| Harmonic distortion is one of the main power quality problems for power system utilities. Nowadays, there are many harmonic-generating loads in a given distribution or sub-transmission system. Developing methods and techniques to quantify the harmonic contributions of the customers and the utility system, especially when a harmonic problem occurs in a system, is highly important for power quality management. After identifying the major harmonic-producing customers, utility companies can negotiate with them to reduce their generated harmonic contents by either installing filters or using other harmonic mitigation approaches. However, the first step is to identify the major harmonic-producing loads and quantify their impact.;In the past, this problem was approached from a single-point perspective. The single-point problem is a classic harmonic determination problem. However, the previous methods are circuit-based and classified as invasive methods. The thesis proposes a new non-invasive data-based method. The harmonic impact of a load is calculated by just measuring its voltage and current at the Point of Common Coupling (PCC). The challenge here is data selection. Not all the measured voltage and current sets are suitable for the analysis. The method is verified and characterized by extensive simulation studies. By using field measurement data, the effectiveness of the method is verified.;While the single-point approach is still very important and worthwhile, another type of harmonic-source-detection problem has emerged, primarily because an increasing number of loads now contain some harmonic sources. In this multi-point problem, the goal is to quantify harmonic impacts of the potential suspicious loads in the network on a reported harmonic problem. It must be determined if these loads are causing the problem and, if so, which load is producing the most significant impact. The multi-point problem has never been studied by other researchers. This thesis proposes two new data-based methods for the multi-point problem. For these methods, the harmonic currents of the suspicious customers and the harmonic voltage at the point of the reported problem should be monitored. By using statistical inference, the harmonic impacts of the loads are estimated directly from the measurement. The idea is to correlate the gradual change of a load to the gradual change of the problem. One of the main challenges of this correlation analysis is the data selection. The thesis proposes and studies different data selection algorithms. The methods are verified and characterized through extensive simulation and field measurement studies. |
