Skip to main content


Table 1 State-of-the-art: Anomaly detection or outlier rejection

From: Energy forecasting based on predictive data mining techniques in smart energy grids

Author Problem Targeted Method applied Contribution & Perspective
Panapakidis (Panapakidis et al. 2018) Missing data treatment Data processing (Clustering phase + completion phase)
− Clustering phase is unsupervised machine learning tool K-means
− Completion phase filling technique application
− Proposed a new methodology for data filling
− Applicable for both complete and partial absence of data
− Presents a novel methodology for missing and incomplete data completion
− Methodology is not dependent on data size, data resolution and amount of missing data
− Incomplete data artificially completed with data entries of high similarity
Daliento (Daliento et al. 2017) Monitoring & diagnosis of faults in single and multiple PV array strings Monitoring and diagnosis techniques based data mining i.e. decision tree method, K-nearest neighbour and SVM − Reviewed methods for fault detection
− Presented reliability issues