Advanced Edge Computing and Monitoring for High Impedance Fault Detection in Distribution Systems

Total situational awareness is imperative for a healthy and safe grid. As more distributive energy resources (DERs) are connected to grid, the grid becomes more decentralized, meaning generation can be more widespread than relying solely on large plants burning fossil fuels or using water to spin generators far away from loads. Photovoltaic inverters are one such DER. They convert direct current (dc) which is produced by solar panels into alternating current (ac) to transmit on the grid. However, their advent introduces a variety of issues that must be tackled as utility companies and homeowners begin implementing more in everyday use. Voltage regulation can be difficult during times of high solar generation. In addition, if faults occur, they must be detected and cleared as soon as possible. Inverters are becoming smarter and their features can help solve some of these issues. Edge data (i.e. from inverters) must swiftly and securely reach a centralized controller for system operators to effectively take care of the grid. This is done with Supervisory Control and Data Acquisition (SCADA) protocols as well as internet connectivity through MQTT. In addition, operators must be able to send commands to individual generators to control power generation during peaks and dips to ensure grid stability. Thus, a secure two-way communications system is critical to achieving issues related to introducing greener energy sources. Once information from the edge-side inverter reaches the cloud, machine learning algorithms can use them to infer potential faults and locate them. This report will dive into the details of why green DERs are being added to the grid, various SCADA protocols, the OSI model for internet connectivity, and present the lab work prepared, including modeling a real feeder in real time for communication testing and OpenDSS for fault studies. Two gateway devices are developed and implementation is extensively detailed. With voltage data from the inverters in the cloud, a number of machine learning algorithms are built and tested for high impedance fault detection on the feeder model. A summary of scientific contributions to the community is also given, including publications and presentations.