An Advanced CNN-LSTM-BiLSTM Model Leveraging Attention Mechanisms for Accurate Distributed Photovoltaic Output Prediction

As the installed capacity of photovoltaic power generation continues to expand, its role in overall energy consumption becomes increasingly significant. Consequently, precise forecasting of photovoltaic output is critical for strategic development planning and efficient grid management. To address this, we present an innovative CNN-LSTM-BiLSTM model enhanced with an attention mechanism. Within this framework, the attention mechanism synergistically integrates with the Convolutional Neural Network (CNN) layer to engineer novel features from influential factors affecting photovoltaic output. These newly extracted features are subsequently fed into Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) layers for next-hour PV output prediction. Comparative analyses demonstrate that our proposed methodology outperforms existing models, including CNN-BiLSTM, CNN-LSTM, BiLSTM, and LSTM, in terms of predictive accuracy.