陈家扬 陈华 张旭



摘 要: 对风电场的准确预测,可以为电网调度提供调峰和消纳依据,从而综合评估电网短期内消纳风电的能力,制定科学合理的消纳措施。通过预测风电场24 h内的出力,基于数值天气预报(NWP)数据的出力预测,采用深度学习神经网络算法,建立数值天气预报与风电功率之间的转换模型,计算功率点预测值,然后利用概率密度函数,建立风电出力预测的概率区间。最后通过实际案例仿真,验证了基于NWP和深度学习神经网络短期风功率预测的可靠性,为调度预留调峰容量提供理论依据。
关键词: 风功率预测; 深度学习神经网络; 数值天气预报; 建立转换模型; 概率密度; 案例分析
中图分类号: TN711?34; TK89; TM614 文献标识码: A 文章编号: 1004?373X(2020)08?0063?05
Short?term wind power forecasting based on NWP and deep learning neural network
CHEN Jiayang, CHEN Hua, ZHANG Xu
(College of Electrical Engineering, Xinjiang University, Urumqi 830047, China)
Abstract: The accurate prediction of wind power plant can provide the basis for peak load regulation and absorption of power grid dispatching, so as to comprehensively evaluate the power grid′s short?term absorption capacity of wind power, and formulate scientific and reasonable absorption measures. The conversion model between numerical weather forecast (NWP) and wind power is established based on the output′s prediction of wind power within 24 hours, the NWP data and deep learning neural network algorithm, the power point prediction value is calculated, and then the probability interval of the wind power output prediction is established by means of the probability density function. The reliability of short?term wind power prediction based on NWP and deep learning neural network is verified by the simulation of practical cases, which provides the theoretical basis for scheduling and reserving peak load regulating capacity.
Keywords: wind power forecast; deep learning neural network; numerical weather prediction; building conversion model probability density
0 引 言
随着能源与环境问题的日益突出,风能作为清洁能源中最重要的可再生能源之一,已受到世界各国的高度重视。2010年底,我国风电装机4 182.7万kW,超过美国,居全球第一。新疆风能资源丰富,新疆风电场依据风资源区域优势,其风电场装机容量逐渐增大,预计到2020年将形成55个百万千瓦级风电基地,总装机规模达到千万千瓦[1] 。
风力发电输出电力的功率多少取决于自然风资源,而国内风电场在规划阶段,基本都是依托设计规划机构实地勘测进行选址和规划,这些机构大多利用的是临时测风和气象部的历史数据。早期的风电场在建设时只是建设了一个或多个测风塔,技术先进的风电场也只是在每台风力机上装备了实时测风装置,大多都没有按照要求建设完善的风速预测系统。……