Detection and Forecasting of Micro-Scale Variability in Electric Power Profiles
Abstract: The development of advanced metering infrastructures (AMI) for low-voltage (LV) networks enables the fine-grained collection of rich and diverse datasets for in situ power quality assessment. Resulting datasets, collected at high reporting rates, support detection and labelling of micro-scale events that are affecting the correct operation of LV networks and have been so far overseen through window-based averaging using typical approaches and measurement equipment. The tutorial focuses on methods and techniques to first detect and label such events as anomalies in a data processing and learning pipeline. Subsequently, the labeled datasets are used in a forecasting framework as early-warning system for potential imbalances in the local energy network. One key novelty is the combination of extracted features using time series data mining methods, such as the matrix profile, with state-of-the-art machine learning algorithms, including automated machine learning to optimize classification metrics in real time, across various model/algorithm structures and hyper-parametrisation options. Hands-on activities will highlight the practical use of the Python matrixprofile, scikit-learn and auto-sklearn open-source packages on publicly available residential power measurements collected in the context of two active research projects.