Neural Network Clustering on Large Dataset: Energy Community Case Study

As interest in Energy Communities (ECs) continues to grow, segmenting users who may be inclined to participate has gained significant importance. However, the highly heterogeneous nature of individual energy consumption patterns complicates the development of accurate assessments regarding the economic viability of EC membership. While neural network clustering methods, such as Self-Organizing Map (SOM), offer a valuable approach for discovering an undefined number of user groups, their practical application is hindered by the substantial volume of data required for effective analysis. Building upon previous work that involved analyzing a large dataset, the methodology was extended by incorporating a Self-Organizing Map (SOM) technique. This extension maintained the significant computational reduction already achieved. Additionally, a sensitivity analysis on input parameters was performed to enhance the approach. The process was finalized with the evaluation of performance metrics, which demonstrated high F1-score and accuracy.