基于遗传算法改进BP神经网络的短期风电功率预测研究 - 图文

2012年“挑战杯”大学生

课外学术科技作品竞赛及创业设计大赛

作品名称：基于遗传算法改进作品类别：社会科学类项目成员：刘知发联系电话： BP神经网络的短期风电功

率预测研究

陈 军 母桑妮 向亚军 唐艳利 15196009869 完成时间： 2012 年 3 月 20 日0

基于遗传算法改进BP神经网络的短期风电功率预测研究

摘 要

风能发电作为21世纪重要的研究课题之一，是清洁、可再生资源的之首。对降低污染，舒缓能源消耗带来的压力有着至关重要的作用。本文通过时间序列、遗传算法和BP神经网络等方法建立了4个风电功率预测模型。通过Matlab编程，得出了不同方法预测结果，并对其准确性进行比较。

本文首先对国内外风电产业发展现状做了分析。在此基础上，第2章确定以移动平均预测法、随机时间序列预测法、BP神经网络预测法对问题进行探讨，通过Excel与

Matlab混合编程，得出移动平均预测法、随机时间序列预测法、BP神经网络预测法的

准确率分别为82%、70%、84%，合格率分别为85%、65%、92%。得出BP神经网络预测法明显优越于其他两种方法。接着运用BP神经网络预测出的数据做了预测的相对误差分析，从中得出了6组预测值的相对误差（见表3.1），并做了对比误差分析图，通过误差分析图得出“风电机组汇聚会减小风电功率预测误差”的结论。并对造成该结论的原因做了解析，提出了在风力允许范围内，增加风电机组的汇聚度，可进一步减小误差的预期判断。

在第2章的基础上，为了进一步提高风电功率实时预测的准确度，建立了遗传算法与BP神经网络相结合来对风电功率进行预测的模型。通过模型对东北某发电厂一周进行预测，并与实测值进行比较，得到其准确率与合格率高达89%与95%。

最后，对本次课题得出的结论做了分析，总结风电功率的预测结果和存在的问题，以及提出需要进一步改进的地方。

关键词：风电功率预测 随机时间序列 BP神经网络 误差分析 遗传算法

1

ABSTRACT

Wind power generation, one of the most significant research subjects in 21st century, tops among the clean and renewable resources, which plays a critical role in reducing pollution and easing stress produced by energy consumption. The thesis, using methods including time series, genetic algorithm and BP nerves network, establishes 4 prediction models with wind power. Through Matlab programming, the author arrives at various prediction results, and checks their accuracy.

First of all, the thesis analyses development situation of wind power generation at home and abroad. And on the basis of the first step, the second chapter discusses the issue by mobile average prediction method, random time series prediction method and BP nerves network prediction method. Through mixed programming of Excel and Matlab, the accuracy rate of average prediction method, random time series prediction method and BP nerves network prediction method is 82%, 70%, and 84% respectively, while qualified rate of them is 85%, 65%, and 92%. Apart from the above-mentioned, the author also draws that BP nerves network prediction method is superior to the other two ones. Then using the data forecast by BP nerves network prediction method, the author, making predicatively comparative error analysis, acquires six groups of comparative errors of prediction figure (see Chart 1), and draws an analysis chart of comparative errors. A sum-up that gathering of wind turbine is able to decrease predictive errors of wind power is reached by means of error analysis. The causes giving rise to such a conclusion are interpreted as well, and an idea that increasing gathering density of wind turbine can reduce expectant judgment of errors is also put forward.

Based on the Chapter 2, a model, combing with genetic algorithm method and BP nerves network, is established to forecast wind power, so as to increasingly improve real-time monitoring accuracy of wind power. With the model, the author, making a week-long prediction towards a northwest power plant and comparing with actual figures, comes to a conclusion that the accuracy and qualified rates are up to 89% and 95%.

Ultimately, the author analyzes the conclusion of the subject, summarizes the predictive results and existing problems, and proposes some aspects that demand improving.

