基于ISSA-BiGRU-Attention的短期电力负荷预测研究  

Short-term Load Forecasting Based on ISSA-BiGRU-Attention

作　　者：沈梦燕 韦文山[1] 荣新 Shen Mengyan;Wei Wenshan;Rong Xin(School of Electronic Information,Guangxi Minzu University,Nanning 530006)

机构地区：[1]广西民族大学电子信息学院,南宁530006

出　　处：《现代计算机》2022年第20期61-64,共4页Modern Computer

摘　　要：为了提升短期电力负荷预测的精准度,提出了基于改进麻雀搜索算法(ISSA)优化的引入注意力机制(Attention)的双向门控循环单元(BiGRU)的预测模型ISSA-BiGRU-Attention。麻雀搜索算法的改进过程首先由引入非线性余弦调整因子和柯西变异算子改进的正余弦算法更新发现者位置;其次通过结合Levy飞行和T分布扰动算法更新加入者位置;最后引入精英反向学习策略赋予算法迭代时更好的寻优性能。实验结果表明,ISSA-BiGRU-Attention模型较其他方法预测效果更好,预测结果更稳定。In order to improve the accuracy of short-term power load forecasting the prediction model(ISSA-BiGRUAttention)was proposed based on the improved sparrow search algorithm and attention mechanism optimize bidirectional gate recurrent unit. The improvement process of sparrow search algorithm was in these ways. Firstly, the improved sines and cosines algorithm by introducing the nonlinear cosine adjustment factor and Cauchy mutation operator was used to update the discoverer position. Secondly, the levy flight and T-distribution perturbation algorithm was combined to update the position of the entrants;finally,elite opposition-based learning strategy is introduced to give the algorithm better performance. Experimental results show that ISSA-BiGRU-attention model has better prediction effect than other methods in this paper, and the prediction results are more stable.

关 键 词：改进麻雀搜索算法 BiGRU 注意力机制 短期电力负荷预测 精英反向学习 

分 类 号：TM715[电气工程—电力系统及自动化] TP18[自动化与计算机技术—控制理论与控制工程]