基于组合式神经网络的短期电力负荷预测模型

通过对电力负荷变化规律和影响因素的分析，提出了一种基于组合式神经网络的短期电力负荷预测模型。该模型综合运用神经网络、模糊聚类分析和模式识别理论方法进行建模。首先，采用模糊聚类分析方法，以每天的24点负荷数据、天气数据以及天类别数据为指标，将历史数据分成若干类别；其次，对每一类别建立相应的神经网络预测模型；预测时通过模式识别，批出与预测天相符的预测类别，利用相应的神经网络预测模型进行24小时的短期电力负荷预测。对绍兴地区2年多的实际负荷变化数据进行预测分析的结果表明，该模型不仅对普通工作日有较高的预测精度，对双休日、节假日和一些特殊情况也有较好的预测精度。

SHORT-TERM LOAD FORECASTING WITH MODULAR NEURAL NETWORKS

A novel short-term load forecasting model based on modular neural networks is presented in this paper. The model employs fuzzy clustering analysis, pattern recognizing and neural networks to forecast hourly loads for the next hour to 24 hours out. The practical historical data within one year is divided into several groups by fuzzy clustering analysis. Each group is modeled by a separate module based on neural networks. During the forecasting phase, pattern recognizing is employed to activate the corresponding module for hourly loads forecasting. Using data from the Shaoxing utilities ,the satisfactory accurate results are obtained on the weekday , weekend and holidays . Moreover ,the model is robust ,and produces accurate results in some special cases.