Harmonic Source Identification Using Partial Least Squares Regression

Harmonic distortions in power systems are an increasingly significant concern for utility companies and can adversely impact the efficient utilisation of a power system, driving up costs. Being able to identify the sources of these distortions would have a sizable impact on network and business planning. This project explores the methodologies proposed in literature for harmonic source identification and evaluates their practicality and applicability to the existing Western Power network and available resources. A databased approach, based on Partial Least Squares (PLS) regression, has been chosen and further developed into a software tool that can be used by power engineers. The algorithm has been tested on real world field-data with preliminary results illustrating how it can be used by power engineers to aid network planning and analysis. The methodology will be subject to further validation by comparing results with an alternative methodology, based on system impedances, carried out in the same test environment. Further recommended study includes testing in a simulated environment and testing the effect of increasing the sampling rate.