Transient Waveform Clustering : Developing efficient data analytics toolchains applying unsupervised machine learning techniques on power quality events

Sammanfattning: High Voltage Direct Current (HVDC) transmission systems appropriate for bulk power transfer to meet increasing power demands and ideal for interconnecting power systems with distant renewable sources of energy without any chances of loss synchronism, efficiency, and reliability. The main obstacle is however connected with the DC grid protection where the timely diagnosis of faults is critical to prevent any rapid built-up leading to failure of the power electronic devices. Monitoring the Power Quality (PQ) necessitates establishing novel criteria and techniques to deal with the abundance of data that are ever-growing with data flow from sensors and measuring units in the electric grid. This study developed a scalable and efficient clustering methodology for a transient waveform database from a HVDC station. The output could help HVDC Service better characterize the data and develop qualitative criteria for monitoring and analytics. The thesis expects to contribute towards a sustainable and reliable electric grid.