基于改进CEEMDAN-TCN模型的风电功率预测研究

Research on wind power prediction based on improved CEEMDAN-TCN model

针对风力发电时间序列数据随机性大、单一算法难以获得准确预测结果的问题,本文采用改进的完全自适应噪声集合经验模态分解(ICEEMDAN)与时域卷积网络(TCN)相结合的模型预测风电功率。首先,针对集成模态分解的残余噪声和杂散模式问题,采用改进的CEEMDAN将原始序列数据分解为多个复杂度差异明显的子序列;其次,利用样本熵评估各分量复杂度,采用注意力机制的TCN,分别对低复杂度子序列和高复杂度子序列进行预测;最后将各子序列的预测结果叠加,得到最终的预测结果。经在弗兰德伦地区的数据集上测试结果表明,所提的ICEEMDAN-TCN模型的MAPE为1.74%,RMSE为127.36,优于其它对比模型,预测效果表现最优。

Aiming at the problem that wind power time series data is highly randomized and it is difficult for a single algorithm to obtain accurate prediction results,this paper uses an improved model combining fully adaptive noise ensemble empirical mode decomposition(ICEEMDAN)and time-domain convolutional network(TCN)to predict wind power power.Firstly,aiming at the residual noise and spurious mode problems of integrated modal decomposition,the improved CEEMDAN is used to decompose the original sequence data into multiple subsequences with obvious complexity differences.Secondly,the sample entropy is used to evaluate the complexity of each component,and the TCN of the attention mechanism is used to predict the low complexity subsequence and the high complexity subsequence respectively.Finally,the prediction results of each subseries are superimposed to obtain the final prediction results.The results of the dataset in Flandren show that the proposed ICEEMDAN-TCN model has a MAPE of 1.74%and an RMSE of 127.36,which is better than other comparison models and has the best prediction effect.