The current international trend in distribution networks is towards increased monitoring. This trend is being driven by distribution network operators (DNOs) who hope that through increased monitoring, they will be able to optimise capital and operational expenditure and thus operate a more efficient networks. One of the key areas of focus relating to the increased interest in distribution network monitoring is power quality. Power quality disturbances affect consumers by interrupting equipment or halting industrial processes and can result in very significant financial losses. DNOs are also financially impacted by power quality issues if they breach regulatory limits or contractual arrangements. To extract value from power quality monitoring, DNOs must process and then interpret data from a variety monitoring devices placed at different locations all potentially measuring different quantities. The challenge of how best to extract useful and practical power quality information from disparate monitoring devices is the subject of this thesis. This thesis describes and develops monitoring techniques for two power quality phenomena: voltage sags and unbalance. The research presents new techniques which can graphically identify the weakest areas and the worst served customers for voltage sags and unbalance. All the developed techniques utilise non-deterministic methods (such as statistics and artificial intelligence) to deal robustly with network and measurement uncertainties. This thesis can be dissected into four areas: voltage sag monitoring, optimal power quality monitor placement, voltage unbalance monitoring and identification of the weakest areas and worst served customers for both issues. The first section of this thesis is dedicated to voltage sags. This section introduces a multi-step process to identify and estimate the impacts of voltage sags within networks. The first stage in this process is classification and detection where several different classification methods (including immune inspired techniques) are compared to determine which algorithms work best under the context of limited monitoring. The research then proposes a novel robust method for performing fault location and voltage sag profile estimation using multiple monitors. The method pays particular attention to the errors in measurement inputs and identifies the most likely location for both the fault location and the voltage magnitude using statistical methods. The voltage sag monitoring research concludes by defining the probable impacts of voltage sags on customers, and by introducing a new measure known as the sag trip probability. The second major section covered by this thesis is optimal monitor placement. This thesis presents a comprehensive methodology which enables network operators to place monitors in locations best suited for voltage sag monitoring based on future likely topological and loading changes. The third major section covered by this thesis is unbalance monitoring. A three phase distribution system state estimation model is developed which can estimate the location and impact of unbalance within the network, without assuming the loading is balanced. The final section of this thesis shows how the worst served customers and the weakest areas of the network can be identified presents for both voltage sag and unbalance using limited monitoring and the developed techniques. The results are presented graphically using a series of topological heat maps, and these show visually how the techniques could work to monitor a distribution network.

Harmonic propagation studies of public distribution networks require accurate models of aggregate residential customers (groups of customers) that simulate the harmonic emission of the multitude of household appliances in the network. Most of the present models were developed with the component-based approach, where models of individual household appliances are combined to build the model of multiple customers. This approach requires high amount of input data, like models of individual household appliances and detail information of customer behavior and device composition, which is usually not easy to acquire. However, with the increasing number of PQ-analyzers in the networks, the measurement-based approach is now more and more considered for the modeling of aggregate customers. The measurement-based approach uses measurements of the network in combination with top-down methodologies to obtain models of the aggregate customers. Compared to the component-based approach it has several advantages, like inherent consideration of the real operating changes of the individual household appliances, variation of customer behavior, effect of line impedances, cancellation and attenuation effects, etc. This thesis presents the development of a time-series stochastic model of the low-order harmonic emission of aggregate residential customers based on a top-down measurement-based approach. The model represents the daily variation of the harmonic magnitudes and phase angles. Besides, the model includes the representation of the harmonic unbalances, which is of great importance for the proper analysis of harmonic propagation in medium-voltage networks. The model is parametrized for German networks, but the methodology can be applied to find the models of other regions or countries.