Transient stability-constrained optimal power flow using improved differential evolution and parallel computing

Optimal Power Flow (OPF) finds the delicate balance between economy and security in power systems. With the rapid increase of demand and deregulation of electricity markets, power systems tend to operate closer to stability boundaries. Thus, consideration of the transient stability limits in the OPF problem is becoming more and more imperative. It is, however, an open question as how to handle the stability constraints since transient stability is a dynamic concept with differential equations involved. Some conventional mathematical methods have been attempted mainly by approximating the differential equations to algebraic ones. However, conventional methods are sensitive to the starting points and have convergence difficulties in handling nonlinear, non-convex problems. Besides, the discretizing scheme will lead to computational inaccuracy and discrete control variable handling such as transformer tap settings is another problem for the conventional methods.;To address the above mentioned problems, this thesis developed an alternative solution based on Differential Evolution (DE). An improved version of DE with population re-initialization scheme is reported to ameliorate the premature problem of DE. As for transient stability constraints, a hybrid method which combines time domain simulation and transient energy function is employed to assess the stability of each individual with no limitation in system modeling. Since transient stability assessment is the most time-consuming part of the whole method, strategies called "stable-space push" and "fitness sorting" are also developed to reduce the searching space as well as the computation time. Other non-convex and discontinuous practical constraints that are difficult for conventional methods are also considered in this thesis. Performance of the proposed algorithm has been studied and compared with the reported results from conventional methods. Results show that the method developed is very powerful in solving nonlinear, non-convex, discontinuous complex optimization problem with both continuous and discrete control variables.;A parallel computation platform implemented on a Beowulf PC-cluster using Message-Passing Interface (MPI) technology is also built to speed up the proposed method. Case studies shows that parallelization does significantly improve the speed of DE and gives the possibility to realize online TSCOPF with moderate scale PC clusters and meet the real-world online application requirement.