基于季节调整的ARIMA模型与BP神经网络的电力负荷组合预测.pdfVIP

中国科技论文在线 基于季节调整的ARIMA 模型与BP 神经网 络的电力负荷组合预测 马开良，胡建明，牛明飞** 5 （兰州大学数学与统计学院，兰州730000 ） 摘要：电力负荷预测一直是电力系统计划和运行的重要工具，通过预测，电力系统得以合理 的调度电力以满足系统电力需求。所用的负荷预测方法及其准确性，将决定调度系统的准 确性，电力系统的经济运行乃至整个社会的稳定。提高电力负荷预测方法的精度对这些活 10 动有重大的影响，目前大量的研究工作致力于解决这个问题。文提出了一种先用季节调整 消除实际数据集的季节性影响，再用 ARIMA 模型与BP 神经网络模型分别建模，最后再组 合成最终模型的电力负荷预测方法。这种组合预测模型相对于单个的预测模型能够以较小 的MSE 和MAPE 预测一周的每半小时电力负荷值，得到比较满意的预测结果。 关键词：季节调整，ARIMA 模型，BP 神经网络，组合预测模型，电力负荷预测 15 中图分类号：TK-9 Combined model for electric power load forecasting based on the ARIMA model and BP neural network using seasonal exponential adjustment 20 MA Kailiang, HU Jianming, NIU Mingfei (School of Mathematics and Statistics,Lanzhou University,Lanzhou 730000) Abstract: The electric power load forecasting has been an important tool for the planning and operating in power system, through which an electric utility schedules power load reasonably in order to meet system demand. The method and its accuracy used in power load forecasting will 25 determine the accuracy of the scheduling system, the economic operation of power system even the stability of the whole social. Improving the accuracy of load forecasting method has a significant impact on these activities, and recently a lot of researches have devoted to solving this problem. This paper presents a combined model which firstly eliminates the seasonal effect of the actual datasets by seasonal exponential adjustment, then makes the ARIMA model and BP neural 30 network model respectively, and finally gets the final combined model of the electric power load forecasting. This combined forecasting model, whic