基于模糊C均值聚类和案例推理的风电功率预测研究

发布时间：2018-03-12 13:08

本文选题：风电功率预测　切入点：模糊C均值聚类　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着人类对生态环境的逐渐重视以及在国家新能源战略的重大部署下,风能作为一种清洁能源得到了大规模的发展,不仅全球的风电装机容量逐年上升,同时也使得风力发电技术越来越成熟。但是风的间歇性、波动性使得风能具有不可控性,始终是风电技术的一大难题。因此,解决能源问题以及风电场长期稳定运行的关键所在是迫切需要提高风电功率的预测精度。本文以风电功率预测为研究对象,对测风塔数据进行分析,以便发掘和更好的利用测风数据自身的信息。神经网络模型的建立对训练数据具有较高的依赖性,训练数据的选择既要包含足够广的选择范围,这样模型具有更强的泛化能力;同时又要保证模型具有更高的预测精度。针对特殊天气下普通预测模型难以满足要求的问题,本文采用案例推理技术对风电功率预测模型进行改进。本文主要研究内容有:(1)对国内外风力发电的现状、发展趋势以及目前主要存在的问题进行了总结,引出了风电功率预测的意义及必要性,同时对风电功率预测模型的研究现状进行总结分析。(2)针对测风数据不能很好地满足预测样本要求的问题,根据国标GB/T18709-2002及相关参考文献补充对数据进行了检验、填补和修正。采用山西省某风电场的实测数据分析了测风数据自身的变化规律,重点分析了模型的预测误差与风电功率影响因素之间的关系。对预测模型评价指标进行了必要的选取说明。(3)本文提出了基于模糊聚类分析的风电功率预测方法。首先采用减法聚类来确定模糊C均值的聚类数和聚类中心,采用风电场实际运行的数据聚类分析,然后进行神经网络模型的训练,既充分考虑了样本空间的特征,又使得模型具有针对性,因此保证网络模型的泛化能力的同时又提高了预测的精度。(4)为了提高特殊天气下风电功率的预测精度,本文通过对普通模型的预测效果和风电场实际测风数据进行分析,建立了基于案例推理的特殊天气下风电功率预测模型,采用基于模糊聚类和粒子群优化的K近邻算法进行案例检索,提高了检索的速度和精度。采用山西省某风电场的实际数据进行了大量的仿真实验,将仿真结果与GRNN、LSSVM、GABP模型进行了对比,预测误差有了不同程度的改善。特别是预测数据发生突变时,效果更为明显,从而验证了该方法的有效性,为解决风电场特殊天气下的风电功率预测提供了一种可行性方法。
[Abstract]:With the gradual attention of human beings to the ecological environment and the important deployment of national new energy strategy, wind energy as a clean energy has been developed on a large scale, not only the global wind power installed capacity is increasing year by year. At the same time, it also makes wind power technology more and more mature. However, the intermittent and volatility of wind makes wind energy uncontrollable, which is always a big problem in wind power technology. The key to solve the problem of energy and the long-term stable operation of wind farm is to improve the accuracy of wind power prediction. The establishment of neural network model is highly dependent on the training data, the selection of training data should include a wide range of selection, so the model has a stronger generalization ability; At the same time, it is necessary to ensure that the model has higher prediction accuracy. In view of the problem that the ordinary prediction model is difficult to meet the requirements under special weather, This paper uses case-based reasoning technology to improve the prediction model of wind power. The main research content of this paper is to summarize the current situation, development trend and existing problems of wind power generation at home and abroad. The significance and necessity of wind power prediction are introduced. At the same time, the research status of wind power prediction model is summarized and analyzed. According to the national standard GB/T18709-2002 and related reference supplement, the data are checked, filled and corrected. The variation law of wind measurement data itself is analyzed by using the measured data of a wind farm in Shanxi Province. The relationship between the prediction error of the model and the influence factors of the wind power is analyzed. The necessary selection of the evaluation index of the prediction model is given.) in this paper, a forecasting method of wind power based on fuzzy clustering analysis is proposed. First, subtraction clustering is used to determine the clustering number and center of fuzzy C-means. The data clustering analysis of the actual operation of wind farm is adopted, and then the neural network model is trained, which not only fully considers the characteristics of the sample space, but also makes the model have pertinence. In order to improve the prediction accuracy of wind power under special weather, this paper analyzes the prediction effect of the common model and the actual wind data of wind farm. In this paper, a case-based reasoning (CBR) model for predicting wind power in special weather is established, and the K-nearest neighbor algorithm based on fuzzy clustering and particle swarm optimization (PSO) is used for case retrieval. The speed and precision of retrieval are improved. A large number of simulation experiments are carried out using the actual data of a wind farm in Shanxi Province, and the simulation results are compared with the GRNNX LSSVMU GABP model. The prediction error has been improved in different degrees, especially when the prediction data is abrupt, the effect is more obvious, which verifies the effectiveness of the method, and provides a feasible method for wind power prediction under special weather conditions.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM614;TP311.13

【参考文献】