Parallel Domain Decomposition Based Power System Simulation

Abstract

With the continuous developing of power systems, simulation of modern systems requires more computational resources and more simulation time. Conventional techniques are not efficient due to the large size of equation set and inconvenient modelling methods. Although computing power has increased dramatically, we still need new algorithms to improve the simulation speed. To address the above problems, this thesis proposes an efficient parallel calculation algorithm based on domain decomposition method. The new approach reduces the computational cost by calculating in parallel and splitting the equation set. The proposed approach partitions the system into several subsystems to be solved in parallel on different processors and at the same time preserves the details of the original system. To reduce the cost of communication overhead and computation load, latency characteristic is utilized. Not all the Jacobian matrix is updated in every iteration. Traditional simulators (transient stability and electromagnetic transient) as well as the hybrid simulator are described and compared with this algorithm. Computational examples are used to validate the accuracy of new proposed algorithm and its efficiency.