Eigenvector selection in load classification and model generalization

It has been well recognized that the load model has great effects on the power system analysis and control. However, it is also widely known that the load modeling is a quite difficult problem due to its special properties such as time-variant characteristics of the load. In this paper, the measurement-based load modeling approach and multi-curve identification technique are applied in Hushitai power station in Northeast China. In order to build practical load models that can describe the actual load characteristics without changing the model parameters too often, cluster analysis is used to classify the loads. Two main kinds of eigenvectors used in cluster analysis are also compared. Finally, based on the load classification, a new aggregation method is proposed to build load models with good generalization capability. Case studies show the efficiency of the proposed method.