Synthetic Agricultural Load Data Generation Using TimeGANs

In power system applications, accessing real data is one of the main challenges. Load modeling is an example that can be hampered by a lack of real data. This paper proposes synthetic data generation as a solution to deal with insufficient datasets. To this end, Timeseries Generative Adversarial Networks (TimeGANs) are proposed to create synthetic electricity data for an agricultural load. Center pivot irrigation system data is targeted, which does not exhibit the same trends as residential or commercial utility data, and information is presented in the lower dimensions. Obtained results show that TimeGANs can leverage the lower dimensional information to produce synthetic data with the same characteristics as the actual data. The proposed method can be utilized for modeling different electric loads.