Graph Data Driven Power Flow Model for Offshore Wind Farm Considering Internal and External Characteristics

Abstract

Data-driven power flow modeling method for offshore wind farm (OWF) has attracted much attention because it does not depend on system parameters and has fast response time. However, the existing method is only established for the fixed external power grid scenario, and does not considering the imbalance of power flow samples, resulting in limited fitting accuracy and application scenarios. Therefore, in this paper, the grid impedance information hidden in the historical power flow data (PFD) is further extracted, and constitutes a new PFD set considering internal and external characteristics. And then, by virtue of the advantages of graph depth learning and the structural characteristics of OWFs, a multi-head graph attention network based on Huber loss function is proposed to fit the PFD, effectively avoiding the problem of limit fitting accuracy caused by the imbalance of PFD. Based on Simulink-Python joint simulation platform, the physical-digital model of 400MW OWF is established for experimental verification, and the results show that the proposed method can accurately predict the power flow information under the change of internal - external characteristics and unbalanced PFD samples.

Keywords: Offshore wind farm, Data driven, power flow model, Power flow data set, Grid impedance, Graph attention network