Applications of Compressive Sensing In Smart Grids
StatisticsView Usage Statistics
The proliferation of distribution energy resources (DER) e.g. electric vehicles, distributed generators (DGs), and demand response necessitates visibility and situational awareness in smart networks. Unlike legacy radial networks with one-way power flow, smart distribution networks with multiple sources contain notable variability and uncertainties that must be continuously observed. Distribution systems are not fully telemetered so far due to the huge cost of deploying the required infrastructure in thousands or tens of thousands of feeders. Therefore, pseudo measurements are not widely available in real distribution networks. For this reason, those algorithms working with a few low-cost and highly accurate micro-phasor measurement units (µPMUs) have more practical benefits. To improve the reliability and resiliency of the power networks, efficient fault location schemes must be provided to facilitate rapid response to abnormal grid conditions. As a result, utilities will be able to better detect and head off potential blackouts, while improving day-to-day grid reliability and enhance the integration of clean renewable sources of energy onto the grid.In order to improve the situational awareness of grid conditions and promote real-time wide-area visibility, active management (AM) must be provided in smart networks. AM will be feasible if a distribution management system (DMS) with various functionalities, such as state estimation (SE), fault detection, isolation, and restoration (FDIR), and insulation assessment are enabled for system operation. Also, wide-area fault location in transmission networks plays an essential role to enhance the reliability and stability of power systems by promptly isolating, repairing, and restoring the faulted elements.In this dissertation, compressive sensing (CS) and ℓ1 regularization techniques are exploited to propose novel methods for distribution system state estimation (DSSE), single and simultaneous fault location in smart distribution and transmission networks, and partial discharge (PD) pattern recognition.In the DSSE studies, an electrical characteristic of distribution networks is used to cast the state estimation problem into a sparse vector recovery problem. Then, the system states are estimated using the voltage phasors at a few buses measured by µPMUs. In the fault location studies, a few smart meters/PMUs measure the pre- and during-fault voltages. The voltage sag values for the measured buses produce a vector whose dimension is less than the number of buses in the system. By concatenating the corresponding rows of the bus impedance matrix, an underdetermined set of equation is formed and used to recover the fault current vector. Since the current vector ideally contains few nonzero values corresponding to fault currents at the faulted points, it is a sparse vector which can be determined by ℓ1-norm minimization. Also, sparse representation classifier (SRC) is used for pattern recognition of partial discharges (PDs) in different artificial insulations, which were measured experimentally in a high voltage laboratory. An artificial neural network (ANN) classifier is trained and compared with the SRC method.