摘要
准确的辐照度预测是光伏发电系统预测输出功率的关键,而辐照度受纬度、天气类型、海拔等因素的影响巨大,不同地区差异较大;目前对辐照度的短时预测研究中复杂的气象数据获取难度大,因此提出了一种利用便于获取气象数据进行辐照度短时预测的简单方法;根据武汉市特有的地理位置特点,将天气类型分为四类,将环境监测仪实时测量的温度、辐照度数据及不同时刻的太阳高度角作为网络的输入,用多变量BP神经网络模型对05:00到20:00时的每小时辐照度进行短期预测;将得到的预测结果与仅用历史辐照度数据作为输入得到的预测结果进行对比,该模型准确性有很大的提高;最终以持续性方法为基准得出预测技能;结果显示该模型在A、B类天气时预测技能均在0.75以上,大部分分布在0.80~0.85,表明该模型在仅利用便于获取的气象信息的基础上能够较准确地对短时辐照度进行预测。
The precise prediction of irradiation is the critical factor of predicting the output power of photovoltaic power generation system, yet the irradiation vary from region to region because of the different latitude, weather types and altitude etc. Because complicated meteorological data in currently research about the forecast of short--term irradiation is hard to be collected, this paper proposes a simple method for short-term prediction of irradiance only using the meteorological data which easy to obtain. In consideration of the unique geographical location of Wuhan, the weather is classified to four types. The input variables are temperature, irradiance measured by real-time envi ronmental monitoring and solar elevation angle at different times. Then the irradiation is calculated per hour from 05 ： 00 to 20 ： 00 using the multivariate backward feedback neural network model. Contrasting the result obtained from the model mentioned above and that only taking historical irradiation data as input, the former has a better outcome. At last, the forecasting skills is calculated based on the Persistence Method. The final results show that all the skills of the proposed model are greater than 0.75, almost distributed between 0. 80 and 0.85 for the weather type A and B, which demonstrates the model proposed in this paper is capable of predicting short-term irradiation accurately based on the meteorological information which easy to be obtained.
出处
《计算机测量与控制》
2017年第7期181-185,共5页
Computer Measurement &Control
关键词
辐照度短时预测
太阳高度角
BP神经网络模型
预测技能
forecast of short-term irradiation
solar elevation angle
BP neural network model
forecasting skills
