基于DBN和多元线性回归的风电功率预测  

Wind Power Prediction Based on DBN and Multiple Linear Regression

作　　者：石慧[1] 李芷萱 马旭 SHI Hui;LI Zhi-xuan;MA Xu(Taiyuan University of Science and Technology Engineering,Taiyuan Shanxi 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,山西太原030024

出　　处：《计算机仿真》2023年第6期90-95,448,共7页Computer Simulation

基　　金：国家自然科学基金青年科学基金项目(.61703297;72071183;71701140);山西省基础研究计划(自由探索类)面上项目(20210302123206);山西省回国留学人员科研资助项目(2021-135,2021-134);山西省留学回国人员科技活动择优资助项目(20220029);山西省留学人员科技活动择优资助项目(20220029);山西省高等学校科技创新项目(2021L322)。

摘　　要：提出一种结合深度置信网络(Deep Belief Network, DBN)和多元线性回归的风电功率预测方法,同时考虑风速、风向对功率的影响进行预测研究。首先运用多元线性回归模型建立风速、风向和风电场输出功率之间的映射关系,寻找最优拟合解;然后对深度置信网络进行分层预训练并在最顶层加入多元线性回归最优解,构建预测模型在监督微调下达到参数最优;最后通过某风电场风速、风向对风电功率模型进行预测,表明提出的模型有更好的拟合效果、更精确的预测结果。A wind power forecasting method combining deep belief network(DBN)and multiple linear regression is proposed,and the influence of wind speed and wind direction on power is also considered for forecasting research.First,the multiple linear regression model was used to establish the mapping relationship between wind speed,wind direction,and wind farm output power to find the optimal fitting solution.Then,the deep confidence network was pre-trained hierarchically,the multiple linear regression optimal solution was added at the top level.,and the prediction model was constructed to achieve the optimal parameters under supervision and fine-tuning.Finally,the wind power model was predicted by the wind speed and direction of a wind farm.The results show that the model proposed in this paper has a better fitting effect and more accurate prediction results.

关 键 词：风电功率预测 风向 多元线性回归 深度置信网络 分层训练 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]