电力调度交换机短期负荷智能预测方法研究  

Intelligent forecasting method for short term load of power dispatching switches

作　　者：钱毅慧 陆萍 黄蓓雯 QIAN Yi-hui;LU Ping;HUANG Bei-wen(State Grid Shanghai Customer Service Center,Shanghai 200030,China)

机构地区：[1]国网上海客服中心

出　　处：《电子设计工程》2019年第12期129-132,137,共5页Electronic Design Engineering

摘　　要：针对电力调度交换机短期负荷预测,采用传统方法受到外界因素影响,导致预测效率较低,为了解决该问题,提出了基于神经网络短期负荷智能预测。综合时间和气温数据,设计训练样本选取流程。根据选取结果,分析神经网络接入方式,通过明确输入、输出和隐含层节点数量,对不同神经元进行训练,形成10-6-1的神经网络结构。在该结构下,计算一天某刻负荷均值和方差,并对温度和天气数据分别进行量化和编码处理,由此实现短期负荷智能预测。通过对比结果可知,该方法最高预测效率可达到98%,为电力调度交换机稳定运行提供数据支持。For short-term load forecasting of power dispatching exchanges,traditional methods are affected by external factors,resulting in low forecasting efficiency.To solve this problem,an intelligent short-term load forecasting method based on neural network is proposed.Integrating time and temperature data,the training sample selection process is designed.According to the selected results,the access mode of neural network is analyzed.By defining the number of input,output and hidden layer nodes,different neurons are trained to form a 10-6-1 neural network structure.Under this structure,the average and variance of load at acertain moment of a day are calculated,and the temperature and weather data are quantified and coded respectively,thus realizing intelligent short-term load forecasting.The results show that the maximum prediction efficiency of this method can reach 98%,which providesdata support for the stable operation of power dispatching switch.

关 键 词：电力调度交换机 短期负荷 智能预测 神经元 量化 编码 

分 类 号：TP715[自动化与计算机技术—检测技术与自动化装置]