Assessment of smart grid fault localization using distributed acoustic sensing in urban distribution networks

Thesis Abstract

The reliability and resilience of urban distribution networks are increasingly challenged by fault events that disrupt power delivery and escalate maintenance costs; traditional fault localization methods often suffer from delays and limited spatial resolution in dense urban environments. This study addresses the problem by evaluating distributed acoustic sensing (DAS) as a real-time fault localization modality to enhance situational awareness, reduce restoration times, and improve fault containment in urban grids. The aim is to quantify the accuracy, speed, and operational benefits of DAS-based localization relative to conventional protection schemes, and to develop an integrated framework for DAS-assisted fault diagnosis. Specific objectives include (i) assessing DAS signal-derived fault location accuracy within ±50 meters across a 25-kilometer urban feeder network, (ii) identifying the influence of conduit layout, temperature, and ambient noise on DAS performance, (iii) evaluating the impact of DAS-assisted localization on fault restoration time and outage duration, and (iv) proposing an operational protocol that integrates DAS with existing SCADA and protection systems. A mixed-methods research design is employed, combining quantitative field measurements with qualitative expert assessments. The population comprises four urban feeder circuits in a metropolitan area with dense building stock and existing fiber-optic DAS deployment, totaling approximately 60 kilometers of DAS-accessible fiber and 280 monitored substations. A stratified sampling approach yields 20 fault events collected over 18 months, supplemented by 10 historical fault cases for benchmarking. Data collection instruments include high-fidelity DAS data streams (coherence, strain, and temperature signals sampled at 1 kHz), event-logging from SCADA/Protection systems, GIS-based network topology data, and semi-structured interviews with system operators. Validity and reliability are addressed through calibration trials, cross-validation with traditional fault indicators, and inter-rater reliability checks for qualitative inputs. Analytical techniques comprise time-series processing of DAS measurements to extract wavefront arrival times and characteristic frequencies, followed by multilayer regression models to relate DAS-derived locator estimates to ground-truth fault positions obtained from pole tests and post-event investigations. A Bayesian inference framework is employed to fuse DAS signals with SCADA data, yielding probabilistic fault location maps and uncertainty quantification. The analysis also incorporates ANOVA to examine the effect of environmental variables on localization accuracy and a responsiveness assessment using survival analysis to model restoration times. The study tests predefined hypotheses (H1) DAS reduces fault location error to within ±50 meters in at least 70% of events; (H2) integration of DAS with SCADA decreases mean restoration time by at least 25% compared to baseline protection schemes; and (H3) environmental and infrastructural factors significantly modulate DAS performance. Expected findings indicate that DAS-informed localization achieves substantial improvements in speed and accuracy, with a mean location error of 32 meters (SD 14) under stable ambient conditions, and a notable reduction in restoration times from an average of 95 to 70 minutes when DAS-assisted workflows are employed. The study anticipates that network topology, conduit fill, and ambient noise will influence signal-to-noise ratios, with urban canyons and higher temperature gradients correlating with marginally degraded localization precision. The integration framework is expected to demonstrate robust performance across scenarios, supported by a probabilistic fault map that assists operators in prioritizing repair actions and isolating faulted sections rapidly. The anticipated contribution to knowledge includes empirical evidence on the viability and performance gains of DAS-based fault localization in metropolitan distribution networks, a validated data fusion architecture that combines DAS and conventional protection signals, and an operational protocol for assimilating DAS outputs into control room decision-making. The study will extend theoretical understanding by applying Bayesian inference to multimodal fault signatures and by delineating the practical constraints and enablers of DAS deployment in dense urban environments. The primary conclusion is that DAS, when integrated with existing protection systems, substantially enhances fault localization accuracy and reduces restoration times, thereby improving grid reliability and customer service levels. Recommendations emphasize scalable DAS deployment in congested urban feeders, standardized data interfaces for protection- DAS interoperability, operator training programs, and further research into climate-related and architectural factors affecting DAS performance.