基于遗传算法优化的RBF神经网络在光伏发电MPPT中的应用

发布时间：2018-02-07 17:27

本文关键词： 光伏最大功率点跟踪 RBF神经网络遗传算法　出处：《湖南工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着传统燃料的日渐消耗与能源需求量的不断提升,可再生能源逐渐受到关注。由于太阳能具有绿色、安全、可再生等特点,近年来,太阳能光伏发电已经在我国得到了飞速发展。但光伏电池具有生产成本高、光电转换效率低的缺点,因此如何使光伏电池持续有效地输出最大功率以提高发电效率和降低发电成本则成为了当下研究的重点。针对该问题,本文利用神经网络的非线性拟合能力以及遗传算法突出的寻优特点,提出了遗传算法优化的RBF神经网络对光伏发电系统最大功率点进行预测控制。首先,本文对光伏发电的研究背景及国内外研究现状进行了综述,介绍了目前光伏发电MPPT技术的判断标准及不足。详细说明了光伏电池的工作原理,通过MATLAB搭建光伏电池模型获得了U-I及P-V动态变化曲线,并在此基础上得出光照强度和温度为影响最大功率点输出的主要因素。接着,阐述了光伏发电最大功率点跟踪的原理,分析了传统跟踪方法及其改进方法的优缺点。针对传统方法的不足,介绍了基于现代控制理论的神经网络控制法,通过RBF神经网络函数逼近能力的分析,选择其对光伏发电最大功率点进行预测控制。然后,对于RBF神经网络中存在的不足,本文使用了遗传算法对其数据中心、扩展常数及权值进行优化。通过将RBF神经网络的数据中心和其对应的扩展常数以及权值统一编码,加强了隐含层和输出层的合作关系,并利用遗传算法全局搜索的功能特性,使得整个网络模型达到全局最优。此外,对遗传算法本身的机制作出相应的改进,使遗传操作更加完善。最后,将遗传算法优化后的RBF神经网络与优化前的网络对同一光伏发电系统最大功率点进行预测,结果显示优化后的RBF神经网络达到目标误差的训练次数较优化前明显减少,平均误差降低了3.7%,结果证明遗传算法优化后的RBF神经网络不仅提高了预测速度,还提高了预测精确度,从而能更好地实现光伏发电最大功率点跟踪控制。
[Abstract]:With the increasing consumption of traditional fuels and increasing energy demand, renewable energy has attracted more and more attention. Due to the green, safe, renewable and other characteristics of solar energy, in recent years, Solar photovoltaic power generation has been developing rapidly in China, but photovoltaic cells have the disadvantages of high production cost and low photoelectric conversion efficiency. Therefore, how to make photovoltaic cells output maximum power continuously and effectively to improve generation efficiency and reduce generation cost has become the focus of current research. In this paper, based on the nonlinear fitting ability of neural network and the outstanding optimization characteristics of genetic algorithm, a genetic algorithm optimized RBF neural network is proposed to predict and control the maximum power point of photovoltaic power generation system. This paper summarizes the research background of photovoltaic power generation and the current research situation at home and abroad, introduces the judging standard and deficiency of photovoltaic generation MPPT technology at present, and explains the working principle of photovoltaic cell in detail. The dynamic curves of U-I and P-V are obtained by building photovoltaic cell model by MATLAB. On the basis of this, the main factors affecting the output of maximum power point are light intensity and temperature. Then, the principle of maximum power point tracking for photovoltaic generation is described. This paper analyzes the advantages and disadvantages of the traditional tracking method and its improvement, and introduces the neural network control method based on the modern control theory, which is based on the RBF neural network function approximation ability. Then, for the shortcomings of RBF neural network, the genetic algorithm is used to control the data center of photovoltaic power generation. By coding the data center of RBF neural network, its corresponding expansion constant and weights, the cooperative relationship between the hidden layer and the output layer is strengthened, and the global search function of genetic algorithm is used. In addition, the mechanism of genetic algorithm itself is improved accordingly to make genetic operation more perfect. Finally, The RBF neural network optimized by genetic algorithm and the network before optimization are used to predict the maximum power point of the same photovoltaic system. The results show that the training times of the optimized RBF neural network to achieve the target error are obviously reduced compared with those before optimization. The results show that the RBF neural network optimized by genetic algorithm not only improves the prediction speed, but also improves the accuracy of prediction, so that the maximum power point tracking control of photovoltaic generation can be realized better.
【学位授予单位】：湖南工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM615;TP183

【参考文献】