Prediction of voltage distribution using deep learning and identified key smart meter locations

The energy landscape for the Low-Voltage (LV) networks is undergoing rapid changes. These changes are driven by increased penetration of distributed Low Carbon Technologies, both on generation side (i.e. adoption micro-renewables) and demand electric vehicle charging). previously passive ‘fit-and-forget’ approach to LV network management becoming increasing inefficient ensure its effective operation. A more agile operation planning needed, that includes pro-active prediction mitigation risks local sub-networks (such as risk voltage deviations out legal limits). mass rollout smart meters (SMs) advances in metering infrastructure holds promise smarter management. However, many proposed methods require full observability, yet expectation being able collect complete, error free data from every meter unrealistic operational reality. Furthermore, (SM) roll-out has encountered significant issues, with current voluntary nature installation UK other countries resulting low-likelihood SM coverage all networks. Even a comprehensive privacy restrictions, constrain availability meters. To address these this paper proposes use Deep Learning Neural Network architecture predict distribution partial actual operator circuits. results show measurements key locations sufficient distribution, even without high granularity personal power individual customers.