基于ABC-SVM和PSO-RF的光伏微电网日发电功率组合预测方法研究  

COMBINED FORECASTING METHOD OF DAILY PHOTOVOLTAIC POWER GENERATION IN MICROGRID BASED ON ABC-SVM AND PSO-RF MODELS

作　　者：王小杨[1] 罗多[2] 孙韵琳 李超[1] 李进[2] Wang Xiaoyang;Luo Duo;Sun Yunlin;Li Chao;Li Jin(University of Elecronic Science and Technology of China,Zhongshan Instiute,Zhongshan 528400,China;SINGYES Green Building Technology,Zhuhai 519060,China;Shunde SYSU Institute for Solar Energy,Shunde 528300,China)

机构地区：[1]电子科技大学中山学院,中山528400 [2]珠海兴业绿色建筑科技有限公司,珠海519060 [3]顺德中山大学太阳能研究院,顺德528300

出　　处：《太阳能学报》2020年第3期177-183,共7页Acta Energiae Solaris Sinica

基　　金：广东省自然科学基金(2016A030310020);教育部项目(19YJC630185);中山市科技计划(2018B1104)。

摘　　要：综合考虑气象因素,使用ABC-SVM方法,对历史的气象数据和光伏出力数据进行训练,依据发电量情况将气象数据分为4类;之后在4类气象情况下各选取上万条数据,使用PSO-RF模型分别训练每组数据,得到4个带不同参数的模型;最后根据每天的气象情况运行不同的模型。验证本组合方法之后发现,通过气象分类后得到的模型,可大幅提高光伏发电量预测的效果。The instability of photovoltaic power generation has a negative impact on the operation of power system. In order to improve the accuracy of photovoltaic power prediction,a photovoltaic power prediction model based on combination model is proposed. Firstly,according to Extra Trees Regressor method,the influence of various data collected in microgrid on generation power is evaluated. After the evaluation,several factors which have great influence on photovoltaic generation power are selected,such as daily average radiation,daily average temperature,daily average wind speed and daily average humidity. Then,a similar classification model based on ABC-SVM is established. In each type,the PSO-RF model is used to train tens of thousands of historical data to find the optimal parameters of each model. Finally,several models are compared through practical examples. The conclusion shows that the combined forecasting method can effectively improve the accuracy of photovoltaic power forecasting in microgrid.

关 键 词：光伏发电量预测 支持向量机 粒子群优化 人工蜂群 随机森林 微电网 

分 类 号：TM615[电气工程—电力系统及自动化]