基于人工神经网络短期负荷预测的数据处理.pdfVIP

27 3 Vol.27 No. 3 2008 6 Journal of Nanchang Institute of Technology Jun.2008 : 1674- 0076( 2008) 03- 0045- 05 1 1 2 3 章顺华 , 连晶晶, 丁建平, 刘 菁 ( 1. , 330099; 2. , 330052; 3. , 330039) :科学合理的数据处理是提高短期电力负荷预测精度的最基本环节之一. 利用软 滤波方法, 自动平滑坏负 荷数据; 同时根据负荷的不同特性和规律, 将输入数据进行有效分组, 分别建立分组负荷预测模型: 工作日负荷预 测模型周日负荷预测模型以及节日负荷预测模型, 使预测模型不但具有所需样本数据少模型简单精度高等优 点, 同时又具有较强的泛化能力, 从而提高负荷预测的效率和精度. 通过对南昌供电公司的负荷数据进行具体计 算, 表明该方法是有效和可行的. :人工神经网络; 短期负荷预测;软 滤波; 数据处理; 预测精度 :TM714 :A Data processing technology for shortterm load forecasting based on artificial neutral networks 1 1 2 3 ZHANG Shunhua , IAN Jingjing ,DING Jianping , IU Jing ( 1. Department of Electrical and Electronic Engineering,Nanchang Institute of Technology, Nanchang 330099, China; 2.Jiangxi Stateowned Medicine Co. , td. ,Nanchang 330053, China; 3.Jiangxi CPI Power Engineering Co. , td. , Nanchang 330053, China) Abstract : Processing data scientifically and rationally is necessary to improve the load forecasting precision. In this paper the software filter method is applied to smoothing the bad load data.According to the different prop erties of the load, different forecasting models have been established respectively including workday model (Monday to Friday) , off days model ( Saturday, Sunday) and festival model.These forecasting models have such merits as simplicity and accuracy, and stronger generalization ability.The practical calculation results of daily power load data from the Nanchang power supply bureau show tha