User Identification in Low-Voltage Distribution Networks Using Support Vector Machine and Random Forest

Accurate identification of users within the low-voltage distribution area directly impacts grid loss calculations, fault detection, and electricity supply safety. With the rise of smart grid technology and the increasing complexity of power grids, this task has become more challenging. This study addresses these challenges by focusing on user identification based on simultaneous electricity consumption data, aiming to minimize the impact of time synchronization issues and data quality problems. The research employs Support Vector Machine (SVM) and Random Forest (RF) algorithms to classify users across multiple substations, enhancing generalization and adaptability to different user types. To tackle the issue of missing data, three interpolation methods-spline interpolation, piecewise cubic Hermite interpolation (PCHIP), and improved Makima interpolation-are tested. The study reveals that the Makima interpolation method consistently achieves the lowest average error across various missing data ratios. Validation examples demonstrate that using the Makima interpolation method results in high classification accuracy for both SVM and RF classifiers, with RF showing superior performance.