<ul><li>Researchers in China developed a new hybrid deep learning model called CRAK for PV power prediction in scenarios with fluctuations, outperforming 10 existing models.</li><li>CRAK model integrates causal convolution, recurrent structures, attention mechanisms, and Kolmogorov–Arnold Network (KAN) to capture critical factors and characteristics of PV data.</li><li>The model uses convolution, recurrent (GRU and BiLSTM), attention mechanism, and KAN layer to predict power generation, achieving superior accuracy and stability.</li><li>Tested on real-world data from a PV power station in China, CRAK model showed exceptional performance with MAPE 0.024, RMSE 0.032, MAE 0.015, and R2 0.999, demonstrating remarkable accuracy and effectiveness.</li></ul>

Hybrid deep learning model for PV forecasting in scenarios with considerable fluctuations

Discover more