微电网光伏发电的Adaboost天气聚类超短期预测方法

发布时间：2018-04-04 00:44

本文选题：光伏发电　切入点：微电网　出处：《电力系统自动化》2017年21期

【摘要】：微电网光伏发电预测精度与天气状态呈高度相关性,非晴空条件下气象因素的随机波动使得超短期预测精度较低。对此,文中提出一种改进Adaboost天气聚类和马尔可夫链的组合预测方法。首先采用滑动平均法提取辐照度特征变量,设计并训练Adaboost改进的K近邻(KNN)分类器,实现历史样本的分类;为进一步提高多云和阴雨天的预测精度,引入天气类型衰减系数对Hottel太阳辐射模型进行校正,形成完整描述各天气类型的辐照度基准模型;建立多阶加权马尔可夫链模型输出辐照度预测值;最后由光电转换模型实现间隔5 min的微电网光伏超短期预测。仿真结果表明,所述预测方法提高了各天气类型下的预测精度,对提高微电网经济调度水平具有重要意义。
[Abstract]:The prediction accuracy of photovoltaic power generation in microgrid is highly correlated with the weather state, and the ultra-short-term prediction accuracy is low due to the random fluctuation of meteorological factors under the condition of no clear sky.In this paper, an improved combined forecasting method of Adaboost weather clustering and Markov chain is proposed.Firstly, the irradiance feature variables are extracted by moving average method, and the K-nearest neighbor KNN classifier improved by Adaboost is designed and trained to realize the classification of historical samples, in order to further improve the prediction accuracy of cloudy and rainy weather.The attenuation coefficient of weather type is introduced to correct the solar radiation model of Hottel, and the irradiance reference model describing each weather type is formed, and the multi-order weighted Markov chain model is established to output irradiance prediction value.Finally, the photovoltaic ultra-short term prediction of microgrid with 5 min interval is realized by photoelectric conversion model.The simulation results show that the prediction method improves the prediction accuracy of different weather types and is of great significance to improve the economic dispatching level of microgrid.
【作者单位】：武汉大学电气工程学院;
【基金】：国家重点研发计划资助项目(2017YFB0903700,2017YFB0903705) 武汉市科技创新计划资助项目(2013072304020824)~~
【分类号】：TM615

【相似文献】