Fig. 1: The proposed model structure.
From: Data-driven dynamic modeling for inverter-based resources using neural networks
In this schematic, the top panel shows the structure applied to different inverter-based resources (IBRs), including a wind farm, a photovoltaic (PV) power station, and a battery energy storage station (BESS). In this schematic, t is the time step, h and c are the hidden and cell states, x and y are the input and output variable vectors, respectively. V and θ denote the voltage magnitude and phase angle. The active and reactive currents are given by Ip = P/V and Iq = Q/V, where P and Q are the active and reactive power. Φ denotes the other factors, such as environmental information (wind speed, solar irradiance, etc.) and power commands from the control center. Neural networks (including long short-term memory (LSTM), cross-layer, and fully connected layer (FC)) are designed to represent the dynamics beyond inverters. The inverter model simulates the behavior of the inverter and determines the outputs (Ip and Iq) that represent the IBR's interface with the grid.
