A Renewable Energy Generation Power Forecasting Method Based on Double Attention Mechanism-Long Short-Term Memory

Due to the intermittence and fluctuation of renewable energy generation, the accurate power forecasting of renewable energy generation is crucial for the stability and security of power grid operations. However, existing power forecasting methods are not sufficient for information extraction in historical data, which will reduce the power forecasting accuracy. Therefore, this paper proposes a renewable energy generation power forecasting method based on double attention mechanism-long short-term memory (LSTM), which adopts the self-attention mechanism to extract the key information of the time dimension and the attention mechanism to extract the key information of the variable dimension. By fusing the extracted time and variable dimension features with the current moment hidden layer vector, the input data features can be more fully extracted. Finally, the forecasting results of photovoltaic (PV) and wind power generation are given to verify the effectiveness of the proposed method.