Journal of Arid Meteorology ›› 2021, Vol. 39 ›› Issue (3): 524-532.

Previous Articles     Next Articles

Research on Wind Power Prediction Algorithm Under Complicated Terrain in Mountainous Area of Hubei Province

XU Peihua1,2, CHEN Zhenghong2, SUN Yanwei3, WANG Biqiang2, JIAN Shilue2#br#   

  1. 1. School of Education Information Technology, Faculty of Artificial Intelligence Education,
     Central China Normal University, Wuhan 430079, China;
    2. Hubei Meteorological Service Center, Wuhan 430074, China;
    3. School of Computer Science, Hubei University of Education, Wuhan 430205, China
  • Online:2021-06-30 Published:2021-07-16

湖北山区复杂地形条件下风电功率预报算法研究

许沛华1,2,陈正洪2,孙延维3,王必强2,简仕略2   

  1. 1.华中师范大学人工智能教育学部教育信息技术学院,湖北武汉430079;
    2.湖北省气象服务中心,湖北武汉430074;3.湖北第二师范学院计算机学院,湖北武汉430205

Abstract: Based on the data of 3 wind farms in Hubei Province from May 2018 to December 2019, including numerical weather forecast, wind speed observed by wind measuring tower and actual power of wind farm, the output results of numerical model forecast were input into three power forecasting models, namely physical method, partial least squares method and neural network method. Then, the rolling correction and modeling of wind speed and power curve were carried out based on the observed wind measurement data of wind farm, and the results were input into the above three power forecasting models. On this basis, the softmax ensemble forecasting algorithm was proposed, and three algorithms with the highest accuracy were selected for ensemble forecasting. Through indepth analysis of three wind farms, the number of qualified days in each month of this algorithm was 1 to 3 days higher than the average value ensemble forecasting algorithm.

Key words: complex terrain, wind farm, power prediction, softmax ensemble prediction, qualified ratio, big data analysis

摘要: 利用湖北省3个风电场2018年5月至2019年12月的数值天气预报、测风塔观测风速、电站实况功率资料,首先将数值模式预报输出结果输入到物理法、偏最小二乘法、神经网络法3种功率预报模型中,然后结合风电场实际测风数据对风速和功率曲线进行滚动订正和建模,再输入到上述3种功率预报模型中。在此基础上,提出softmax集成预报算法,对上述6种预报算法优选出3种预报准确率最高的算法进行集成预报,结果表明该算法较均值法集成预报每个月的合格天数提升1~3 d。

关键词: 复杂地形, 风电场, 功率预报, softmax集成预报, 合格率, 大数据分析