A Hybrid Wavelet Transform and Neural-Network-Based Approach for Modelling Dynamic Voltage-Current Characteristics of Electric Arc Furnace

• Published in: IEEE transactions on power delivery 2014-04, Vol.29 (2), p.815-824
• Main Authors: Chang, Gary W., Min-Fu Shih, Yi-Ying Chen, Yi-Jie Liang
• Format: Article
• Language: English
• Subjects: Applied sciences; Computer simulation; Current measurement; Discrete wavelet transform (DWT); Discrete wavelet transforms; Dynamics; EAF; electric arc furnace; Electric arc furnaces; Electric power generation; Electrical engineering. Electrical power engineering; Electrical power engineering; Exact sciences and technology; Mathematical models; Matlab; Miscellaneous; Multiresolution analysis; Neural networks; Operation. Load control. Reliability; Power electronics, power supplies; Power networks and lines; radial basis function neural network (RBFNN); Table lookup; Training; Various equipment and components; voltage fluctuation; Voltage measurement
• ISSN: 0885-8977; 1937-4208
• DOI: 10.1109/TPWRD.2013.2280397

This paper proposes a discrete wavelet transform (DWT) and radial basis function neural network (RBFNN)-based method for modeling the dynamic voltage-current (v- i)characteristics of the ac electric arc furnace (EAF). The objective of the study is to develop a complete model of the EAF including different operation stages, and the model can be used as a harmonics and flicker source in its connected power system for the power-quality penetration or mitigation study, where the developed model can be embedded in the power system implemented by a commonly seen simulation tool, such as Matlab/Simulink. In the study, a combination of the DWT and the sequential RBFNN with parameters initialization algorithm is proposed to build the EAF v- i characteristics with enhanced lookup tables for different operation stages, where the field measurements of the EAF voltage and current are used to train the RBFNN for modeling the EAF load. Simulation results obtained by using the proposed model are compared with different measured data. It shows that the solution procedure accurately models the EAF dynamic v- i behavior. The proposed method also can be applied to model other highly nonlinear loads to assess the effectiveness of compensation devices or to perform relative penetration studies.