A Review of Transfer Learning Approaches for Load, Solar and Wind Power Predictions

With the increasing penetration of renewables, the operation of power system is facing significant challenges. To ensure a safe and reliable operation of power system, it is necessary to precisely predict the power consumption of load and also the power generation of solar and wind power. Owing to the advantages of quickly extract abstract complex features and process vast quantities of data, AI techniques have been widely adopted to predict load, solar and wind power. In particular, the transfer learning has emerged as a promising solution when there is not sufficient data available for training. In this paper, we aim to give a brief introduction to the transfer learning technique employed for load, solar and wind power predictions, from the perspective of working principle, transferability and classification. The advantages and disadvantages of each transfer learning technique are discussed. The motivation, state-of-the-art, and future directions are also thoroughly discussed.