摘要
随着社会的不断发展，传统能源的大量消耗使人们在工业发展和日常生活中面临关于不可再生能源耗尽和严重的环境污染等问题。太阳能作为一种优秀的可再生能源而受到世界各国的重视并具有较大发展潜力。随着光伏发电系统容量的不断扩大，准确地预测光伏系统未来几天的发电量对保证电网的稳定运行和大规模光伏发电系统的发展有着重要义。而城市建筑屋顶作为城市中利用率较低的部分，如果在闲置的屋顶上均安装太阳能光伏发电系统，对整个城市都将带来许多利益。本文提出了一种利用气象信息和历史发电量来预测次日光伏发电量的模型，整个模型采用非线性映射能力较强的BP神经网络来建立。原始数据由杭州电子科技大学光伏发电微网实验室提供，包括实验平台记录下的历史气象信息和对应当天的光伏发电量。由于原始数据有限，本文采用模块化的思想，先将模型按季节划分为春、夏、秋、冬四个子模型，再将每个季节模型按日气象类型划分为晴天、云天和雨天三个子模型，共计十二个子模型。以2010年10月的发电数据和气象数据为例，输入数据为预测前一日的光伏发电量和预测当日的温度和光照强度，对建立的神经网络进行训练，并对训练好的模型进行了测试、预测及评估。结果表明，预测模型的预测精度较高，对发电量的预测有较好的参考作用。最后结合杭州电子科技大学，查阅下沙校区的建筑物屋顶面积，推广到校园建筑物所有屋顶都安装上太阳能电池板，预测每日总发电量。
关键词：光伏发电量预测气象因子  BP神经网络模块化
ABSTRACT
With the development of society, large consumption of traditional energy makes people face the problem of non-renewable energy depletion and serious environmental pollution in industrial development and daily life. As a predominant energy,solar energy has been paid attention to and will be a potential one new energy.With the increase of the capacity of PV system, forecast the generating capacity of PV systems in the next few days accurately has an important meaning to ensure the stable operation of electric grid and large-scale development of PV system. And the roof of city’s construction is unemployed. if the PV system are installed on these idle roof, it will bring much benefits to the city. This paper presents a prediction of the PV system model using the historical meteorological information and historical generation, BP neural network which has the ability of nonlinear mapping is used to establish the model. The original data is provided by photovoltaic micro-grid Laboratory of Hangzhou Dianzi University, which includes the historical meteorological information and the corresponding amount of the photovoltaic power generation. Because the original data is limited, this paper modularize the project .According to the fact of the season ,the first model is divided into four sub-models which are named spring, summer, autumn and winter , and then divided each season into three sub-models named sunny day, cloudy day and the rainy day, these twelve sub models combines the project. Take the point data and meteorological data of October, 2010 as an example, the PV power generations which were measured on the day before the prediction and the temperature and th
e amount of PV power generations which were measured on the predicted day are the input data , the ANN will be trained, and the trained model will be tested and predicted, also the prediction will be
estimated. The results show that the model has a high prediction accuracy, It has a good reference to the calculation of power generation. Finally, according to Hangzhou Dianzi University, find the data of the roof area of Xiasha campus's buildings, extended to the case that all roof of the campus are installed with solar panels, predict total power generation capacity.
Keywords: the forecasting of PV system  meteorological factor  BP neural network
modularity
目录
1.绪论 (2)
1.1课题研究背景及意义 (2)
1.2 光伏发电系统发电量预测方法综述 (3)
1.2.1 原理预测法 (3)
1.2.2 统计预测方法 (4)
1.2.3 智能预测方法 (5)
1.2.4 不确定理论预测方法 (7)
1.3 国内外光伏发电量预测的研究动态 (8)
1.4 本文主要内容及章节安排 (10)
2.光伏发电系统概要 (11)
2.1 太阳能电池发电原理 (11)
2.2 光伏发电系统的组成 (12)
杭州电子科技大学
2.3太阳能电池随环境变化的输出特性 (13)
3.BP神经网络基本原理 (16)
3.1 BP神经网络的结构 (17)
3.2 BP神经网络的学习算法 (18)
3.3 BP神经网络的设计 (21)
3.4 BP网络的限制与不足 (23)
4.基于BP神经网络的光伏发电量预测模型设计 (25)
4.1影响光伏发电量的环境因素 (25)
4.1.1 辐照强度对光伏系统发电功率的影响 (25)
4.1.2 日气象类型对光伏系统发电功率的影响 (26)
4.1.3 温度对光伏系统发电功率的影响 (28)
4.1.4 季节对光伏阵列发电量的影响 (29)
4.2 预测模型设计 (29)
4.2.1 输入层节点的确定 (30)
4.2.2 隐含层节点的确定 (30)
4.2.3 输出层节点的确定 (31)
4.2.4 预测模型的训练与评估 (32)
4.3 预测结果分析 (34)
4.3.1 晴天模型预测结果分析 (34)
4.3.2 云天模型预测结果分析 (38)
4.3.3 雨天模型预测结果分析 (41)
4.3.4 预测分析总结 (44)
4.4 结合杭州电子科技大学全校实际可利用建筑楼屋顶面积进行
评估 (45)
5.结论 (499)
致谢 (51)
参考文献 (53)
附录 (56)