Key Words: prediction of wind power; random time series; BP nerves network; error analysis; genetic algorithm

2

目 录

摘 要 ............................................................................................................................................1 ABSTRACT...................................................................................................................................2 目 录 ..........................................................................................................................................3 第1章 引 言 ............................................................................................................................5

1.1风电产业发展现状

[1,2]

........................................................................................................5

1.2背景分析与研究意义 .........................................................................................................5 1.3国内外研究动态 ................................................................................................................6 1.3.1国外研究现状[3] ..............................................................................................................6 1.3.2国内研究现状 .................................................................................................................6 第2章 风电功率实时预测及误差分析 .........................................................................................8

2.1移动平均预测法预测风电功率 ...........................................................................................8 2.1.1 移动平均法基本概念......................................................................................................8 2.1.2 移动平均法基本原理......................................................................................................8 2.1.3 移动平均法的特点 .........................................................................................................8 2.1.4 一次移动平均法.............................................................................................................8 2.1.5 二次平移预测法.............................................................................................................9 2.1.6 二次平移预测法基本算法...............................................................................................9 2.1.7 考核指标模型的建立.................................................................................................... 10 2.1.7.1 准确率 ...................................................................................................................... 10 2.1.7.2 合格率 ...................................................................................................................... 10 2.1.8 考核指标模型的求解.................................................................................................... 10 2.2 随机时间序列方法预测风电功率..................................................................................... 12 2.2.1 随机时间序列方法介绍 ................................................................................................ 12 2.2.2 自回归滑动—平均混合模型[5] ...................................................................................... 14 2.2.3 模型的参数估计........................................................................................................... 14 2.2.3.1 AR模型的参数估计 ................................................................................................. 14 2.2.3.2 MA模型的参数估计 ................................................................................................. 15 2.2.3.3 ARMA模型的参数估计 ............................................................................................ 15 2.2.4 模型的定阶[5,6] ............................................................................................................. 16 2.2.4 随预测机时间序列方法的求解...................................................................................... 16 2.3 BP神经网络对风电功率的预测 ....................................................................................... 18 2.3.1 神经网络背景[1,7].......................................................................................................... 18 2.3.2 神经网络概述 .............................................................................................................. 19 2.3.3 BP神经网络建模 .......................................................................................................... 19 2.3.3.1背景知识 ................................................................................................................... 19 2.3.3.2 BP神经网络的学习过程............................................................................................. 21 2.3.4 BP神模型经网络对风电功率的预测的建立 ................................................................... 22 2.3.5 BP神经网络对风电功率的预测方法的求解 ................................................................... 22 第3章 风电机组的汇聚对于预测结果误差的影响 ...................................................................... 26

3

3.1 相对误差定义................................................................................................................. 26 3.2分析减小风电功率误差的原因及预期............................................................................... 28 第4章 进一步提高风电功率实时预测精度的探索 ...................................................................... 29

4.1对问题一中BP神经网络存在的缺陷分析......................................................................... 29 4.2 基于遗传算法的BP神经网络学习算法原理 .................................................................... 29 4.3 遗传算法的主要特征 ...................................................................................................... 29 4.4 算法实现的关键技术及步骤............................................................................................ 29 4.4.1 编码与解码.................................................................................................................. 29 4.4.2 根据适应值评价解的适应程度并据此生成新群体.......................................................... 30 4.4.2.1计算个体适应值 ......................................................................................................... 30 4.4.2.2计算每个染色体的选择概率 ....................................................................................... 30 4.4.2.3 交叉操作 .................................................................................................................. 30 4.4.2.4 变异操作 .................................................................................................................. 31 4.4.2.5 对进化的数据进行检验 ............................................................................................. 31 4.4.3由上述关键技术及其步骤可以画出如下遗传神经网络算法流程图 .................................. 31 4.5 模型求解........................................................................................................................ 32 第5章 最优方案的确定 ............................................................................................................ 33

5.1 遗传BP神将网络算法下预测情况 .................................................................................. 33 5.2 两种神经网络同时对单台发电机组进行相对误差对比 ..................................................... 35 5.3 两种神经网络同时对多组合力发电机组进行相对误差分析 .............................................. 36 5.4 对两种平均相对误差进行比较 ........................................................................................ 37 第6章 结论与展望 ................................................................................................................... 38

6.1结论................................................................................................................................ 38 6.2展望................................................................................................................................ 38 参考文献 ..................................................................................................................................... 40 致 谢 ........................................................................................................................................ 42 附 录 ........................................................................................................................................ 43

4

百度搜索“77cn”或“免费范文网”即可找到本站免费阅读全部范文。收藏本站方便下次阅读，免费范文网，提供经典小说综合文库基于遗传算法改进BP神经网络的短期风电功率预测研究 - 图文在线全文阅读。