基于OLS-SFLA-RBF神经网络的风电出力预测.pdf

清洁能源
Clean Enel
第29卷第⑨期
电与清能源
Vol 29 No,9
013年9
Power System and Clean En
文章编号:1674-3814(2013)09-0062-06
中图分类号:TM614
文献标志码:A
基于OLS- SFLA-RBF神经网络的风电出力预测
彭健,王斐2,洪翠,江岳文,温步瀛
(1.福州大学电气工程与自动化学院,福建福州350100;2.组约大学理工学院电气与计算机工程系,
纽约布鲁克林11201)
Wind Power Forecasting Based on OLS-SFLA-RBF Neural Network
PENG Jian, WANG Fei, HONG Cui, JIANG Yue-wen, WEN Bu-ying
(1. College of Electrical Engineering and Automation, Fuzhou University, Fuzhou 350108, Fujian, China; 2. Department l
Electrical and omputer Engineering, Polytechnic Institute of New York University, Brooklyn 11201, New York USA
ABSTRACT: Increasing the forecasting accuracy of wind power
In order to mitigate climate warming and shortage
can alleviate the negative influence caused by wind rower of fossil fuel energy, the development and utilization
e paper
Idial basis function neural
network(RBF) to establish the wind power forecasting model
of new energy has particularly been paid attention
and uses the orthogonal least squares algorithm(OLS) to preli
the world. According to the published data from the
minarily train the RBF neural network to detemine the neawork's world wind energy association, by the end of June
structure and central nodes in the hidden layer. In addition. the 2012, the worldwide installed wind capacity had
paper introduces Shuffled Frog Leaping Algorithm(SFLA) to reached 254 000 MW, out of which 16 546 MW were
optimize the width value of each radial basis function on the new added in the first si] months of 2012. And the
foundation of preliminary-trained network to further improve the annual growth rate of wind power was more than
nctwork's generalization ability. The forecasting example shows 20g over the past ten years.However, due to the
that the forecasting accura
acy of Rbf'nenlral network with the
width value further optimized is improved with the
characteristics of volatility and intermittent of wind
network s structure and central nodes in the hidden layer
power, it will bring austere challenge to the safety
KEY WORDS: OLS; SFLA; RBF neural network; wind power
and stabilization of power system when large capacity
forecast
wind power connected to the grid. The practical opera
摘要:提高风电出力的预测精度,可以减轻风电并网带来的 ting experience shows that the wind power forecasting
不利影响。利用径向基函数神经网络(RBF)建立风电出力预
can provide scientific decision for the power system s
测模型,并通过正交二乘算法(OES)对RBF神经网络进行初 dispatch to ensure the security and economic opera
步训练,以确定网络结构及隐含层各节点中心。在OLS算法训
练的网络基础上引入蛙趴算法(SFLA),进一步对隐含层基函
tion of the system, and to alleviate the adverse effect
数的宽虔值进行优化以提高?络的泛化能力。实例预测表
aused by the wind power integration2-l
明,在相同的网络结构及隐含层中心下,基函数宽度值优化
Wind power forecasting is a complicated nonli
后的RBF神经网络模型预测精度得到了提升。
near problem which involves many factors. Currently
关键词:正交最小二乘法;混合蛙跳算法;径向基神经网络; The existing forecasting methods include time series
风电出力预测
anal ysis, support vector machine, artificial neural
network(ANN), physical prediction and etc(+-s5l. The
existing research showed that the artificial
This research was supported by Innovation Program for Young
lents in Science and Technology of Fujian Province (Number
twork can complete the task of nonlinear wind
2011J05124)and N