基于AL-BILSTMDN的输电线动态热极限概率预测  

Probabilistic Prediction of Dynamic Thermal Rating for Transmission Lines Based on AL-BILSTMDN

作　　者：孙辉[1] 卢雪立 高正男 胡姝博 金田 王钟辉 SUN Hui;LU Xueli;GAO Zhengnan;HU Shubo;JIN Tian;WANG Zhonghui(School of Electrical Engineering,Dalian University of Technology,Dalian 116024,China;Electric Power Research Institute,State Grid Liaoning Electric Power Co.,Ltd,Shenyang 110055,China;Dispatch&Communication Center,State Grid Liaoning Electric Power Co.,Ltd,Shenyang 110055,China)

机构地区：[1]大连理工大学电气工程学院,大连116024 [2]国网辽宁省电力有限公司电力科学研究院,沈阳110055 [3]国网辽宁省电力有限公司调度通信部,沈阳110055

出　　处：《电力系统及其自动化学报》2024年第6期110-118,共9页Proceedings of the CSU-EPSA

摘　　要：针对输电线动态热极限概率预测精度不足的问题,提出一种基于交替学习-双向长短时混合密度网络模型的动态热极限概率预测方法。该模型基于双向长短时混合密度网络抓取训练数据时序信息并实现动态热极限概率预测。同时,使用交替学习方法对时序数据集中的复杂模式数据进行强化学习,即区分出训练集中具有较为复杂模式的部分,让复杂模式训练集和全部训练集在模型中交替迭代直至最优,从而解决不平衡数据集混叠造成的局部最优问题。通过辽宁省某地区实例分析显示,所提模型可提升预测精度、降低过载概率。To address the problem of insufficient probabilistic prediction accuracy of dynamic thermal rating(DTR)for transmission lines,a DTR probabilistic prediction method based on an alternate learning-bidirectional long short-term mixture density network(AL-BILSTMDN)model is proposed.On the basis of BILSTMDN,this model captures the in-formation about the training data timing characteristics and achieves the DTR probabilistic prediction.At the same time,the AL method is used to enhance the learning of complex pattern data in which the timing data is concentrated,i.e.,the part of the training set with more complex patterns is distinguished,and this training set and the whole training set are alternately iterated in the model until an optimum is reached,so as to solve the local optimum problem caused by the overlapping of imbalance data sets.The analysis of an example in one region of Liaoning province shows that the proposed model can improve the prediction accuracy and reduce the overload probability.

关 键 词：线路输送能力 动态热极限 概率预测 动态增容 神经网络 

分 类 号：TM751[电气工程—电力系统及自动化]