Data-Driven Hybrid Equivalent Dynamic Modeling of Multiple Photovoltaic Power Stations Based on Ensemble Gated Recurrent Unit

Abstract

With the increasing penetration of the photovoltaic (PV) in the distributed grid networks, the dynamic response analysis of the system becomes more and more complex and costs lots computational time in the simulation. To cut down the computational resources while maintaining the accuracy, this paper proposes a data-driven hybrid equivalent model for the dynamic response process of the multiple PV power stations. The data-driven hybrid equivalent model includes the simple equivalent model and data-driven error correction model. In the equivalent model, all the PV arrays in the central PV power station are simplified into one equivalent model based on the related parameter equivalence. The distributed PV power stations in the same branch are equivalent to one power station model based on the parameter equivalence and feeder equivalence. Since the difference between the simple equivalent dynamic response model and precise one exists, the data-driven error correction model is proposed. The ensemble Gated Recurrent Unit (GRU) model utilizes the bagging ensemble structure and uses the simple equivalent dynamic response as input to learn the accurate dynamic response. The simulation results validated the super-performance of the proposed model both in the response speed and accuracy.