【作者】 陈静；

【导师】 周腊吾；

【作者基本信息】 湖南大学 ， 电气工程， 2011， 硕士

【摘要】 在能源危机日益严重的今天,新能源的开发利用受到越来越多的关注。在这些新能源中,风力发电更是得到了广泛应用。然而,风电功率的随机性,给电力系统运营带来了一系列问题,当大规模的风电场并入电网时,更是给电网的安全和稳定带来了极大的负担。为了解决该问题,提出了风电场风速和风电功率预测。其中风速预测为风电功率的间接预测方式。通过对风电场风速的预测,可以降低风电功率的随机性,从而可有效缓解风电场给电力系统造成的不利影响。风速预测近年来有一定发展,但是还存在许多不足之处,仍然具有很大的进步空间。通过考虑多种风速短期预测模型,发现单一的风速预测模型的预测精度提升空间有限,于是根据对支持向量机和组合预测模型的研究,利用它们的算法优点,提出了基于一种改进的支持向量机——最小二乘支持向量机的组合预测模型。利用组合预测模型可综合各单项预测模型的信息,并且最小二乘支持向量机可以简化、优化组合预测模型。该组合模型采用RBF神经网络、BP神经网络和遗传神经网络的风速预测值作为输入,实际风速值作为输出,并建立了线性组合预测模型,且以此为参照来分析基于最小二乘支持向量机的组合预测模型的预测性能。各模型的预测性能,采用预测平均绝对误差、平均相对误差以及误差平方和三个误差指标来比较分析。以江西老爷庙的小时风速数据作为研究样本,运用MATLAB进行仿真,利用各模型对风速进行短期预测,证明了基于最小二乘支持向量机的风速组合预测模型的有效性。仿真试验表明,组合预测模型可进一步提升风速预测精度,而且相较于传统线性组合预测模型,基于最小二乘支持向量机的组合预测模型具有比较大的精度优势。最后建立了基于MATLAB/GUI的风速预测系统,以便风速预测的进一步研究,通过该系统可以方便的调整参数、预测时段、模型类型,简化了复杂的调试过程,结果证明该系统相较于原始程序实现具有明显的优势,验证了风速预测系统的可行性。更多还原

【Abstract】 Today with energy resource crisis becomes more severe, using and developing of new energy resource is attracting more and more attention. Among these renewable energy resources, wind power is widely used. However, the randomness of wind power brings a series of problems to power system operation. When large-scale wind farms are integrated to power grid, a great burden will act on the security and stability of power grid In order to resolve the problem, wind speed forecasting and wind power forecasting were proposed and wind speed forecasting is an indirect way to wind power forecasting. Through the forecasting of wind speed, the randomness of wind power will reduce, which can effectively alleviate the adverse effects on power system. In recent years, wind forecasting has made some progress, but there are also shortcomings, it has rooms for further research.According to the consulted literatures and references, this paper reviewed and summarized the techniques on the short-term forecast model of wind speed, it was found that the precision of single forecast model needed to improve. Because of the advantages of support vector machines and combined forecast model, this paper presented a wind speed forecast model based on least square support vector machine. The combined forecast model which optimized by least square support vector machine could synthesize many kinds of message in single model. The forecasted wind speed which calculated by BP neural networks、RBF neural networks and genetic neural networks was separately set as the input of the combined model , the actual wind speed was set as the output of the combined model, then the forecast model was build based on the self-learning capability of the support vector machine. At the same time, this thesis elaborated the Linear combination forecast model which was used to analysis the performance of wind speed forecast model based on least square support vector machine. In order to compare the forecast performance of all kinds model, the average absolute deviation、average relative deviation、square deviation were used to analysis the forecast results. A simulation system was designed during MATLAB, the simulation results showed that the presented model could improve the accuracy of the wind speed forecast. The feasibility of the presented model was verified by the studied samples which the wind speeds per hour were in jiangxi Laoyemiao.Finally, the wind speed forecasting system is established based on MATLAB/GUI to promote convenience on further study. The system can adjust the parameters of the model; setting predict times, selecting model and so on. Simulation results show that the system achieved obvious advantage compared to the original programs; it is proved the feasibility of wind forecasting system.更多还原