基于动态自适应神经网络和人体舒适度的短期负荷预测

Short-term load forecasting based on dynamic adaptive artificial neural network and human body amenity indicator

针对BP(Back Propagation)神经网络的适应性较差的问题,提出了自适应神经网络的模型,并将其应用到短期负荷预测中。在神经网络进行数据训练时,对于大量的训练数据,提出采用动态自适应的方式进行处理。分析了实时气象因素对短期负荷的影响,以人体舒适度作为气象因子的处理模型。采用杭州地区数据对提出的模型进行验证,与BP模型预测的结果对比,具有更快的预测速度、更高的预测精度。所构建的预测模型具有很好的适应性,并充分考虑了气象因素、日期类型,预测结果表明所提出的预测方法是有效且实用的。

In view of the bad adaptive performance of BP neural network, an adaptive artificial neural network model is proposed and applied to the short-term forecasting. When a great deal of data are trained, a dynamic adaptive method is used to process them. Based on the analysis of the meteorology influence on short-term load forecasting, the human body amenity indicator is chosen as the way of input of online meteorological factors. The model proposed is verified by the load and meteorological data of Hangzhou region and it is proved that it has quicker rate and higher precision compared to BP model. The new method has good adaptability, meanwhile the meteorological factors and date types are considered, thus achieving good performance and practicability.