基于天牛须搜索算法的短期风电功率组合预测

为了提高风电功率预测精度,提出了一种完全集成经验模态分解(complete ensemble empirical mode decomposition adaptive noise, CEEMDAN)、极限学习机(extreme learning machine, ELM)和改进天牛须搜索算法(improved beetle antennae search algorithm, IBAS)的组合预测模型来预测风电功率。引入动态惯性权重改进天牛的位置更新方式,提高天牛须搜索算法的寻优能力。在预测过程中,首先通过CEEMDAN对原始风电功率数据进行预处理,将非平稳信号分解为一组按照频率和振幅大小排列的序列分量,减少数据波动带来的预测误差。然后利用IBAS优化ELM构建预测模型,分别预测每个序列分量,最后叠加每个序列分量的预测值得到最终预测值。仿真结果表明,与其他预测模型相比,本预测模型预测精度最高,评价指标平均绝对误差(mean absolute error, MAE)、均方根误差(root mean square error, RMSE)、平均绝对百分比误差(mean absolute ...

Short-term Wind Power Combination Prediction Based on CEEMDAN-IBAS-ELM

In order to improve the accuracy of wind power prediction, a combined prediction model of complete ensemble empirical mode decomposition (CEEMDAN), extreme learning machine (ELM) and improved beetle search algorithm (IBAS) is proposed to predict wind power. The dynamic inertia weight was introduced to improve the position update method of the beetle, and improve the optimization ability of the beetle search algorithm. In the prediction process, the original wind power data was preprocessed by CEEMDAN, and the non-stationary signal was decomposed into a set of sequence components arranged according to frequency and amplitude to reduce the prediction error caused by data fluctuations. Then IBAS was used to optimize ELM to build a prediction model, and each sequence component was predicted respectively, and finally, the predicted value of each sequence component was superimposed to get the final predicted value. The simulation results show that, compared with other prediction models, this prediction model has highest prediction accuracy, and the evaluation indicators MAE, RMSE, and MAPE are all the minimum, which has broad practical application prospects.