基于JANET神经网络的短期负荷预测模型  

Short-term Load Prediction Model Based on JANET Neural Network

作　　者：许飞 武志刚[1] 张文倩 Xu Fei;Wu Zhigang;Zhang Wenqian(College of Electric Power, South China University of Technology, Guangzhou Guangdong 510640, China)

机构地区：[1]华南理工大学电力学院,广东广州510640

出　　处：《电气自动化》2021年第1期71-75,共5页Electrical Automation

基　　金：国家自然科学基金资助项目(5187070167)。

摘　　要：对长短时记忆网络结构进行了简化,提出了一种改进的基于长短时记忆网络(LSTM)的短期负荷预测方法,称为JANET(just another network)网络,以充分利用海量数据中所蕴藏的信息,从而在资源受限的实际场景中实现应用。模型保留了长短时记忆网络中最重要的遗忘门,在保证预测精度的同时,能够有效地缩短训练模型和预测的时间。使用所提出的方法对某电网历史负荷数据集进行预测试验,结果表明,简化模型与标准LSTM网络预测方法有相近的预测精度,在某些数据点甚至有更高的预测精度,同时在使用相同的框架和预测平台的前提下,JANET网络具有更快的运算速度,能够大大缩短负荷预测的计算时间,为电网提供及时有效的负荷预测数据。In this paper,the structure of the long-short-term memory(LSTM)network was simplified and an improved short-term load prediction method based on the long and short-term memory(LSTM)network,called JANET(Just Another Network)was proposed to make full use of information hidden in massive data so that application might become possible in actual sceneries with limited resources.The model retained the most important forgotten gate in the long and short-term memory network,and could effectively shorten the time needed for model training and prediction,while ensuring prediction accuracy.Prediction experiments were made on historical load data of a power grid in the proposed method.The results indicated that,in comparison with standard LSTM network prediction,the proposed simplified model had a similar prediction accuracy and even higher prediction accuracy at some data points.Furthermore,under the precondition that the same framework and prediction platform were used,JANET had a faster calculation speed and could substantially reduce the calculation time for load prediction and provide the power grid with timely and effective load prediction data.

关 键 词：短期负荷预测 LSTM神经网络 相关性分析 模型简化 数据预处理 

分 类 号：TM76[电气工程—电力系统及自动化]