作者：吴坚, 郑照红, 薛家祥
作者单位：华南理工大学机械与汽车工程学院, 广东 广州 510640
关键词：光伏发电;短期功率预测;深度置信网络;仿真验证
摘要：
针对现有光伏功率预测技术存在提取特征不充分导致预测精度低的问题，提出一种基于深度置信网络的光伏发电短时功率预测方法。根据光伏发电系统的运行特征和深度置信网络的特点，阐述该预测方法的可行性和科学性。搭建功率预测模型，通过无监督学习过程逐层提取输入序列的内在特征；模型顶层采用BP神经网络对特征矩阵和偏移量进行有监督训练，经过误差微调后输出预测结果。综合考虑可能对光伏发电功率产生影响的多种因素（如辐射强度、温度等），并将上述因素做归一化处理后作为模型的初始输入量，在Matlab上对预测模型进行仿真验证。最后将该预测模型与常用的BP神经网络方法进行比较，结果显示所提模型性能优于BP神经网络，证明该模型具有较好的预测准确度。
Research on short-term power prediction of photovoltaic power generation based on deep belief network
WU Jian, ZHENG Zhaohong, XUE Jiaxiang
School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510640, China
Abstract: In view of the fact that the existing photovoltaic power prediction technology has insufficient extraction characteristics leading to insufficient prediction accuracy, a short-term power prediction method based on deep belief network for photovoltaic power is proposed. According to the characteristics of photovoltaic power system and the characteristics of deep belief network, the feasibility and scientificity of the prediction method are expounded. A short-term power prediction model based on depth belief network(DBN) for photovoltaic power is also proposed. Through unsupervised learning process, the model extracts the intrinsic features of the output sequence layer by layer. At the top of this model, with the help of supervised learning, the BP neural network is used to adjust the feature matrix and offsets, and the prediction results are obtained after the reset of the deviations. A variety of factors, such as radiation intensity and temperature, which may affect photovoltaic power, are comprehensively considered. And after the above factors are normalized as the initial input of the model, the prediction model is simulated and verified on the Matlab. Finally, the model is compared with the commonly used BP neural network method, and the results show that the performance of the proposed model is better than that of BP neural network, which proves that the model has good prediction accuracy.
Keywords: photovoltaic power generation;short-term power prediction;deep belief network;simulation and verification
2018, 44(5): 6-11 收稿日期: 2017-11-28;收到修改稿日期: 2018-01-19
基金项目: 广东省自然科学基金项目（2015A030313675）；广东省交通厅科技项目（科技-2017-02-041）；广州市南沙区科技计划项目（2016CX010）；2015东莞市引进第三批创新科研团队项目（2017360004004）；2015年东莞市东城区产学研合作项目（东城府办复［2016］447号）
作者简介: 吴坚(1992-),男,广东茂名市人,硕士研究生,专业方向为测试计量技术及仪器。
参考文献
[1] 冉晓洪,苗世洪,刘阳升,等. 考虑风光荷联合作用下的电力系统经济调度建模[J]. 中国电机工程学报,2014,34(16):2552-2560.
[2] 丁明,王伟胜,王秀丽,等. 大规模光伏发电对电力系统影响综述[J]. 中国电机工程学报,2014,34(1):1-14.
[3] 王新普,周想凌,邢杰,等. 一种基于改进灰色BP神经网络组合的光伏出力预测方法[J]. 电力系统保护与控制,2016,44(18):81.
[4] 姚仲敏,潘飞,沈玉会,等. 基于GA-BP和POS-BP神经网络的光伏电站出力短期预测[J]. 电力系统保护与控制,2015,43(20):83-89.
[5] 王继东,宋智林,冉冉. 基于改进支持向量机算法的光伏发电短期功率滚动预测[J].电力系统及其自动化学报,2016,28(11):9-13.
[6] LORENZ E, HURKA J, HEINEMANN D, et al. Irradiance forecasting for the power prediction of grid-connected photovoltaic systems[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sen-sing,2009,2(1):2-10.
[7] KUDO M, TAKEUCHI A, NOZAKI Y, et al. Forecasting electric power generation in a photovoltaic power system for an energy network[J]. Electrical Engineering in Japan,2009,167(4):16-23.
[8] 尹邵龙,赵亚楠. 深度学习在城市交通流预测中的实践研究[J]. 现代电子技术,2015,38(15):158-162.
[9] 万杰,胡清华,刘金福,等. 基于深度学习理论的短期风速多步预测方法研究[C]//智能化电站技术发展研讨暨电站自动化2013年会论文集,2013:73-78.
[10] 郑毅,朱成璋. 基于深度信念网络的PM2.5预测[J]. 山东大学学报(工学版),2014,44(6):19-25.
[11] HINTON G, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation,2006,18(7):1527-1554.
[12] 高相铭,杨世凤,潘三博. 基于EMD和ABC-SVM的光伏并网系统输出功率预测研究[J]. 电力系统保护与控制,2015,43(21):86-92.
[13] 袁晓玲,施俊华,徐杰彦. 计及天气类型指数的光伏发电短期出力预测[J]. 中国电机工程学报,2013,33(34):57-64,12.
