电业超短期负荷预测仿真研究

Simulation and Research on Ultra-Short-Term Load Forecasting of Electricity Industry

超短期负荷预测是电力负荷预测中很重要的环节,数据的提取和预测方法的选择则是最难的环节。为了准确预测,采用电行业监测分析系统数据提取平台,能够实时监测及提取负荷数据。反向传播(BP)神经网络和极限学习机(ELM)具有预测能力强和全局搜索显著特点而成为超短期负荷预测中常用的两种方法。实验数据通过重点用电行业监测分析系统获取,通过建立预测数据提取模型,用BP神经网络和ELM通过不同的隐含层节点数设置进行超短期预测。实验结果表明,提出的数据提取平台在提取数据上的可靠性,同时BP神经网络及ELM在超短期负荷预测中的可行性,并且相比较于BP神经网络,ELM在超短期负荷预测上具有较高的预测精度和较短的运算时间。

Ultra-Short-Term Load Forecasting is a very important part in power load forecast,and collecting data and the methods for forecasting are the most important. Monitoring and analysis system for the key electricity industry,which can monitor and extract the data in real time,is a new platform for extracting data. Back Propagation（ BP） Neural Network and Extreme Learning Machine（ ELM）,which have notable features of a good ability of forecast and global search ability,and have been two common methods in Ultra-Short-Term Load Forecasting. The tested data were obtained from the monitoring and analysis system for the key electricity industry,and an extraction model for forecasting data was built. Then the Ultra-Short-Term Load Forecasting was conducted through different hidden layer nodes setting under algorithms of BP Neural Network and ELM. The result shows that the data came from the platform are reliable,and BP Neural Network and ELM are available in Ultra- Short- Term Load Forecasting,and ELM has higher prediction accuracy and less computing time for USTLF compared with BP Neural Network.