Uncertainty Quantification Using Real-world Data for NAERM Model Validation

Abstract

Several co-simulation models are being developed under the North American Energy Resilience Model (NAERM) project initiative. The co-simulations primarily simulate the Bulk Energy Systems (BES) model and other interdependent systems, such as the models of natural gas (NG) pipeline supplying gas to the power generators. The co-simulation results help in understanding the interdependencies, which in turn help to derive management policies to bolster the power grid's resilience. However, the predictions of these models must be validated against the real-world data for their practical adoption. To validate the co-simulation models, we need to validate each simulation model involved in the co-simulation separately before validating their interdependent behavior. Since these models are approximations of the physical systems they represent, their simulation predictions deviate from the actual physical system behavior. Uncertainty quantification(UQ) estimates the prediction deviations of the model from the actual system behavior. By considering only the model predictions within the pre-determined uncertainty limits as successful, UQ can be used to differentiate between the success and failure of a co-simulation validation exercise. For example, consider the NG and BES co-simulation validation scenario, where the real-world data on NG utilization by the generators of a BES system is available, and the real-world data corresponding to the BES system state is also available. Then, the power generated from the NG real-world data and the load profile from the BES system data can be used to determine the state of the BES system. If the deviation of the BES model's computed state falls within the specified uncertainty limits, we can claim success of the NG-BES co-simulation validation exercise. However, the proposed NG-BES validation method depends on the knowledge of uncertainty limits for the BES model, i.e., the success criteria for the BES validation. In this project, we use a state estimation data from a utility company, obtain the simulation results for a set of randomly chosen state estimation scenarios, and estimate the uncertainty. Further, with the knowledge that the NG model can interface with the Multi-regional Modeling Working Group's (MMWG) eastern interconnect planning model, we generate a use-case scenario that removes the buses and corresponding branches missing in the MMWG planning model from the real-world state-estimation scenario. Using the results from several of these scenarios, we determine uncertainty using a metric developed for this task. Our approach, methods, metrics, and corresponding results are detailed in this report.