基于优化神经网络的短期电力负荷预测  

Short- Term Load Forecasting Based on Neural Network Optimized

作　　者：吴桂峰[1] 王轩 陈东雷[1] 

机构地区：[1]扬州大学能源与动力工程学院,江苏扬州225127 [2]扬州北辰电气设备有限公司,江苏扬州225127

出　　处：《计算机仿真》2013年第11期95-99,111,共6页Computer Simulation

基　　金：江苏省科技支撑计划(BE2011133)

摘　　要：研究短期电力负荷预测问题,短期电力负荷具有时变性、不确定性、非线性等特点,传统的线性预测方法无法正确描述短期电力负荷准确预测的变化规律,且神经网络存在局部极小值、过拟合、泛化能力不强等缺陷,预测精度比较低。为了提高短期电力负荷预测精度,提出了一种变参数量子粒子群(VPQPSO)算法优化BP神经网络的短期负荷预测模型(VPQPSO-BPNN)。首先变参数量子粒子群算法优化BP神经网络参数,然后采用优化的BP神经网络对短期电力负荷预测的非线性变化规律进行建模,最后采用对某地区短期电力负荷数据进行仿真。仿真结果表明,VPQPSO解决了BP神经网络存在的难题,提高了短期电力负荷的预测精度,减少了预测误差。In order to forecast short-term load accurately and quickly, this paper proposed a short load forecasting model based on BP neural network optimized by quantum particle swarm optimization algorithm. Firstly, the data of short load were reconstructed by chaotic theory, and then the parameters of BPNN were considered. The position vec- tor of quantum particle and the optimal parameters of BPNN were found by quantum particle optimization algorithm to reduce blindness and inefficiency. Lastly, the optimal model for network traffic was built and the performance of mode was tested with short load data. The simulation results show that VPQPSO algorithm has solved the problems of the BP neural network, and the proposed model can improve the power load forecasting accuracy.

关 键 词：电力负荷 预测精度 量子粒子群算法 变参数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]