An Incremental Clustering Algorithm with Pattern Drift Detection for IoT-Enabled Smart Grid System.

Area of Science:

• Electrical Engineering
• Computer Science
• Data Science

Background:

• Smart grids utilize IoT-enabled smart meters to collect electricity consumption data.
• Analyzing load patterns from this data is crucial for understanding consumer behavior and enabling demand response.
• Dynamic changes in consumer behavior necessitate frequent updates to load patterns for accurate segmentation.

Purpose of the Study:

• To propose a novel incremental clustering algorithm, ICluster-PS, for efficient load pattern updates in smart grids.
• To reduce computational scale and training time compared to traditional clustering methods.
• To enable continuous updating of load patterns with new incoming data.

Main Methods:

• ICluster-PS employs a probability strategy for incremental clustering, avoiding complete data re-clustering.
• It extracts new load patterns, integrates them with existing ones, and optimizes the combined set.
• The algorithm incorporates parameter updating and generalization for continuous operation.

Main Results:

• Experiments on a real-world dataset demonstrate ICluster-PS's superior performance over existing incremental clustering algorithms.
• Clustering validity indices and accuracy measures confirm the effectiveness of ICluster-PS.
• Case studies show ICluster-PS can effectively track pattern drifts and evolution.

Conclusions:

• ICluster-PS offers an efficient and accurate solution for updating load patterns in IoT-enabled smart grids.
• The algorithm facilitates precise consumer segmentation and effective demand response strategies.
• Its incremental nature and ability to detect pattern drifts make it suitable for dynamic energy management systems.