A Forecasting Model of Support Vector Machine Based on Ensemble Empirical Mode Decomposition and Improved Particle Swarm Optimization

doi:10.11985/2016.04.008

Abstract

Abstract:

A forecasting model of support vector machine (SVM) based on ensemble empirical mode decomposition (EEMD) and improved particle swarm optimization (IPSO) is proposed to tackle with the problem of the accuracy of the short-term forecast of photovoltaic system (PVs) hourly output. Both of historical data for the hourly output of PVs and the related meteorological factors, belong to the days that are similar to the forecast day, are taken into consideration. The historical data for hourly output of PVs is decomposed into a series of relatively stable components by using EEMD method. SVM models with different kernel functions and parameters optimized by IPSO are established to forecast different decomposed components. Then different prediction models are established to compare with each other. And the combined prediction model proposed in this paper is validated to has high prediction accuracy, which has a great significance for economic dispatch of power system incorporating large-scale photovoltaic plant.

Key words: Short-term photovoltaic power output prediction, EEMD, IPSO, SVM

CLC Number:

TM615

Zhu Meimei,Su Jianhui,Chen Zhihui. A Forecasting Model of Support Vector Machine Based on Ensemble Empirical Mode Decomposition and Improved Particle Swarm Optimization[J]. Journal of Electrical Engineering, 2016, 11(4): 47-54.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Tab.1

Tab.2

Tab.3

Tab.4

Fig.6

Fig.7

Tab.5

References 17

[1]	董雷, 周文萍, 张沛 , 等. 基于动态贝叶斯网络的光伏发电短期概率预测[J]. 中国机电工程学报, 2013,33:39-40.
	Dong Lei, Zhou Wenping, Zhang Pei , et al. Short-term photovoltaic output forecast based on dynamic bayesian network theory[J]. Proceeding of the CSEE, 2013,33:39-40.
[2]	丁明, 徐宁舟 . 基于马尔科夫链的光伏发电系统输出功率短期预测方法[J]. 电网技术, 2011,35(1):153-154.
	Ding Ming, Xu Ningzhou . A method to forecast shrot-term output power of photovoltaic generation system based on markov chain[J]. Power System Technology, 2011,35(1):153-154.
[3]	Cai Tao, Duan Shanxu, Chen Changsong . Forecasting power output for grid-connected photovoltaic power system without using solar radiation measurement[C]. 2nd International Symposium on Power Electronics for Distributed Generation Systems (PEDG 2010), Hefei, China, 2010.
[4]	Deshmukh M K, Deshmukh S S . Modeling of hybrid renewable energy systems[J]. Renewable and Sustainable Energy Reviews, 2008,12(1):235-249. doi: 10.3390/s19112485 pmid: 31151324
[5]	Yona A, Senjyu T, Funabashi T , et al. Application of neural network to one-day-ahead 24 hours generating power forecasting for photovoltaic system[C]. Intelligent Systems Application to Power System, 2007: 16.
[6]	Atsushi Yona, Tomonobu Senjyu, Ahmed Yousuf Saber , et al. Application of neural network to one-day-ahead 24 hours generating power forecasting for photovoltaic system[C]. 2007 International Conference on Intelligent Systems Applications to Power Systems (ISAP), Toki Messe, Niigata, 2007.
[7]	Mellit A, Pavan A M . A 24h forecast of solar irradiance using artificial netural network: Application for performance prediction of a grid-connected PV plant at Trieste, Italy[J]. Solar Energy, 2010,84(5):807-821. doi: 10.1016/j.solener.2010.02.006
[8]	Duan Xiaobo, Fan Lei . Based on improved BP neutral network model generating power predicting for PV system[C]. 2012 World Automation Congress (WAC), 2012: 1-4.
[9]	王飞, 米增强, 杨奇逊 , 等. 基于神经网络与关联数据的光伏电站发电功率预测方法[J]. 太阳能学报, 2012,33(7):1171-1177.
	Wang Fei, Mi Zengqiang , Yang Qixun et al. Power forecasting approach of PV plant based on ANN and relevant data[J]. Acta Energiae Solaris Sinicam, 2012,33(7):1171-1177.
[10]	栗然, 李广敏 . 基于支持向量机回归的光伏发电出力预测[J]. 中国电力, 2008,41(2):74-78.
	Li Ran, Li Guangmin . Photovoltaic power generation output forecasting based on support vector machine regression technique[J]. Electric Power, 2008,41(2):74-78.
[11]	袁晓玲, 施俊华, 徐杰彦 . 基于BP神经网络的光伏发电短期处理预测[J]. 可再生能源, 2013,31(7):11-16.
	Yuan Xiaoling, Shi Junhua, Xu Jieyan . Shrot-term power forecast for photovoltaic generation based on BP neutral network[J]. Renewable Enerdy Resources, 2013,31(7):11-16.
[12]	罗毅, 邢校萄 . 基于小波变换和支持向量机的短期光伏发电功率预测[J]. 新能源进展, 2014,2(5):381-382.
	Luo Yi, Xing Xiaotao . Short-term forecasting of photovoltaic power generation based on wavelet decomposotion and support vector machine[J]. Advancing in New and Renewable Energy, 2014,2(5):381-382.
[13]	茆美琴, 龚文剑, 张榴晨 , 等. 基于EEMD-SVM方法的光伏电站短期出力预测[J]. 中国机电工程学报, 2013,33(34):18-34.
	Mao Meiqin, Gong Wenjian, Zhang Liuchen , et al. Short-term photovoltaic generation forecasting based on EEMD-SVM combined method[J]. Proceeding of the CSEE, 2013,33(34):18-34.
[14]	孟繁林 . 集合经验模态分解的理论及应用研究[D]. 江苏:江苏科技大学, 2013.
[15]	Vapnik V . The nature of statistical learning theory[M]. New York: Spdnge-Verlag, 1995.
[16]	吴杰康, 韩军锋, 刘蔚 , 等. 基于反捕食粒子群算法的电力系统经济调度方法[J]. 电网技术, 2010,34(6):60-62.
	Wu Jiekang, Han Junfeng, Liu Wei , et al. Economic dispatching of power system based on anti-predatory particle swarm algorithm[J]. Power System Technology, 2010,34(6):60-62.
[17]	宁阳天, 李相俊, 麻秀范 , 等. 基于改进粒子群算法的微网动态经济调度算法[J]. 电力建设, 2014,35(6):26-30. doi: 10,3969/j.issn,1000-7229,2014,06,006
	Ning Yangtian, Li Xiangjun, Ma Xiufan , et al. Microgrid dynamic economic dispatch based on improved particle swarm optimization[J]. Electric Power Construction, 2010,35(6):26-30. doi: 10,3969/j.issn,1000-7229,2014,06,006

