Application of machine learning techniques for asset management and proactive analysis in power systems

Abstract

Asset Management is one of the foremost vital chapters within the power system's operation and, in general, within energy systems. Electric utilities are a capital-intensive industry with assets such as power transformers, power lines, and switch gears spread across a large geographic area. This paper examines the business drivers, challenges, and innovations for maximizing power network reliability through Asset Management (AM). It presents the main features of an open-source software platform that can be used to evaluate indicators that guide the process of making decisions. This tool is being developed inside a European research project named ATTEST. The machine learning algorithms implemented in the tool for AM and described in the paper can assess indicators for evaluating asset health and prioritize preventive and proactive maintenance strategies. The article describes the tool's outcomes, including an overall health score and risk ranking.

Today, the generation, transmission, and distribution business of power systems are suffering essential and fast changes in their operative and management strategies worldwide. Some decades ago, the liberalization of the electricity markets introduced a new significant competition factor within the sector, followed by increasing demand for quality of service from customers and public administrations. More recently, the digitalization of power systems and new concepts about Smart Grids allow the collection of more and more information about the power system components. This enables recording data in a more orderly and reliable way that can be used for better operation and maintenance and, ultimately, better management of the available resources in power systems, the main idea of the Asset Management concept.