基于改进灰色神经网络组合模型的光伏电站短期出力预测  

Short-term Output Forecasting for Photovoltaic Power Station Based on Improved Grey Neural Network Combined Model

作　　者：刘博洋[1] 潘宇[1] 许伯阳 刘文[1] 李焕奇 王苏 

机构地区：[1]东北电力大学电气工程学院,吉林吉林132012 [2]国网吉林省电力有限公司吉林供电公司,吉林吉林130021 [3]国网浙江德清县供电公司,浙江湖州313200

出　　处：《广东电力》2017年第4期55-60,共6页Guangdong Electric Power

摘　　要：针对单一灰色模型和神经网络模型自身的缺陷对传统灰色神经网络组合模型预测精度的影响问题,提出一种基于改进的灰色神经网络组合模型的光伏电站短期出力预测方法。通过把历史日最高和最低气温以及功率数据作为输入,将改进后的单一灰色模型和神经网络模型进行串联组合。采用粒子群优化算法对该组合模型的权值和阈值进行优化,得到改进的灰色神经网络组合模型,可实现提前一天功率预测。某光伏电站群的实测数据验证了该预测方法能够有效提高预测精度。In allusion t o the problem t ha t defec ts of the grey modei and t he neural ne t work modei have influences on forecas-ting precision of the traditionai grey neurai network combined model, this paper presents a method of photovoltaic power based on improved grey neural network modd. By taking the highest and the lowest tempera-ture and power data in history days as input, it combines the improved grey model andneural network model in series.Then, it adopts particle swarm optunization algorithm to optunize weight and threshold of this combined model so as to getthe improved grey neural network combined model for power one photovoltaic power station group verifies effectiveness of forecasting in one day in advanced. Actual measuring data in this forecas ting method in improving forecast ing precision.

关 键 词：光伏短期预测 灰色模型 神经网络模型 平滑处理 粒子群算法 

分 类 号：TM715[电气工程—电力系统及自动化] TP183[自动化与计算机技术—控制理论与控制工程]