Load profiling and Monte Carlo simulation for load variety and variability in voltage optimization

Voltage optimization has increasingly turned to the demand side for greater flexibility, a pursuit complicated by variety and variability of the loads. To address these challenges and maximize demand-side potential, a load profiling method embedded in a Monte Carlo framework is proposed in this study. The variety of loads is captured by load profiling that delineates the power system’s operational boundaries by identifying typical consumption patterns, which is achieved via a novel clustering technique that uniquely combines supervised and unsupervised learning. Unlike existing combinations of the two learning algorithms that use unsupervised learning to set the classes and supervised learning to fill in them in two separate steps, the newly developed clustering integrates both unsupervised and supervised learning exclusively for clustering. The variability of loads is represented by the active – reactive load curves, sampled by the Monte Carlo simulation to create multiple scenarios for the coordinated dispatch of active and reactive powers. This multi-scenario voltage optimization, enabled by the new load profiling technique, aims to enhance a wide range of power system operation and planning applications, particularly voltage evaluation and reactive power planning, which are utilized here to demonstrate the effectiveness of the proposed method.