摘要
针对传统神经网络收敛速度较慢且实时性较差的缺点,在考虑气候因素的情况下,分别用改进BP网络、径向基函数网络和Elman网络算法对某地区的负荷进行预测。通过对预测误差的分析,用证据理论的Dempster合成法则对算法进行融合,通过选取待预测日之前几天的数据作为融合样本,规定相应的基本信度函数,得到融合后的信度分配,从而决定相应时刻的预测模型。仿真结果表明,经过证据理论融合后选择的负荷预测算法具有较高的预测精度。
Aiming that the traditional neural network has slow concentration speed and weak follow-time function, The distance changed BP network, RBF network and Elman network calculating ways are used in load forecasting at an area, according to the situation of weather factor. By analyzing the error of load forecasting, calculating ways have fusion through Dempster fusion rule of evidential theories. Through choosing the data of several days before estimated day as fusion samples and ruhng the homologous basic behef degree function, behef degree allotment is fetched after the fusion.The load forecasting model is decided according to the corresponding time.The results of simulation demonstrate that the load forecasting calculating ways after fusion of evidential theories have higher accuracy.
出处
《华中电力》
2007年第3期1-4,7,共5页
Central China Electric Power
