动态调整蚁群-BP神经网络模型在短期负荷预测中的应用  

Short-term load prediction based on dynamic adjustment ant colony system and back propagation neural network hybrid algorithm

作　　者：师彪[1] 李郁侠 于新花[2] 闫旺[1] 

机构地区：[1]西安理工大学水利水电学院,西安710048 [2]青岛科技大学高职业技术学院,青岛261000

出　　处：《水力发电学报》2011年第3期5-9,194,共6页Journal of Hydroelectric Engineering

基　　金：国家火炬计划基金(07C26213711606);陕西省自然科学基础研究计划(SJ08E220);山东省软科学基金(2009RKB190);西安理工大学优秀博士学位论文和科学研究基金(106-210912;106-210917)

摘　　要：为提高短期电力负荷预测精度,提出了动态调整人工蚁群算法(DAACS),对BP神经网络参数进行优化,建立了动态调整人工蚁群-BP神经网络组合算法(DAACS-BP)预测模型,综合考虑气象、天气、电价、日期类型等影响因素,对四川电网进行短期电力负荷预测。仿真表明,该方法克服了BP神经网络和人工蚁群算法易陷于局部极值、搜索质量差和精度不高的缺点,改善了BP神经网络的泛化能力,收敛速度快,输出稳定性好,提高了电网短期负荷预测的精度,可有效用于电力系统短期负荷预测。In this paper,a dynamic adjustment ant colony system（DAACS） is proposed to improve short-term load forecast accuracy,and a short-term load forecast model is developed by using the DAACS-BP hybrid algorithm of DAACS combined with a back propagation（BP） neural network that is trained by the DAACS algorithm.This model can automatically determine the parameters of neural network based on the data sample.A prediction of the Sichuan power grid loads was made by using the model and the history load data,with a comprehensive consideration of various load-impacting factors such as meteorology,weather,price and date types.The results show a faster convergence and a better forecast accuracy of the hybrid method than those of the traditional ant colony system algorithm-BP neural network or the BP neural network,and also a significant improvement on the generalization capacity of BP neural network.Thus,the hybrid algorithm enhances the efficiency of short-term load forecast of a power system.

关 键 词：水电工程 短期负荷预测 DAACS-BP网络算法 动态调整蚁群算法 BP神经网络 

分 类 号：TV752[水利工程—水利水电工程] TM614[电气工程—电力系统及自动化]