基于改进BP神经网络和工业重构理论的用电增长预测方法  

Predictive method for power growth based on improved BP neural network and rebuilding of new industry structure

作　　者：李蒙[1] 郭鑫[1] 谭显东[1] 袁家海[1] 

机构地区：[1]华北电力大学工商管理学院,北京102206

出　　处：《中南大学学报（自然科学版）》2007年第1期143-147,共5页Journal of Central South University:Science and Technology

摘　　要：运用改进BP神经网络方法和工业重构理论,建立新型工业化与用电增长的关系,提出附加动量和自适应学习率的改进BP神经网络构建工业用电增速的预测模型,利用该模型对未来工业用电增长的趋势进行判断。利用Matlab7．0对该模型进行设计,并运用该模型对工业用电增长趋势进行仿真。与传统的BP神经网络模型相比,采用改进后的预测方法仅经过47次训练就满足预定误差要求,而采用传统的预测模型易陷入局部极小点,很难满足预定误差要求;改进预测模型的预测值与实际值拟合程度好,平均预测精度比传统模型预测精度高1％～3％。此外,验证了改进模型的优越性和合理性。An improved BP Neural Network predictive method based on rebuilding of new industry structure was proposed. The relationship between industrialization and power growth was established, the predictive model for power growth of industry was put forward with additional momentum and adaptive learning rate, and the future trend of power growth was estimated by modeling. Matlab7 was applied as the modeling tool to design the predictive model, The predictive model based on improved method reaches error demands by only 47 training times. But it enters into local minimum using the traditional BP（Back propagation） neural network model. The improve method makes the precision of simulation result improve by 1%-3% compared with that of traditional model. The simulation result validates the rationality and feasibility of predictive method. Furthermore, it shows the superiority of improved BP Neural Network.

关 键 词：神经网络 工业结构 预测 

分 类 号：TM714[电气工程—电力系统及自动化] TP18[自动化与计算机技术—控制理论与控制工程]