A Sequential Bayesian Approach to Online Power Quality Anomaly Segmentation

Increased observability of power distribution networks can reveal signs incipient faults which develop into costly and unexpected plant failures. While low-cost sensing communications infrastructure is facilitating this, it also highlighting the complex nature fault signals, a challenge entails precisely extracting anomalous regions from continuous data streams before classifying underlying signature. Doing this incorrectly will result in capture uninformative data. Extraction processes be confounded by operational noise on network including harmonics produced embedded generation. In article, an online model proposed. Our Bayesian Changepoint quality anomaly segmentation allows automated anomalies current waveforms, irrespective noise. Demonstration effectiveness proposed technique carried out with field as well challenging simulated network, ability to accommodate typical penetration levels electronic devices.