Parameter Identification of Composite Load Model Based on Bayesian Optimization

Abstract

Load modeling has a great impact on the analysis, operation, and planning of power systems. Using the composite load model (CLM) as the model structure, parameter identification is a major challenge in load modeling. Measurement-based parameter identification of CLM can be expressed as a complex optimization problem with differential-algebraic equations (DAEs)-based dynamic constraints. Conventional population-based evolutionary programming algorithms are computationally intensive as the fitness evaluation has to be carried out exhaustively for all the individuals. To enhance the computation efficiency in load modelling. In this paper, Bayesian optimization is proposed. Case studies are presented to demonstrate the effectiveness of the proposed Bayesian optimization method. Comparing with the conventional particle swarm optimization, the proposed Bayesian optimization can significantly reduce the computation time as fewer times of time-domain simulation-based fitness evaluation is needed. But both algorithms can achieve similar performance on the estimation accuracy.