基于改进深度信念网络的风电场短期风速预测  

Short-term wind speed forecasting of wind farm based on improved deep belief network

作　　者：王硕禾[1] 张嘉姗 陈祖成 张国驹 郭威 Wang Shuohe;Zhang Jiashan;Chen Zucheng;Zhang Guoju;Guo Wei(Shijiazhuang Tiedao University,Shijiazhuang 050043,China;Tianjin Municipal Engineering Design&Research Institute,Tianjin 300380,China;Tianjin Tiancheng Smart Energy Technology Co.,Ltd.,Tianjin 300450,China)

机构地区：[1]石家庄铁道大学,河北石家庄050043 [2]天津市市政工程设计研究院,天津300380 [3]天津天诚智慧能源科技有限公司,天津300450

出　　处：《可再生能源》2020年第11期1489-1494,共6页Renewable Energy Resources

基　　金：天津市科技计划项目(19YFZGQY00040);河北省分布式能源应用创新中心资助项目(SG20182050)。

摘　　要：受大气湍流影响,滨海地区风能的随机性与波动性更为明显。为提高滨海风电场风速的预测精度,文章针对传统深度信念网络易陷入局部最优以及训练时间过长等问题,提出改进的深度信念网络模型(SA-ALS-DBN)来预测风速。该方法首先采用模拟退火算法对DBN的初始参数进行优化,而后引入自适应学习步长算法缩短了训练DBN所需要的时间。实验结果表明,相比其他方法,SA-ALS-DBN方法提高了预测精度,缩小了预测误差,对于滨海地区风速预测具有更高的可行性和有效性。Due to the influence of atmospheric turbulence,the randomness and fluctuation of wind energy in coastal areas are more obvious.In order to improve the prediction accuracy of wind speed of coastal wind farms,this paper proposes an improved deep-belief network model(SA-ALS-DBN)to predict wind speed in view of the problems that traditional deep-belief networks tend to fall into local optimality and training time is too long.This method first uses the simulated annealing algorithm to optimize DBN's initial parameters,and then introduces an adaptive learning step algorithm to reduce the time required to train the DBN.The experimental results show that compared with other methods,SA-ALS-DBN method improves the prediction accuracy and reduces the prediction errors,which has higher feasibility and effectiveness for wind speed prediction in coastal areas.

关 键 词：风电场风速预测 深度信念网络 模拟退火算法 自适应学习步长 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]