基于相似日算法及集成学习的短期光伏预测模型

Short-term photovoltaic power forecasting model based on similarity day algorithm and ensemble learning

随着分布式发电系统的日益增多,光伏发电预测逐渐成为影响电力系统运行及调度的关键。本文提出一种基于改进相似日算法和集成学习的短期光伏发电混合预测模型。应用改进的相似日算法在历史数据中找到相似日,将相似日数据和气候因素等作为Bagging集成学习的输入变量,对其进行建模训练。通过公开的光伏数据集进行验证,并与传统的神经网络模型和支持向量机进行对比。结果表明,该模型具有较高的预测精度。

With the increasing of the distributed generation in the power system,photovoltaic power forecasting has become essential in the planning and operation of the electric power system.This paper proposes a hybrid model which utilizing the improved similarity day algorithm and the Bagging ensemble learning for short-term photovoltaic power forecasting.By using the improved similarity day algorithm,the similar day is found out from the history data of photovoltaic output.The similar day and other climate factor make up the input vector of the decision tree model,which has been trained by using the Bagging ensemble learning algorithm.To confirm the effectiveness of the proposed modeling strategy,the model has been tested on the publicly available data set of photovoltaic output and compared with the classical neural network model and SVM.The results of the proposed model show a better accuracy.