Wind and PV Power Generation Forecasting Based on VMD-TCN-SA-BiLSTM

Wind and photovoltaic (PV) power generation exhibit significant instability and volatility, primarily due to their strong dependence on weather conditions. To improve the forecasting of wind and PV power generation and enhance the utilization of renewable energy in microgrid systems, this study proposes a novel forecasting model that integrates variational mode decomposition (VMD), a temporal convolutional network (TCN) with an embedded self-attention mechanism (SA), and a bidirectional long short-term memory (BiLSTM) network. First, Spearman's rank correlation analysis is applied to identify the weather features that significantly affect PV power forecasting, thereby constructing a multidimensional feature set. Then, the variational mode decomposition method is used to decompose the original wind and PV power data into sub-series with distinct frequencies. Next, the TCN-SA-BiLSTM hybrid neural network model is proposed. Finally, simulation experiments are conducted to predict wind and PV power. The experimental and simulation results show that the proposed VMD-TCN-SA-BiLSTM model behaves better forecasting accuracy compared with other models.