Enhanced ambient signals based load model parameter identification with ensemble learning initialisation

Load modelling is significant to ensure the accuracy of power system simulation. In previous research on load modelling, various optimisation algorithms have been widely applied. However, the achievement of the global optimal solution depends on the quality of the initial feasible solutions (IFSs). In this study, an enhanced measurement-based load modelling approach with ensemble learning-based initialisation is proposed to solve this problem. In the proposed method, an ensemble intelligent machine (EIM) is trained offline to provide high-quality IFSs based on which the load model parameters can be identified through optimisation. The input features of the EIM are extracted through numerical subspace state-space system identification from the measurement data, while the output of the EIM is the estimated load model parameters. Then, based on the offline generated samples, a group of individual intelligent units (IIUs) is trained and selected first, after which they are integrated to form an EIM. The enhanced load modelling approach is tested in a simulation case for the Guangdong power grid. The results show that the EIM has better performance than all the IIUs, and the identification accuracy of the load model parameters can be improved with the EIM estimated parameters as the IFSs.