| Modern electric power system is becoming more and more complex due to the deregulation of power markets, the increasing amount of interconnections and different types of control mechanisms, which make the detailed simulation of the planning of the system more urgent which requires more computation than before. In addition, from the security point of view, fast on-line security assessment software is needed to provide the on-line contingency analysis and preventive controls to mitigate the potential conditions. In the other side, with the development of parallel computation, the hardware for high performance computation is becoming cheap and software for message passing, such as MPICH and PVM, are public available now. This makes the high performance computation feasible and promising for power system operation and planning.; The distributed high performance computing applications in real-time power systems operations will be investigated in this dissertation.; The first application is the power flow. Power flow is the one of most basic function in EMS. In this dissertation, we develop two parallel linear solvers, parallel direct LU and GMRES linear solvers, to solve the linear update equations, which arise in the Newton power flow process. Based on these two parallel linear solvers, one class of linear multisplitting parallel power flow algorithms are presented in this paper. Also, another class of parallel power flow algorithms based on nonlinear multisplitting methods are presented here. In addition, this dissertation compares these two class of parallel power flow algorithms on a message-passing distributed-memory multiprocessor architecture. For four large-scale power system cases, the simulation results show that it is possible to achieve relatively good performance for the real time power system applications using a small cluster of workstations, which are connected via 100Mbit/s Ethernet.; Another application is the security constrained optimal power flow (SCOPF). The correct representation of security constraints may cause an increase of orders of magnitude in problem size, especially when considering nonlinear AC power system model. A new distributed SCOPF algorithm based on interior point methods is presented in this dissertation. It shows a very potential tool for the large scale deregulated power market.; Some other real-time power systems applications, such as state estimation and transient stability analysis, could be the future work. |
