Power Forecasting of Regional Wind Farms via Variational Auto-Encoder and Deep Hybrid Transfer Learning

• DOI: 10.3390/electronics11020206
• Publication Date: 2022-01-10T22:42:25Z
• Authors: Mansoor Khan; Muhammad Rashid N...; Essam A. Al-Ammar; Wonsuk Ko; Hamsakutty Vettik...; Irfan Ahmad
• Keywords: Transfer of learning; Feature (linguistics)

Wind power is a sustainable green energy source. Power forecasting via deep learning essential due to diverse wind behavior and uncertainty in geological climatic conditions. However, the volatile, nonlinear intermittent of makes it difficult design reliable models. This paper introduces new approach using variational auto-encoding hybrid transfer forecast for large-scale regional windfarms. Transfer applied windfarm data collections boost model training. multiregional windfarms consist different weather conditions, which apply learning. Therefore, we propose method consisting two feature spaces; first was obtained from an already trained model, while second, small set current retraining. Finally, transferred neural networks were fine-tuned achieve precise forecasting. A comparison with other state-of-the-art approaches revealed that proposed outperforms previous techniques, achieving lower mean absolute error (MAE), i.e., between 0.010 0.044, lowest root square (RMSE), 0.085 0.159. The normalized MAE RMSE 0.020, accuracy losses less than 5%. overall performance showed offers maximum minimal error.