登录    |    注册

您好,欢迎来到中国测试科技资讯平台!

首页> 《中国测试》期刊 >本期导读>基于EEMD-MPE-LSSVM的光伏发电功率预测

基于EEMD-MPE-LSSVM的光伏发电功率预测

183    2021-09-23

¥0.50

全文售价

作者:朱瑞金, 龚雪娇, 张娟娟

作者单位:西藏农牧学院电气工程学院,西藏 林芝 860000


关键词:光伏发电功率预测;最小二乘法支持向量机;集合经验模态分解;多尺度排列熵


摘要:

为提高光伏并网的调度效率和运行稳定性,提出一种基于EEMD-MPE-LSSVM的光伏发电功率预测方法。首先,选取光伏发电功率部分历史数据作为训练样本,采用集合经验模态分解(EEMD)方式对历史功率曲线进行分解;然后,对不同频率特性的分解模态分量进行最小二乘支持向量机(LSSVM)预测,并结合初始功率曲线迭代误差完成预测值重构;最后,利用多尺度排列熵(MPE)量化不同天气类型,构建在晴天、阴天、雨雪、突变天气下输入特征向量,同时参与光伏发电功率 LSSVM预测,减少天气因素对预测值的影响。通过对光伏发电功率50天内的真实值和预测值进行对比试验,结果表明该预测算法的平均相对误差(MRE)和均方根误差(RMES)分别为1.56%、3.14%,证明其有效,同时具有小样本、自适应的优势。


Forecast of photovoltaic power generation based on EEMD-MPE-LSSVM
ZHU Ruijin, GONG Xuejiao, ZHANG Juanjuan
College of Electric Engineering, Tibet Agriculture & Animal Husbandry University, Linzhi 860000, China
Abstract: In order to improve the scheduling efficiency and operational stability of photovoltaic grid-connected, proposes a photovoltaic power generation prediction method based on EEMD-MPE-LSSVM. First, select the historical data of photovoltaic power generation power as the training sample, and use the ensemble empirical mode decomposition (EEMD) method to decompose the historical power curve; Then, perform least squares support vector machine (LSSVM) prediction for the decomposition modal components of different frequency characteristics, and complete the prediction value reconstruction by combining the iterative error of the initial power curve; Finally, use the multi-scale permutation entropy (MPE) to quantify different weather types, construct input feature vectors under sunny, cloudy, rain and snow, and sudden weather conditions, and participate in the LSSVM prediction of photovoltaic power generation power at the same time, reducing the impact of weather factors on the predicted value. By comparing the real value and predicted value of photovoltaic power generation within 50 days, the results show that the mean square error (MRE) and root mean square error (RMES) are 1.56% and 3.14%, the effectiveness of the algorithm is proved, and it has the advantages of small samples and self-adaptation.
Keywords: photovoltaic power forecast;least squares support vector machine;ensemble empirical mode decomposition;multi-scale permutation entropy
2021, 47(9):158-162  收稿日期: 2020-11-09;收到修改稿日期: 2020-12-27
基金项目: 国家自然科学基金项目(51667017)
作者简介: 朱瑞金(1986-),男,西藏林芝市人,副教授,硕士,研究方向为人工智能在能源领域内的应用
参考文献
[1] 龚莺飞, 鲁宗相, 乔颖, 等. 光伏功率预测技术[J]. 电力系统自动化, 2016, 40(4): 140-151
[2] 田壁源, 李永新, 文玉玲, 等. 光伏功率预测技术与方法[J]. 电工技术, 2019(15): 25-27
[3] JI L, NIU D, WANG P. Photovoltaic load forecasting based on the similar day and Bayesian neural network[J]. Chinese Journal of Management science, 2015, 23(3): 118-122
[4] 谢晓頔, 潘霄峰, 高云峰, 等. GSIG分布式光伏云缓存调度研究[J]. 中国测试, 2020, 46(6): 13-17
[5] 叶興, 薛家祥. 改进型LSTM网络光伏发电功率预测研究[J]. 中国测试, 2019, 45(11): 14-20
[6] SHANG L, WANG S. Application of improved principal component analysis in comprehensive assessment on thermal power generation unitis[J]. Power System Technology, 2014, 38(7): 1928-1933
[7] 徐先峰, 刘阿慧, 陈雨露, 等. 基于气象因素充分挖掘的BiLSTM光伏发电短期功率预测[J]. 计算机系统应用, 2020, 29(7): 205-211
[8] 王志忠, 韩茂林, 胡海, 等. 基于晴朗系数和多层次匹配的光伏功率预测方法研究[J]. 电测与仪表, 2019, 56(8): 45-50
[9] 侯松宝, 王侃宏, 石凯波, 等. 基于相似日和主成分分析的光伏发电系统短期出力预测[J]. 可再生能源, 2018, 36(1): 15-21
[10] 李攀. 基于稀疏贝叶斯学习的光伏发电功率预测方法[J]. 数据通信, 2017(5): 35-38, 49
[11] 卢鹏宇, 刘宏勋. 基于CEEMDAN-UWT降噪的暂态电能质量检测方法[J]. 中国测试, 2020, 46(7): 60-67
[12] 刘宇韬, 盛文娟. 基于AdaBoost-LSSVM的纤维复合材料损伤识别[J]. 中国测试, 2020, 46(9): 148-153
[13] LI H, GONG S, LAI L, et al. Efficient and secure wireless communications for advanced metering infrastructure in smart grids[J]. IEEE Transaction on Smart Grid, 2012, 3(3): 1540-1551