基于局部相似与神经网络的超短期负荷预测方法

为提高超短期负荷预测精度,特别是负荷曲线在拐点处的精度,在模糊聚类分析的基础上提出了一种选取局部相似日的改进模型。通过该模型选取局部相似日,并结合RBF网络提出一种新的预测超短期负荷方法。采用该方法建立2个预测模型,分别用于预测下个时刻和下一个小时段的数据。将该方法与灰色关联分析预测方法进行比较,发现该方法在预测精度上具有显著优势,证明新的预测超短期负荷方法有较高的可靠性与有效性。仿真结果表明,新的预测超短期负荷方法在工作日或休息日负荷曲线拐点处的预测上,均具有较高精度。

Very Short-term Load Forecasting Method Based on Partial Similarity and Neural Network

In order to improve the precision of very short-term load forecasting,especially in the load curve at the inflexion point,a modified model to select partial similarity day based on fuzzy cluster analysis is proposed. A new method of very short- term load forecasting is proposed based on the partial similar day model and Radial Basis Function（RBF）network. Two prediction models based on the method are proposed to respectively predict the next time of load and the next hour period of load. By Comparing with gray relative analysis forecasting method,it is found that the new very short-term load forecasting method has higher reliability and validity,Simulation results show that the new very short-term load forecasting method has higher accuracy at the inflexion point of load curves of weekday or weekend.