USE CASE APPLYING MACHINE-LEARNING TECHNIQUES FOR IMPROVING OPERATION OF THE DISTRIBUTION NETWORK

Abstract

This paper discusses the use of machine learning (ML) techniques to improve fault handling in distribution networks. The paper includes a short survey on the use of ML techniques in fault handling and shows that little published work has been done on using weather data and smart metering data as data sources. It can be argued that this is desired to increase the performance and usability of ML in operational support systems. Previous work also focuses almost exclusively on statistical machine learning aiming to replace traditional simulation models, overlooking other ML methods which can support operations. Here it is illustrated that Case based reasoning (CBR) can be used to aid the decision-making for example, when trying to restore service after an outage. The paper also describes the use of experience databases to aid the operator during fault handling. To illustrate potential use of ML and CBR, the paper presents a use case for future fault handling in low voltage distribution network and discusses the usefulness of this approach. This example shows that implementation of ML techniques in daily operation can be expected to contribute to reduction of costs for the network companies and increased security of supply for the customers.