基于相似日和EM-ELM的短期光伏功率预测

Short-term photovoltaic power prediction based on similar days and EM-ELM

光伏功率精准预测有助于电网调度中心提前编制出科学的调度计划,从而提升经济效益。提出一种基于相似日理论结合极限学习机(extreme learning machine,ELM)的光伏功率预测模型。首先,采用皮尔森相关系数法挑选出与光伏出力相关程度最强的气象特征;其次,采用模糊C均值聚类法(fuzzy-c-means algorithm,FCM)将历史样本数据按照所选的气象特征划分相似日;最后,根据预测日天气类型选取对应相似日数据样本集,利用增量型极限学习机进行预测,将影响光伏出力的主要天气特征量作为输入,光伏发电功率作为输出,并采用改进的果蝇优化算法对极限学习机隐含层参数进行优化。预测结果表明,提出的方法对光伏功率预测精度有明显的提升。

Accurate prediction of photovoltaic power can enable power grid dispatching center to work out scientific dispatching plan in advance and improve economic benefits.This paper presents a photovoltaic power prediction model based on the similar day theory combined with extreme learning machine(ELM).Firstly,Pearson correlation coefficient method was used to select the meteorological features with the strongest correlation degree with photovoltaic output.Secondly,the fuzzy c-means algorithm(FCM)was used to divide the historical sample data into similar days according to the selected meteorological characteristics.Finally,the data sample set corresponding to similar days was selected according to the weather types of the forecast days,and the incremental extreme learning machine was used for prediction.The main weather features affecting the photovoltaic output were taken as the input,and the photovoltaic power was taken as the output.The parameters of the hidden layer of the extreme learning machine were optimized by the improved fruit fly optimization algorithm.The prediction results show that the method proposed in this paper can significantly improve the prediction accuracy of photovoltaic power.