物联网学报 ›› 2021, Vol. 5 ›› Issue (4): 81-89.doi: 10.11959/j.issn.2096-3750.2021.00221

• 理论与技术 • 上一篇    下一篇

基于多预测模型与非线性组合的神经网络风速预测

王家君1, 曹薇2, 张贵龙2, 张淮智1, 凌子兴1, 赵小强1   

  1. 1 西安邮电大学,陕西 西安 710121
    2 雅砻江流域水电开发有限公司,四川 成都 610051
  • 修回日期:2021-11-08 出版日期:2021-12-30 发布日期:2021-12-01
  • 作者简介:王家君(1997− ),男,西安邮电大学通信与信息工程学院硕士生,主要研究方向为物联网技术及应用、风能资源在线测量等
    曹薇(1982− ),女,雅砻江流域水电开发有限公司高级工程师、主任工程师,主要研究方向为战略规划、新能源管理等
    张贵龙(1985− ),男,雅砻江流域水电开发有限公司高级工程师,主要研究方向为电气工程与自动化、太阳能和风力发电等
    张淮智(1996− ),男,西安邮电大学通信与信息工程学院硕士生,主要研究方向为物联网技术及应用、风能资源在线测量等
    凌子兴(1998− ),男,西安邮电大学通信与信息工程学院硕士生,主要研究方向为物联网技术及应用、神经网络等
    赵小强(1977−2021),男,博士,西安邮电大学教授,主要研究方向为物联网技术及应用、智慧农业、环境监测及风能资源在线测量等
  • 基金资助:
    国家自然科学基金-雅砻江联合基金资助项目(U1965102);陕西省创新人才推进计划-物联网科技创新团队(2019TD-028)

Neural network wind speed prediction based on multiple prediction model and nonlinear combination

Jiajun WANG1, Wei CAO2, Guilong ZHANG2, Huaizhi ZHANG1, Zixing LING1, Xiaoqiang ZHAO1   

  1. 1 Xi'an University of Posts and Telecommunications, Xi’an 710121, China
    2 Yalong River Basin Hydropower Development Co., Ltd., Chengdu 610051, China
  • Revised:2021-11-08 Online:2021-12-30 Published:2021-12-01
  • Supported by:
    The Yalong River Joint Funds of the National Natural Science Foundation of China(U1965102);The IoT Innovation Team for Talent Promotion Plan of Shanxi Province(2019TD-028)

摘要:

针对复杂山地风速在时空特性上存在强随机性的问题,为了提高风速数据预测的准确性,提出一种基于多预测模型与非线性组合的神经网络风速预测算法。在算法第一层,利用灰狼优化器(GWO)并引入动态收敛因子改进鲸鱼优化算法(WOA),将改进后的 WOA 应用于反向传播神经网络(BPNN)权值及偏置项的更新过程。同时,基于改进后的鲸鱼优化算法的反向传播神经网络(IWOABP)、极限学习机(ELM)和长短期记忆(LSTM) 3种优势互补的方法构建组合预测方法,并在此基础上利用算法第二层的ELM混合机制,以非线性方式学习第一层与最终结果的关系。仿真结果表明,所提算法相较于BPNN、小波神经网络(WNN)及灰狼优化器导向的BPNN (GWOBP),预测误差均有所降低。

关键词: 复杂山地, 风速预测, 鲸鱼优化算法, 神经网络

Abstract:

For the problem of strong randomness in space-time characteristics of wind speed in complex mountains,in order to improve the accuracy of wind speed data prediction, a neural network wind speed prediction algorithm based on multi prediction model and nonlinear combination was proposed.In the first layer of the algorithm, the grey wolf optimizer (GWO) and the dynamic convergence factor were used to improve the whale optimization algorithm (WOA), and the improved WOA was applied to the updating process of BPNN weights and bias items.At the same time, the improved whale optimiza-tion algorithm of back propagation neural network (IWOABP), ELM and LSTM three complementary single methods were constructed to build a combination prediction method, and on this basis, the ELM mixing mechanism of the second layer of the algorithm was utilized to learn the relationship between the first layer and the final result in a non-linear way.Simulation results show that compared with BPNN, WNN and GWOBP, the proposed algorithm has lower prediction errors.

Key words: complex mountain areas, wind speed prediction, whale optimization algorithm, neural network

中图分类号: 

No Suggested Reading articles found!