Test Scenario Generation for Distribution Grid Control

Abstract:
The increasing integration of renewable energy sources and electrification of buildings and mobility sectors presents new challenges for distribution grid operations. While developing advanced grid control strategies is crucial, obtaining real distribution grid data that captures various configurations and scenarios remains a significant barrier for both academia and industry. This paper introduces a novel tool for generating diverse distribution grid scenarios to facilitate the devel-opment and validation of advanced grid control strategies. The tool generates realistic time series data for loads, renewable generation, and EV charging profiles that can be applied to any distribution grid model, regardless of whether it is bal-anced or unbalanced, radial or meshed. It employs Generative Adversarial Networks (GANs) to create load and PV/wind power profiles while maintaining spatiotemporal correlations. For EV charging profiles, the tool utilizes probability dis-tributions derived from real data, considering different charging levels and days of week. Additionally, users can specify participation factors to control the deployment of renewable generation and EV chargers across the network. This com-prehensive approach enables the creation of numerous realistic scenarios, supporting the development and testing of grid control strategies and machine learning applications in distribution network management. This tool has been successfully tested in the development congestion management algorithms for distribution grids.