Unsupervised Analysis of Public Street Lighting Energy Consumption Using Self-Organizing Maps

Street public lighting is an important component of urban energy consumption and is being studied to improve its monitoring through increasingly efficient and accurate measurement systems. Consumption data can provide insights into the daily operation of infrastructure, but their high dimensionality and lack of labeled information pose a challenge for analysis. In this work, an unsupervised approach based on self-organizing maps is proposed, to analyze the quart-hour energy consumption profiles of public street lighting systems. To this aim, the neural network-clustering tool available in MATLAB is used, applied to a dataset presenting real and simulated measurement data. The results show that this approach groups similar daily profiles, supporting intuitive visual interpretation through SOM maps.