A Bayesian Hierarchical Model to create synthetic Power Distribution Systems

Abstract

The growing complexity of Power Distribution Systems, driven by distributed generation, renewable energy integration, and increasing demand, has led to restricted access to DS data due to security and privacy concerns. This study addresses limited data accessibility by proposing a hybrid approach for crafting synthetic power distribution systems tailored for power system analysis and control. Synthetic power distribution systems refer to artificially generated models that faithfully replicate real-world DS features while upholding security and privacy constraints. This innovative methodology merges a Bayesian Hierarchical Model with Markov Chain Monte Carlo techniques, utilizing georeferenced data to capture intricate system dependencies, feeder configurations, switch statuses, and load node distributions. Leveraging OpenStreetMaps for DS topology, the approach incorporates expert knowledge and real-world data. Results highlight the methodology's ability to evaluate credible intervals for parameters, facilitating a probabilistic assessment of uncertainties and enhancing decision support in power system analysis and control. Findings affirm the hybrid approach's efficacy in generating realistic synthetic DSs, bridging the gap between statistical and georeferenced methodologies for advanced power system analysis and control. The capacity to generate synthetic DSs provides valuable insights into power system dynamics, addressing security, privacy, and data accessibility concerns for a more informed decision-making process.