e. Mengxi, Shanxi and Northeast systems. Swing curves and simulation timeare compared with those obtained by BPA. They validate that the simulation speed oflarge scale system has been greatly enhanced...