PF-FEDG: An open-source data generator for frequency disturbance event detection with deep-learning reference classifiers

Abstract

Accurate and fast classification of FDEs is crucial to power systems situation awareness and stability control. However, various database used in recent studies makes it hard to directly compare different classification methods. To relieve the lack of a standardized database that can be used as a benchmark, this work proposes an open-source package which generates frequency events based on PowerFactory named PF-FEDG. PF-FEDG can produce various labeled FDEs with random parameters such as generator trips, load disconnections, line outages, frequency ramp ups, frequency ramp downs, and frequency oscillations. Furthermore, the package includes three reference FDEs classifiers based on deep learning techniques. These models are tested on the IEEE 39-bus system and IEEE 118-bus system and achieve high performance. The package generating capability and the reference classifiers can be used by the community as benchmarks for comparison and development of new algorithms for FDEs detection. The code of PF-FEDG is available on GitLab.

Keywords

• Frequency disturbance event
• Public dataset
• Classification
• Deep-learning
• LSTM
• CNN