预测模型	MAPE	RMSE
SVM	20.29	20.319 8
IPSO-SVM	19.66	20.282 1
EEMD-SVM	18.86	1.284 6
EEMD-IPSO-SVM	4.72	0.550 6

预测模型	MAPE	RMSE
SVM	14.38	16.006 8
IPSO-SVM	13.49	12.823 1
EEMD-SVM	10.14	1.301 6
EEMD-IPSO-SVM	8.12	1.083 1

预测模型	MAPE	RMSE
SVM	24.38	17.004 8
IPSO-SVM	16.53	15.254 1
EEMD-SVM	15.3	3.373 4
EEMD-IPSO-SVM	9.68	1.032 4

预测模型	MAPE	RMSE
SVM	9.58	10.102 4
IPSO-SVM	9.44	9.792 5
EEMD-SVM	8.08	0.685 2
EEMD-IPSO-SVM	4.8	0.412 9

时间	EEMD-IPSO-SVM	EEMD-SVM	IPSO-SVM	SVM
7:00	1.59	10.88	28.27	36.48
8:00	1.37	21.25	19.97	28.73
9:00	1.38	18.2	19.5	38.42
10:00	1.13	4.37	0.44	0.04
11:00	6.66	1.43	17.9	1.36
12:00	5.82	3.72	2.56	1.45
13:00	4.08	0.47	2.62	0.04
14:00	1	12.84	1.09	1.16
15:00	4.56	4.86	0.19	0.34
16:00	1.469	3.07	14.09	2.02
17:00	4.56	24.88	13.01	15.37
18:00	12.95	34.16	28.89	33.56