A new method for power system load modeling using a nonlinear system identification estimator

This paper proposes a new method for measurements-based modeling of nonlinear loads in power systems. This method includes a combination of a binary tree algorithm with Nonlinear Auto-Regressive with Exogenous Input (NARX) identification. The paper demonstrates that the new method performs well without any prior knowledge of the system structure. In contrast to other load modeling methods, which are typically aimed for particular studies or load types, the proposed method can be used with any load type and for any study. Accurate load modeling is particularly important for studies of industrial networks and grids. In this study, a field data set was collected in a mine site from a large electrical rope shovel. This data set has been used to develop a model of the rope shovel based on the proposed binary tree - NARX algorithm. When compared to other known methods, such as Wavelet and Sigmoid networks, the proposed method has shown the fastest training time and the highest accuracy. Finally, the modeling results have been verified against another set of field measurements from an existing network, and have shown a very good agreement.