Tirza Routtenberg

The problem of imbalance detection in a three-phase power network using a phasor measurement unit (PMU) is considered. In this paper, a general model for the positive sequences is presented based on time-domain PMU measurement at off-nominal frequencies. Then, the detection problem at the control center is formulated as a composite multihypothesis testing framework, in which the goal is to identify the unbalanced buses in the power network. A method for unbalanced bus identification is developed based on the minimum description length (MDL) model selection criterion. In addition to determining the unbalanced buses, the algorithm also provides an estimation of the off-nominal frequency and the system states. In contrast to the existing imbalance detection techniques, the proposed method is based solely on the single-phase positive sequence data, without the negative or zero sequences. Thus, it can be carried out at the control center without additional communication and processing costs. In addition, the novel networked MDL-based approach takes advantage of all the information from different buses in the network. The use of the networked information is expected to yield a lower probability of error compared to existing local bus-level detection approaches. Numerical performance analysis shows the applicability of the proposed networked MDL method for imbalance detection in a two-port π-model.