周恒俊,王璇,王志远,许若冰

• 1:国网江苏省电力有限公司南京供电分公司
• 2:天津相和电气科技有限公司

摘要(Abstract)：

高精度的光伏出力预测有助于电力系统的安全经济运行和电力资源的协调利用。本文提出了一种新颖的基于互信息矩阵主成分分析与门控循环单元网络的光伏出力短期预测方法。此方法为解决传统主成分分析只能衡量特征变量间线性关系的局限性,将互信息引入主成分分析中以优化主成分分析结果和预测模型的输入变量。结合基于互信息的主成分分析结果及历史光伏出力数据,构建了有更强适用性的门控循环单元预测模型。此外,为更有效地训练模型参数并防止过拟合现象,在优选模型结构的同时引入Adam算法和Dropout技术以进一步优化光伏出力预测模型。算例分析表明,本文所提预测方法较其他方法能更有效地进行特征分析、更准确地把握光伏出力的变化规律,表现出了更高的预测精度。

关键词(KeyWords)： 光伏出力预测;互信息;主成分分析;门控循环单元神经网络

Abstract:

Keywords:

基金项目(Foundation): 国网江苏省电力有限公司科技项目(J2019001)

作者(Author): 周恒俊,王璇,王志远,许若冰

DOI: 10.19635/j.cnki.csu-epsa.000630

参考文献(References)：

• [1]Ming Mengjun,Wang Rui,Zha Yabing,et al.Multi-objective optimization of hybrid renewable energy system using an enhanced multi-objective evolutionary algorithm[J].Energies,2017,10(5):674-688.
• [2]孔祥玉，郑锋，鄂志君，等（Kong Xiangyu,Zheng Feng,EZhijun,et al）.基于深度信念网络的短期负荷预测方法（Short-term load forecasting based on deep belief network)[J].电力系统自动化（Automation of Electric Power Systems),2018,42(5):133-139.
• [3]张倩，马愿，李国丽，等（Zhang Qian,Ma Yuan,Li Guoli,et al）.频域分解和深度学习算法在短期负荷及光伏功率预测中的应用（Applications of frequency domain decomposition and deep learning algorithms in short-term load and photovoltaic power forecasting)[J].中国电机工程学报（Proceedings of the CSEE),2019,39(8):2221-2230.
• [4]茆美琴，龚文剑，张榴晨，等（Mao Meiqin,Gong Wenjian,Zhang Liuchen,et al）.基于EEMD-SVM方法的光伏电站短期出力预测（Short-term photovoltaic generation forecasting based on EEMD-SVM combined method)[J].中国电机工程学报（Proceedings of the CSEE),2013,33(34):17-24.
• [5]Shi Jie,Lee Wei-Jen,Liu Yongqian,et al.Forecasting power output of photovoltaic systems based on weather classification and support vector machines[J].IEEE Trans on Industry Application,2012,48(3):1064-1069.
• [6]朱乔木，李弘毅，王子琪，等（Zhu Qiaomu,Li Hongyi,Wang Ziqi,et al）.基于长短期记忆网络的风电场发电功率超短期预测（Short-term wind power forecasting based on LSTM)[J].电网技术（Power System Technology),2017,41(12):3797-3802.
• [7]朱想，居蓉蓉，程序，等（Zhu Xiang,Ju Rongrong,Cheng Xu,et al）.组合数值天气预报与地基云图的光伏超短期功率预测模型（A very short-term prediction model for photovoltaic power based on numerical weather prediction and ground-based cloud images)[J].电力系统自动化（Automation of Electric Power Systems),2015,39(6):4-10.
• [8]Guo Jun,Xu Xunjian,Lian Weiwei,et al.A new approach for interval forecasting of photovoltaic power based on generalized weather classification[J].International Transactions on Electrical Energy Systems,2019,29(4):2802-2817.
• [9]Liu Jun,Fang Wanliang,Zhang Xudong,et al.An improved photovoltaic power forecasting model with the assistance of aerosol index data[J].IEEE Trans on Sustainable Energy,2015,6(2):434-442.
• [10]Zhou Hangxia,Zhang Yujin,Yang Lingfan,et al.Shortterm photovoltaic power forecasting based on long short term memory neural network and attention mechanism[J].IEEE Access,2019,7:78063-78074.
• [11]许童羽，马艺铭，曹英丽，等（Xu Tongyu,Ma Yiming,Cao Yingli,et al）.基于主成分分析和遗传优化BP神经网络的光伏输出功率短期预测（Short term forecasting of photovoltaic output power based on principal component analysis and genetic optimization of BP neural network)[J].电力系统保护与控制（Power System Protection and Control),2016,44(22):90-95.
• [12]Qu Xiaoyun,Kang Xiaoning,Zhang Chao,et al.Shortterm prediction of wind power based on deep Long ShortTerm Memory[C]//IEEE PES Asia-Pacific Power and Energy Engineering Conference.Xi'an,China,2016:1148-1152.
• [13]范雪莉，冯海泓，原猛（Fan Xueli,Feng Haihong,Yuan Meng）.基于互信息的主成分分析用于声场景分类（Mutual information based principal component analysis for acoustic environment classification)[J].声学技术（Technical Acoustics),2013,32(3):222-227.
• [14]Liu Huawen,Sun Jigui,Liu Lei,et al.Feature selection with dynamic mutual information[J].Pattern Recognition,2008,42(7):1330-1339.
• [15]Wu Wenyi,Liao Wenlong,Miao Jian,et al.Using gated recurrent unit network to forecast short-term load considering impact of electricity price[J].Energy Procedia,2019,158:3369-3374.
• [16]Wen Lulu,Zhou Kaile,Yang Shanlin.Load demand forecasting of residential buildings using a deep learning model[J].Electric Power Systems Research,2020,179:106073.1-106073.8.