基于神经网络的光伏发电系统发电功率预测

发布时间：2018-01-26 16:46

本文关键词： 光伏发电系统功率预测 BP神经网络小波神经网络回声状态网络经验模态分解　出处：《华北电力大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着能源危机和环境保护形势日趋严峻，光伏发电凭借太阳能清洁、环保、可再生等优点备受各国重视。近年来，在政策的推动下，光伏发电产业得到快速发展，光伏电站容量在电力系统电源中所占比例逐渐增加。但是，，由于光伏发电系统发电功率受多种因素影响，其输出具有明显的周期性、波动性和随机性，使得其并网运行后相对于电力系统是一个不可控源，当光伏电站的规模较大时势必将对电网的安全稳定运行造成影响。因此对光伏发电系统的发电功率预测进行研究具有重大的理论和现实意义。本文首先阐述光伏电池的工作原理及光伏发电系统的分类，详细介绍并网光伏发电系统的组成，给出华北电力大学新能源国家重点实验室的光伏发电系统基本结构和相关参数。然后分析光伏发电系统的发电特性，选取辐照强度、环境温度、光板温度、环境相对湿度作为影响光伏发电系统发电功率的主要因素，为建立发电功率预测模型提供理论支撑。人工神经网络作为一种广泛使用的数学工具，在光伏发电功率预测方面应用普遍。本文先后阐述了BP神经网络、小波神经网络、回声状态网络及基于经验模态分解-回声状态网络的基本原理，并建立了4种相应的光伏发电功率预测模型。最后，采用华北电力大学新能源电力系统国家重点实验室的实测数据对4种预测模型分别进行实验验证。预测结果表明所建立的4种预测模型均能较准确的进行预测，其中基于经验模态分解和回声状态网络的组合预测模型，由于结合了两种算法的优点，预测精度更高、稳定性能更好。
[Abstract]:With the energy crisis and the increasingly serious environmental protection situation, photovoltaic power generation with the advantages of solar cleanliness, environmental protection, renewable and so on, has been attached great importance to. In recent years, under the promotion of policies. With the rapid development of photovoltaic power generation industry, the capacity of photovoltaic power plants in the power system has gradually increased. However, the power generation of photovoltaic power system is affected by many factors. Its output has obvious periodicity, volatility and randomness, which makes it is an uncontrollable source relative to the power system after grid-connected operation. When the scale of photovoltaic power station is large, it will inevitably affect the safe and stable operation of power grid. Therefore, it is of great theoretical and practical significance to study the generation power prediction of photovoltaic power generation system. This paper first describes the working principle of photovoltaic cells and the classification of photovoltaic power generation system, and introduces the composition of grid-connected photovoltaic power generation system in detail. The basic structure and related parameters of photovoltaic power generation system in the State key Laboratory of New Energy of North China University of Electric Power are given. Then, the characteristics of photovoltaic power generation system are analyzed, and the radiation intensity, ambient temperature and plate temperature are selected. Environmental relative humidity (RH) is the main factor affecting the generation power of photovoltaic power system, which provides theoretical support for the establishment of generation power prediction model. As a widely used mathematical tool, artificial neural network is widely used in photovoltaic power prediction. In this paper, BP neural network and wavelet neural network have been expounded. The basic principles of echo state network and empirical mode decomposition-echo state network are presented, and four corresponding photovoltaic power prediction models are established. Finally. The four prediction models are verified by the measured data from the State key Laboratory of New Energy Power system of North China University of Electric Power. The prediction results show that the four prediction models can be accurately predicted. The combined prediction model based on empirical mode decomposition (EMD) and echo state network is more accurate and stable because of combining the advantages of the two algorithms.
【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

【相似文献】