Affordable Access

Exact and Approximate Methods for Computing the Hessian of a Feedforward Artificial Neural Network

Purdue University
Publication Date
  • Computer Science


AUPEC 2004 Full Paper Template ON-LINE IDENTIFICATING OF THE PROPORTION OF DYNAMIC COMPONENT IN COMPOSITE LOAD MODEL SHI Zhenhui* ZHU Shouzhen* ZHENG Jinghong* WANG Guang* and LOU Hongxiang ** *Department of Electrical Engineering Tsinghua University ** Electric Power Research Institute of Shansi P. R. China Abstract The proportion of dynamic load model in composite load model has been determination by using an improved ANN method in this paper. The importance of the proportion of dynamic component in composite load model to power system analyses is discussed first. Then Levenberg-Marquardt back propagation algorithm is used for off-line training of the feed-forward network. The developed identification method is tested by the simulation data generated using Mathlab tools, the dynamic system data measured from dynamic simulation system, and the data obtained from filed tests. The well trained Neural Network model is used online to identify the proportion of dynamic component for a real power system. The test results show that method is feasible and effective. 1. INTRODUCTION The composition of power system load is quite complicated since a large number of components may affect the system model, including static load component such as resistance loads and reactive loads, as well as dynamic load components such as the loads driven by induction motor and other devices. Thus, it would be impractical to describe the system characteristic by detailed system components. The common practice in most of cases is to adopt the composite equivalent load model which is consist of an equivalent static load component represented by constant impedance or other static model, and an equivalent dynamic load component represented by induction-motor model according to a certain proportion for the load modeling in power system stability analyses, control and other studies [1]. In transient stability analyses, the proportion of dynamic load co

There are no comments yet on this publication. Be the first to share your thoughts.


Seen <100 times