Clustering smart meter data for residential load profile analysis

Smart meters typically record half-hourly electricityusage, enabling detailed analysis of household consumption patterns.However, the data is often random and high-dimensional,making pattern detection difficult. K-means clustering is appliedto a real-world smart meter dataset from a London pilotprogram, with data aggregated by season and household tosimplify analysis. The challenges of clustering time series data and selecting the optimal number of clusters are discussed. Results show that k-means can effectively group households with similar usage patterns, supporting applications in demand-side management, grid balancing, and forecasting.