Application of regression models to predict harmonic voltage and current growth trend from measurement data at secondary substations

This paper presents the applications of regression models to predict the future growth trend of total harmonic distortions (TFID) of 40 power transformers from field measurement data. The prediction values are important and can be used for transformer planning. Those transformers which feed different types of customers reveal different harmonic characteristics. While the linear regression model is suitable for harmonic voltage distortion, the nonlinear exponential regression model is better for current. All transformers’ harmonic voltage distortions follow with harmonic current distortions, that is, feeder loads are harmonic sources. The prediction harmonic values of regression analysis are lower than that from the method of average spread rate.