并网型光伏发电功率预测系统的研究与实现

发布时间：2018-03-07 21:46

本文选题：并网光伏电站　切入点：功率预测　出处：《华北电力大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着我国光伏发电系统得到越来越广泛的应用,随之而来的问题也越来越多。由于太阳辐照度与季度、昼夜等周期性因素有关,还与阴晴等天气的非周期因素有关,致使得光伏发电输出功率有着随机性及间歇性等缺陷。当前我国储能技术还不成熟,当大规模的光伏发电系统并网时,对电网的电能质量及系统稳定带来巨大挑战。因此对光伏发电输出功率进行预测对于电力系统调度及电力系统运行的稳定性有着举足轻重的作用。做好光伏发电功率预测工作对于扩大光伏产业规模及提高光伏产业发展速度具有重要意义。本文在研究光伏发电特性的基础上,提出了粒子群优化算法(PSO)来优化稀疏贝叶斯回归(SBR)的混合算法,并将其应用于光伏功率预测问题中。通过对光伏发电特性及影响因素的分析,得出影响出力的主要因素为光照强度、温度,并以此构建样本集,用上述算法构建的模型进行功率预测。本文采用的稀疏贝叶斯回归是一种可以解决非线性回归的有效方法,其参数的选择和预测结果的精度密切相关。本文采取的是用粒子群算法来代替传统的共轭梯度法以解决稀疏贝叶斯的参数优化过程。经过实验验证,在未经参数优化时,稀疏贝叶斯回归算法的预测精度要略高于支持向量机以及神经网络算法。在参数优化后,预测精度在原有的基础上又得到了进一步提高,验证了算法的有效性。本文在最后本文设计了一个光伏发电功率预测系统,给出了系统的各个方面的详细设计,包括数据库、系统架构、系统功能设计和一些操作界面。该系统实现了光伏功率预测的基本功能,包括模型训练、数据展示及查询、预测结果展示、误差比较等,具备一定的实用性。
[Abstract]:With the more and more extensive application of photovoltaic power generation system in China, there are more and more problems. Because solar irradiance is related to periodic factors such as quarterly, day and night, and aperiodic factors such as cloudy and sunny weather, etc. As a result, the output power of photovoltaic power generation has some defects, such as randomness and intermittency. At present, the energy storage technology in China is not mature, when the large-scale photovoltaic power generation system is connected to the grid, Therefore, it is very important to predict the output power of photovoltaic power system for power system dispatching and stability of power system. The work of power prediction is of great significance to expand the scale of photovoltaic industry and improve the speed of development of photovoltaic industry. On the basis of studying the characteristics of photovoltaic power generation, a hybrid algorithm named Particle Swarm Optimization (PSO) is proposed to optimize sparse Bayesian regression (SBR). By analyzing the characteristics of photovoltaic power generation and the influencing factors, the main factors affecting the power are light intensity, temperature, and the sample set is constructed. In this paper, sparse Bayesian regression is an effective method to solve nonlinear regression. The selection of parameters is closely related to the precision of prediction results. In this paper, particle swarm optimization algorithm is used to replace the traditional conjugate gradient method to solve the parameter optimization process of sparse Bayes. The prediction accuracy of sparse Bayesian regression algorithm is slightly higher than that of support vector machine and neural network algorithm. At the end of this paper, a photovoltaic power prediction system is designed, and the detailed design of all aspects of the system, including database, system architecture, is given. The system realizes the basic functions of photovoltaic power prediction, including model training, data display and query, prediction result display, error comparison and so on.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM615

【参考文献】