Context-Referenced Telemetry Data for Distribution Utilities: Quality Assurance/Quality Control by Lateral Sensors

摘要

The notion of enhancing resiliency for electrical grids has become a priority for engineers and researchers within the past few years. Unforeseen natural disasters (e.g., lightning strikes, geomagnetic storms, floods, etc.) can cause devastating damage to electrical grid infrastructures. While disasters may strike with no warning, prototypical weather events can indeed be forecast. However, anticipating and quantifying the impact of weather events is a challenging task due to its stochasticity. In this paper, a weather monitoring system paradigm, as part of a lateral sensor system, is proposed. Lateral sensors for the electrical grid, such as by way of a hyper-locale set of weather sensors equipped with edge analytics and artificial intelligence, provide incredible insight, via various parameters, such as air temperature, barometric pressure, humidity, precipitation, solar radiation, and wind. These lateral sensor parameters can provide indicators regarding impending storms, which could impact power lines (e.g., via lightning strikes, downed trees, etc.) and cause communications interference. Spider radar plots concurrently reflecting both weather sensor data and grid sensor data have proven useful, as weather data can serve to provide contextual reference for the associated grid sensor telemetry data. Moreover, this involved lateral sensor utilizes a deep learning module, which is based upon a Generative Adversarial [Neural] Network (GAN). The results of this study demonstrate that the implementation of lateral sensors based upon a deep learning module can result in enhanced contextual awareness.