Aggregate load parameter identification using general regression neural networks

In this paper, the general regression neural networks (GRNN) paradigm, as originally presented by Specht is summarized. Next, a GRNN implementation of a static power system load model simulation is described, and its performance is discussed. A generic load model is presented as a simple dynamic model suitable for wide-scale implementation using a GRNN and integrating measurement-based data. Finally, ongoing and future work are discussed, and preliminary conclusions are stated based on experience to date. The paper is application oriented. Its major contribution is the structuring of the load parameter identification problem into a form conveniently handled by neural networks. Generation of a training set is facilitated by the use of a simplified dynamic equivalent load model whose parameters are easily identified to form a training set from system measurements using least squares procedures.<>