摘要
首先提出了一种由熵诱导出的度量(称为CIM),这种度量具有非凸性,鲁棒性,平滑性,有界性,逼近行为等。将该度量用作人工神经网络(称为ANN)的损失函数,以提高其鲁棒性,然后建立新的鲁棒人工神经网络框架(称为指数损失人工神经网络,即ELANN)以减少噪声和异常值的影响。ELANN继承了ANN的优势,并提高了ANN在解决回归问题中的预测性能。利用广东电网2016—2018年的实际负荷数据进行仿真分析,结果表明该方法可以有效提高负荷预测的准确性和可靠性,为电力调度部门提供可靠的决策依据。
This paper first proposes a measure induced by correntropy(CIM),which has the characteristics of non-convexity,robustness,smoothness,boundedness,approximation behavior,etc.This metric is used as the loss function of the artificial neural networks(ANN)to improve its robustness and then a new robust artificial neural networks framework,exponent loss artificial neural networks(ELANN),is built to reduce the noise and the outliers.The ELANN inherites the advantages of the ANN and improves the predictive performance of the ANN in solving regression problems.The simulation analysis is performed by using the actual load data of Guangdong Power Grid from 2016 to 2018.The results show that this method can effectively improve the accuracy and reliability of load forecasting,and provide a reliable decision basis for the power dispatching departments.
作者
蔡秋娜
潮铸
苏炳洪
王龙
段秦尉
温亚坤
李冰
CAI Qiuna;CHAO Zhu;SU Binghong;WANG Long;DUAN Qinwei;WEN Yakun;LI Bing(Power Dispatching Control Center of Guangdong Power Grid Co.,Ltd.,Guangzhou 510600,Guangdong Province,China;Beijing Tsintergy Technology Co.,Ltd.,Haidian District,Beijing 100080,China)
出处
《电网技术》
EI
CSCD
北大核心
2020年第11期4132-4139,共8页
Power System Technology
关键词
信息熵
指数损失
鲁棒性
人工神经网络
短期负荷预测
information entropy
exponential loss
robustness
artificial neural network
short-term load forecast