Multi-Start Strategy based Global Optimization Method for Ambient Signal-based Load Modeling

Abstract

Load modeling is critical for power system transient analysis and simulation. In recent studies, ambient signal-based parameter identification is considered as an effective method to obtain time-varying model parameters. However, low fluctuation amplitude of ambient signals may exacerbate the issue of multiple local optimal solutions, which is derived from the nonconvex and nonlinear characteristics of parameter identification problem. In this paper, a multi-start strategy based global optimization method is presented for ambient signal-based load parameter identification. Firstly, the selection of start points is formulated as a maximum diversity problem which can be solved by greedy algorithm. Then, we apply trust region-based local solvers to identify model parameters and obtain global optimum from multiple local solutions. Simulation results in the IEEE-39 bus system have demonstrated the global accuracy and efficiency of the proposed method.