基于二次分解和IDBO-DABiLSTM的短期风电功率预测模型

Short-Term Wind Power Prediction Model Based on Secondary Decomposition and IDBO-DABiLSTM

为提高风电功率预测精度,针对风电的强波动性和高随机性,提出一种基于二次分解和改进蜣螂优化算法(IDBO)-双重注意力双向长短期记忆(DABiLSTM)网络的风电功率预测模型。首先,采用自适应噪声的完全集合经验模态分解(CEEMDAN)和小波包分解(WPD)构成一种二次分解方法对历史风电功率和风速数据进行分解,降低初始序列的随机性和非平稳性。其次,在BiLSTM网络的基础上,加入特征和时间注意力机制,建立DABiLSTM模型,充分挖掘特征间的关联性和时间序列间的长时间依赖性。最后,采用黄金正弦算法来优化滚球蜣螂的位置,从而增强算法在局部和全局的探索能力,同时引入动态权重系数改进偷窃蜣螂的位置,以平衡算法在全局和局部的探索能力,提出IDBO,并用其优化DABiLSTM网络的超参数,防止网络陷入局部最优解。采用贵州某风电场实际数据对所提模型进行实验,结果表明该方法能有效提升模型的预测能力,所提出的模型的均方根误差(RMSE)和平均绝对误差(MAE)在单步预测下分别为0.0449和0.0312 MW,与其他模型相比,分别平均降低了36.9%和31.7%,表现出较好的预测精度和鲁棒性。

To improve the accuracy of wind power prediction,a new wind power prediction model based on secondary decomposition and Improved Dung Beetle Optimizer(IDBO)-Dual Attention Bidirectional Long Short-Term Memory(DABiLSTM)network is proposed for the high stochasticity and strong volatility of wind power.First,a secondary decomposition method based on the Complementary Ensemble Empirical Mode Decomposition of Adaptive Noise(CEEMDAN)and Wavelet Packet Decomposition(WPD)is proposed to disaggregate the raw wind power historical data and windspeed historical data,thereby reducing the randomness and volatility of the raw signals.Second,a DABiLSTM network model is established by incorporating feature and time attention mechanisms.This model fully explores the correlation between features and long-term dependency between time series,thereby improving the accuracy of wind power prediction.Finally,IDBO is proposed based on the golden sine algorithm to improve the rolling ball dung beetle position and enhance the local development and global exploration capabilities of the algorithm.In addition,a dynamic weighting factor is incorporated to improve the stealing cockroach position and balance the global exploration and local development capabilities of the algorithm.IDBO is employed for intelligent optimization of the network working hyperparameters of the DABiLSTM model to further enhance the prediction precision of the model.The proposed model is tested using real data from a Guizhou wind farm,and the findings demonstrate that the proposed approach can successfully increase the predictive power of the model.The Root Mean Square Error(RMSE)and Mean Absolute Error(MAE)values of the proposed model are 0.0449 and 0.0312,respectively,in single-step prediction,which is a reduction by 36.9%and 31.7%on average,respectively,compared with other models in the literature,thus showing better prediction accuracy and robustness.