基于相似日聚类和贝叶斯神经网络的光伏发电功率预测研究

Photovoltaic Load Forecasting Based on the Similar Day and Bayesian Neural Network

光伏发电功率的预测是光伏发电规划和运行的基础,因而受到越来越多的重视。文中提出了FCM相似日聚类与智能算法相结合的光伏阵列功率短期预测模型。该方法的思路是首先通过分析影响光伏阵列输出功率的主要因素,对历史数据与预测日气象环境进行模糊分类,并筛选出相似度高的子集作为样本,以提高预测样本的质量;然后通过神经网络映射出特征空间与光伏功率之间的复杂关系,并用贝叶斯理论对神经网络参数进行优化,提高网络的泛化能力。为检验该方法的有效性和精确性,将所提出方法与常用BP神经网络模型对同一仿真算例进行预测,预测结果表明本文提出的预测模型效果更佳。

Since the limitation of the primary energy,the increasing energy consumption and more concern on environmental protection,much attention has been devoted to explore and utilize the renewable energy.Photovoltaic output power forecasting is the foundation of the planning and operation of photovoltaic system.In this paper,a novel short-term PV power forecasting integrating knowledge mining method and intelligent algorithm has been proposed.The main idea of this method is that firstly though knowledge mining it analyzes the important factor impacting the photovoltaic array output.Then Fuzzy c-mean clustering is adopted to classify the history data and the meteorological data on forecasting day.The selected subset including high similarity days would improve the quality of the training samples.Later,Bayesian neural network is built to mapping the complex relationship among the input data and the PV power,and the parameters of the network is optimized according to Bayesian theory to improve the generalization of the model.At last,to valid the effectiveness and accuracy of the proposed method,simulation is carried out.The forecasting result shows the goodness of this method by comparing with traditional BP network.