基于自适应模态分解和融合双尺度注意力机制的时间卷积网络的超短期风电功率预测  

Ultra-short-term wind power prediction based on adaptive mode decomposition and temporal convolutional network fusing double-scale attention mechanism

作　　者：谢智锋 张展 曾颖 许炫淙 于慧 孟安波[1] XIE Zhifeng;ZHANG Zhan;ZENG Ying;XU Xuancong;YU Hui;MENG Anbo(College of Automation,Guangdong University of Technology,Guangzhou 510006,China;Guangzhou Power Supply Bureau,Guangdong Power Grid Co.,Ltd.,Guangzhou 510620,China;Huizhou Power Supply Bureau,Guangdong Power Grid Co.,Ltd.,Huizhou 516000,China)

机构地区：[1]广东工业大学自动化学院,广州510006 [2]广东电网有限责任公司广州供电局,广州510620 [3]广东电网有限责任公司惠州供电局,广东惠州516000

出　　处：《黑龙江电力》2024年第6期478-485,490,共9页Heilongjiang Electric Power

摘　　要：针对风电功率强波动性限制预测精度的问题,提出一种基于自适应变分模态(adaptive variational mode decomposition,AVMD)和融合双尺度注意力(double-scale attention,DA)的时间卷积神经网络(temporal convolutional network,TCN)的超短期风电功率预测模型AVMD-DATCN。采用纵横交叉(crisscross optimization,CSO)算法对变分模态分解参数进行优化,提出动态混合熵(dynamic mixing entropy,DME)作为适应度函数以兼顾分解损失和分解子序列可预测性,将风电功率自适应分解为一系列稳定有序的子分量。针对各分量建立DATCN预测模型以充分挖掘潜在深层耦合非连续时序特征,将各分量预测值叠加重构得到最终预测结果。多角度对比实验结果表明,所提出模型的预测性能显著优于其他预测方法。To overcome the issue of limited prediction accuracy due to the strong volatility of wind power,an ultra-short-term wind power prediction model based on adaptive variational mode decomposition(AVMD)and the temporal convolutional network(TCN)integrated with the double-scale attention(DA),namely AVMD-DATCN,is proposed.The crisscross optimization algorithm is adopted to optimize the parameters of variational mode decomposition.The dynamic mixing entropy(DME)is proposed as the fitness function to take into account both decomposition loss and the predictability of the decomposed subsequences,and the wind power is adaptively decomposed into a series of stable and ordered sub-components.The DATCN prediction model is established for each component to fully exploit the potential deep-coupled and discontinuous time series features,the predicted values of each component are superimposed and reconstructed to obtain the final prediction result.The results of multi-angle comparative experiments prove that the prediction performance of the proposed model is significantly better than other prediction methods.

关 键 词：超短期风电功率预测 变分模态分解 纵横交叉算法 动态混合熵 双尺度注意力 时间卷积网络 

分 类 号：TM614[电气工程—电力系统及自动化]