Multivariate Statistics and Supervised Learning for Predictive Detection of Unintentional Islanding in Grid-Tied Solar PV Systems

Abstract

Integration of solar photovoltaic (PV) generation with power distribution networks leads to many operational challenges and complexities. Unintentional islanding is one of them which is of rising concern given the steady increase in grid-connected PV power. This paper builds up on an exploratory study of unintentional islanding on a modeled radial feeder having large PV penetration. Dynamic simulations, also run in real time, resulted in exploration of unique potential causes of creation of accidental islands. The resulting voltage and current data underwent dimensionality reduction using principal component analysis (PCA) which formed the basis for the application ofQstatistic control charts for detecting the anomalous currents that could island the system. For reducing the false alarm rate of anomaly detection, Kullback-Leibler (K-L) divergence was applied on the principal component projections which concluded thatQstatistic based approach alone is not reliable for detection of the symptoms liable to cause unintentional islanding. The obtained data was labeled and aK-nearest neighbor (K-NN) binomial classifier was then trained for identification and classification of potential islanding precursors from other power system transients. The three-phase short-circuit fault case was successfully identified as statistically different from islanding symptoms.

Computational geometry‐based methodology for identification of potential islanding initiators in high solar PV penetration distribution feeders

“…The authors had previously used the principal components (PCs) analysis (PCA) on a dataset consisting of both positive and negative half-cycles of instantaneous voltage and current with projection onto the first PC as the single training feature. Although the PC transformation matrix contained mostly ones (with both signs) and zeros, it contained the information of both current and voltage samples and gave good classification results for offline testing as reported in [18]. One important observation was the ease and convenience of matrix-transformation-based methods for extracting features in real-time data transfer.…”

Wide area monitoring, protection, and control application in islanding detection for grid integrated distributed generation: A review

“…Supervised learning techniques or induction models deciphers patterns by connecting and evaluating relationships between properly labeled data, its variables, and the known outcomes. 71,72 Unlike supervised learning where the algorithm exploits known information and the expected outcome is known, unsupervised learning is the total opposite. The learning process is performed using unlabeled feature vectors, unsupervised learning algorithms decipher underlying patterns from the input data, and creates labels by grouping similar patterns together.…”