基于数据挖掘的双模式组合光伏功率日前预测

Day-ahead prediction of dual-mode combined photovoltaic power based on data mining

为了提高光伏发电功率的预测精度,在数据挖掘分析基础上提出双模式组合的日前光伏预测方法。首先,利用波动量分析建立输出功率与天气类型之间更精确的匹配模型,将天气划分为简单与复杂2种天气类型。其次,对于简单天气类型,采用K-means聚类分析选取最相似日和XGBoost回归组合的预测模型;对于复杂天气类型,提出基于变分模态分解(variational modal decomposition,VMD)、采用麻雀算法(sparrow search algorithm,SSA)优化极限学习机(extreme learning machine,ELM)的日前光伏预测模型。最后,选用DKASC Alice Spring光伏电站数据集对2种模型进行验证,并进行仿真实验。实验结果显示,使用双模式组合方法构建的光伏发电功率预测模型在春季和夏季2个不同数据集下,相关系数分别达到96.44%和96.61%,比其他4种常用模型具有更高的预测精度。

In order to improve the prediction accuracy of photovoltaic(PV)power,this paper proposes a dualmode combination prediction method of day-ahead PV power generation based on data mining analysis.Firstly,fluctuation volume analysis is utilized to establish a more accurate matching model between output power and weather type,and the weather is divided into two types including simple weather type and complex weather type.Secondly,a prediction model combining K-means clustering analysis to select the most similar day and XGBoost regression is adopted for simple weather types.For complex weather types,a day-ahead PV forecasting model based on variational modal decomposition(VMD)and using sparrow search algorithm(SSA)to optimize extreme learning machine(ELM)is proposed.Finally,the DKASC Alice Spring PV power plant data set is selected to validate the two models and conduct simulation experiments.The experimental results show that the correlation coefficients of the PV power prediction model constructed by the dual-mode combination method reach 96.44%and 96.61%respectively under two different datasets of spring and summer,which has higher prediction accuracy than the other four commonly used models.