基于DBN与T-S时变权重组合的光伏功率超短期预测模型  

ULTRA-SHORT-TERM PV POWER FORECASTING MODEL BASED ON DBN AND T-S TIME-VARYING WEIGHT COMBINATION

作　　者：谭小钰 刘芳[1] 马俊杰 邹润民[1] Tan Xiaoyu;Liu Fang;Ma Junjie;Zou Runmin(School of Automation,Central South University,Changsha 410083,China;College of Electrical and Information Engineering,Hunan University,Changsha 410082,China)

机构地区：[1]中南大学自动化学院,长沙410083 [2]湖南大学电气与信息工程学院,长沙410082

出　　处：《太阳能学报》2021年第10期42-48,共7页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(61973318,61911530132);湖南省自然科学基金(2020JJ2045);湖南省重点研发计划(2020WK2007)。

摘　　要：针对光伏出力过程复杂、受多种气象因素综合影响,提出一种基于深度信念网络(DBN)和T-S(Takagi-Sugeno)模糊模型的时变权重组合式预测模型对光伏发电功率进行超短期预测。针对单一模型预测精度的时变波动性,结合模型特性采用遗传算法赋予组合模型不同时刻的权重系数。经实测数据验证,所提时变权重组合式预测模型能够结合不同预测模型优势,有效提高光伏功率预测精度。Accurate forecasting of photovoltaic(PV)power has great significance to mitigate the instability of grid-connected PV power.Considering PV power generation process is complicated and affected by various weather factors,a time-varying weight model is proposed to forecast PV power combined with Deep Belief Network(DBN)and T-S(Takagi-Sugeno)fuzzy model.Aiming at the forecasting accuracy volatility with time of single forecasting method,a genetic algorithm(GA)is used to calculate time-varying weight of combined model utilized with the model characteristics.Verified by measured data,the proposed time-varying weight forecasting model can integrate the ascendancy of different forecasting models and effectively improve the forecasting accuracy of PV power.

关 键 词：光伏功率 预测 组合模型 DBN T-S模糊模型 时变权重 

分 类 号：TM615[电气工程—电力系统及自动化]