短期风力发电功率预测的研究

发布时间：2018-04-19 01:16

本文选题：风电功率预测 + BP神经网络　；参考：《太原理工大学》2014年硕士论文

【摘要】：随着世界经济和社会的快速发展,传统的能源结构已经不能满足人们对能源的需求,人们的目光已经从传统的化石能源转移到可再生清洁能源上来,而风力发电是所有可再生能源中开发程度最高的。随着国内外风电装机容量的快速增长,风力发电产业正面临着风电并网、风机维护以及提高风能利用率等问题,对风电的合理、安全、有效的利用越来越受到人们的重视,风电功率的短期预测能够有效地解决这些问题。本文综述了国内外风力发电的发展现状,风电功率预测技术的研究现状、基本原理及预测方法,以及国内外已开发运行的风电功率预测系统。风电功率短期预测的误差主要是由内在随机性因素和外在随机性因素造成的。内在随机性因素是指预测系统本身存在缺陷或者不完善,外在随机性因素是指系统输入的数据不完善或者输入数据本身存在误差。本文系统的、全面地分析了随机性因素对短期风电功率预测带来的影响,通过改进以及设计新的预测系统来解决在风电功率短期预测中随机性因素所带来误差的问题。针对内在随机性因素问题,本文建立了BP神经网络短期风电功率预测模型,探讨了BP建模过程中参数的确定以及隐含层数的选取,最终得到了最优的BP神经网络预测模型。仿真结果显示BP模型的预测精度和稳定性都比较差。针对BP神经网络的易陷入局部极小值、稳定性差的问题,建立了通过遗传算法和模拟退火算法改进BP模型的GASABP短期风电功率预测模型。仿真结果显示预测系统的预测精度和稳定性有了明显的提高,有效的解决了局部极值的问题。针对传统的机器学习理论的局限性,建立了基于统计学习理论的支持向量机预测模型,对预测模型的参数的选取采用交叉验证网格搜索法。仿真结果显示支持向量机预测模型的预测精度明显高于GASABP模型。通过对三种模型的对比显示,经过不断优化预测模型,能够有效的降低风电功率预测过程中内在随机性因素对风电功率预测精度的影响。针对外在随机性因素问题,上述几个预测模型在对风电功率的影响因素的确定没有统一的指导原则,需要人为按照经验来确定,导致确定的影响因素不完善。此外,由于国内风电场没有建立完善和精确的气象预报系统,采集的数据会含有误差。因此本文建立了混沌时间序列支持向量机短期风电功率预测模型,通过遗传算法对该预测模型的参数进行组合优化。仿真结果显示混沌支持向量机预测模型显示了良好的预测性能。通过与支持向量机模型的对比,混沌时间序列可以包含所有影响因素所携带的统计规律,混沌支持向量机预测模型能有效的降低外在随机性因素对风电功率预测精度的影响。
[Abstract]:With the rapid development of the world economy and society, the traditional energy structure has not been able to meet the energy needs of people, people's eyes have shifted from the traditional fossil energy to renewable clean energy.Wind power is the most developed of all renewable energy sources.With the rapid growth of wind power installed capacity at home and abroad, wind power industry is facing problems such as wind power grid connection, fan maintenance and improving wind energy utilization ratio. People pay more and more attention to the rational, safe and effective utilization of wind power.The short-term prediction of wind power can effectively solve these problems.In this paper, the current situation of wind power generation at home and abroad, the research status of wind power forecasting technology, the basic principle and forecasting method, and the developed and running wind power forecasting system at home and abroad are summarized.The errors of short-term wind power prediction are mainly caused by internal and external randomness factors.The intrinsic random factors refer to the defects or imperfections of the prediction system itself, while the external random factors refer to the imperfections of the input data or the errors of the input data itself.In this paper, the influence of randomness factors on short-term wind power prediction is systematically and comprehensively analyzed, and the error caused by randomness factors in short-term wind power prediction is solved by improving and designing a new forecasting system.In order to solve the problem of inherent randomness, this paper establishes the BP neural network short-term wind power prediction model, discusses the determination of parameters and the selection of hidden layers in the BP modeling process, and finally obtains the optimal BP neural network prediction model.The simulation results show that the prediction accuracy and stability of BP model are poor.Aiming at the problem that BP neural network is prone to fall into local minima and has poor stability, a GASABP short-term wind power prediction model based on genetic algorithm (GA) and simulated annealing algorithm (SA) is established.The simulation results show that the prediction accuracy and stability of the prediction system are improved obviously, and the problem of local extremum is solved effectively.Aiming at the limitation of traditional machine learning theory, a support vector machine (SVM) prediction model based on statistical learning theory is established. The cross-validated grid search method is used to select the parameters of the prediction model.Simulation results show that the prediction accuracy of SVM model is obviously higher than that of GASABP model.The comparison of the three models shows that through the continuous optimization of the prediction model, it can effectively reduce the influence of the inherent random factors on the wind power prediction accuracy in the process of wind power prediction.In order to solve the problem of external random factors, the above prediction models have no unified guiding principle in determining the influencing factors of wind power, which need to be determined artificially according to experience, which leads to the imperfection of the determinant factors.In addition, because the domestic wind farm does not set up perfect and accurate weather forecast system, the collected data will contain errors.In this paper, the short-term wind power prediction model of chaotic time series support vector machine is established, and the parameters of the prediction model are optimized by genetic algorithm.Simulation results show that the chaotic support vector machine model shows good prediction performance.Compared with the support vector machine (SVM) model, the chaotic time series can contain the statistical laws carried by all the influencing factors, and the chaotic SVM prediction model can effectively reduce the influence of external random factors on the prediction accuracy of wind power.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM614

【参考文献】