A systematic review of machine learning applications in the operation of smart distribution systems

Abstract

Due to climate changes happening in the past few years, the necessity for the integration of renewable energy sources and other low-carbon technologies is ever-growing. With the integration of low-carbon technologies, the power system is facing changes, that are particularly visible in distribution systems. A high share of distributed energy resources installed at the medium voltage or low-voltage level creates new challenges for Distribution System Operators. To overcome challenges, Distribution System Operators need to look beyond traditional methods and find new ones that will help in the mitigation of potential problems in the planning and operation of distribution networks. Recent research efforts have put the attention on using machine learning based algorithms and methods. Machine learning methods have shown great potential in the prediction of consumption and production, scheduling of flexibility services, near real-time operations, etc. To summarize their advantages, but also shortcomings, a comprehensive review of using machine learning based methods and algorithms in the planning and operation of smart, active distribution systems is provided. In addition to the already developed and presented applications, we identify the current research gap and proposed future research directions with machine learning applications.

Keywords

• Distribution network
• Distributed energy resource
• Flexibility
• Machine learning
• Planning
• Real-time operation