Deep Clustering and Hybrid Neural Networks for Enhanced Ultra-Short-Term PV Power Forecasting

Accurate ultra-short-term photovoltaic (PV) power forecasting is essential for optimizing power system management and maintaining grid stability. This study proposes a novel forecasting framework that integrates the K-DAE clustering method, NRBO, TCN, BiGRU, and Multi-Head Attention. The model effectively captures dynamic PV power characteristics under varying weather and seasonal conditions. Experimental results show that the proposed model significantly reduces prediction errors and improves robustness, with MAPE, RMSE, and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> metrics outperforming CNN-LSTM, LSTM, and TCN by notable margins. This work demonstrates the potential of advanced machine learning techniques in enhancing the accuracy and reliability of PV power forecasting.