【摘要】 提出了一种基于谱分析法进行建模的短期负荷预测方案,该方案利用负荷历史数据的谱分析结果进行人工神经网络(ANN)模式分类和选择输入变量。方案采用快速傅立叶变换(FFT)进行负荷数据预处理,运用滤波算法及小时负荷曲线的频谱分析来研究电网负荷的周期特性,所得结果表明四季负荷的谱特性具有明显差异,应采用不同的模型和方案进行预测。谱分析有助于各时段预测方案提取输入变量。利用该思路构造的基于人工神经网络的负荷预测方案被用于预测广东省网的负荷,与其他普遍采用的输入变量预测结果的对比表明,所提方案在短期负荷预测中的性能良好。更多还原

【Abstract】 A short-term load forecasting method based onthe modeling by spectrum analysis is put forward in which thespectrum analysis results of historical load data is used for theclassification of artificial neural network (ANN) modes and theselection of input variables. In this method the load data ispreprocessed by fast Fourier transform (FFT), the filtrationalgorithm and the spectrum analysis of hourly load curve areused to research the periodical characteristics of power load.The obtained results show that there are obvious differencesamong the spectrum characteristics of the loads in four seasons,therefore different models and method should be used toforecast the load in different seasons. The spectrum analysis ishelpful to the extraction of the input variables of theforecasting scheme for different time intervals. According toabove mentioned idea an ANN based load forecast scheme isapplied in the load forecasting of Guangdong province.Comparing with the load forecasting results by other widelyused input variables, it is shown that the presented schemepossesses good performance for short-term load forecasting.更多还原