Short-term wind power prediction based on modal reconstruction and CNN-BiLSTM

Abstract

Accurate prediction of short-term wind power plays an important role in the safe operation and economic dispatch of the power grid. In response to the current single algorithm that cannot further improve the prediction accuracy, this study proposes a combined wind power prediction model based on data processing, signal decomposition, and deep learning. First, outliers in the original data can affect prediction accuracy. This study detects the outliers by the Z-score method and fills them with cubic spline interpolation to ensure the integrity of the data. Second, for the volatility of wind power, the wind power time series is decomposed using the complete ensemble empirical modal decomposition with adaptive noise (CEEMDAN). The component complexity of the decomposition is calculated using the sample entropy (SE), and the components are reconstructed according to the SE size to improve the prediction accuracy. Finally, the traditional convolutional neural network (CNN) structure is improved and the bi-directional long short-term memory (BiLSTM) network is used to further extract the feature links between wind power data and superimpose the prediction results of each reconstructed component to obtain the final wind power prediction value. The experimental results demonstrate that the hybrid model proposed in this study has better performance in terms of prediction performance.

Keywords

• Wind power prediction
• Complete ensemble empirical mode decomposition with adaptive noise
• Data processing
• Convolutional neural network
• Sample entropy
• Bidirectional long short-term memory