大型风电场分组建模方法及其在功率预测中的应用
发布时间:2018-01-11 09:33
本文关键词:大型风电场分组建模方法及其在功率预测中的应用 出处:《华北电力大学》2015年硕士论文 论文类型:学位论文
更多相关文章: 分组建模 聚类算法 风电场分组个数 轮廓系数 霍普金斯统计量 风电场功率预测
【摘要】:风电固有的波动性影响电力系统的安全、稳定和经济运行,是大规模风电并网运行的主要挑战。风电场功率预测是解决该问题的必要手段之一。利用场内某代表位置的风况来映射整个风电场的输出功率是当前大多风电场功率预测采用的方法,但对于大规模风电场,此方法难以保证精度;若对每台机组进行建模预测,将导致预测计算时间过长,无法满足电力系统对功率预测的要求。因此,研究既能提高预测精度,又保证计算效率的风电场功率预测方法,是大型风电场功率预测领域的关键问题之一。基于聚类算法研究了风电场分组功率预测方法,主要工作包括:(1)研究了风电场分组建模的影响因素。以实测风速、实测功率及二者组合作为模型输入,分析其对风电场分组功率预测精度的影响程度。通过分析可得风速是影响分组效果的主要因素,以风速作为输入变量,预测精度波动范围较小而功率变量会使预测精度波动范围较大。(2)提出了用于确定风电场分组个数的指数。利用提出的轮廓系数和霍普金斯统计量指数,分别从定性和定量的角度确定风电场分组个数,为建立风电场分组模型奠定基础。结果表明:霍普金斯统计量方法效果更好,定量的判定标准保证了确定风电场分组个数的准确性和高效性。(3)提出了一种可提高功率预测精度的风电场分组模型。建立了基于K-means、FCM、SOM、GA-蚁群和谱聚类五种聚类算法的风电场分组模型,通过识别机组的风况特征和发电特征的相似性将大型风电场分成不同的机组群,并利用相关性分析法选择组内代表机组。结果表明:五组聚类结果合理;谱聚类计算效率最高,而GA-蚁群计算时间最长。(4)建立了具有较强适用性的风电场分组功率预测模型。利用多台代表机组位置的风况,预测整场输出功率,比较各模型的预测精度,并对其适用性进行分析。结果表明:风电场分组模型可以有效降低大型风电场功率预测的计算维度,显著提高预测精度,其中,SOM和谱聚类模型较未分组预测精度分别提高1.67%和1.59%;计算效率最高的谱聚类适用于大型风电场的分组建模,预测精度最高的SOM适用于中型风电场,二者可为电力系统调度提供更准确的功率预测信息。
[Abstract]:The inherent volatility of wind power affects the safety, stability and economic operation of the power system. Wind farm power prediction is one of the necessary methods to solve this problem. Using wind condition of a representative position in the field to map the output power of the whole wind farm is currently most of the wind farms. The method of power prediction. However, for large-scale wind farms, this method is difficult to ensure accuracy. If each unit is modeled and forecasted, the prediction time will be too long, which can not meet the power forecasting requirements of power system. Therefore, the research can improve the accuracy of prediction. It is one of the key problems in the field of large-scale wind farm power prediction to ensure the computational efficiency of wind farm power prediction method. Based on clustering algorithm, the wind farm grouping power prediction method is studied. The main work includes: (1) study the influence factors of wind farm grouping modeling, take the wind speed, power and combination of wind farm as the input of the model. The influence of wind speed on the accuracy of wind farm grouping power prediction is analyzed. The wind speed is the main factor affecting the grouping effect, and the wind speed is taken as the input variable. The index used to determine the number of wind farm groups is proposed. The contour coefficient and Hopkins statistic index are used to determine the number of wind farm groups. Determine the number of wind farm grouping from the qualitative and quantitative point of view, for the establishment of wind farm grouping model, the results show that the Hopkins statistical method is more effective. The quantitative criterion ensures the accuracy and efficiency of determining the number of wind farm groups. (3) A wind farm grouping model which can improve the accuracy of power prediction is proposed. Based on K-means, a wind farm grouping model is established. The wind farm grouping model based on FCM-SOMGA- ant colony and spectral clustering algorithm is proposed. The wind farm is divided into different units by identifying the wind characteristics of the units and the similarity of the generation characteristics. The representative units in the group are selected by correlation analysis. The results show that the cluster results of five groups are reasonable; Spectral clustering calculation efficiency is the highest, and GA-ant colony calculation time is the longest. (4) A wind farm grouping power prediction model with strong applicability is established. Forecast the output power of the whole field, compare the prediction accuracy of each model, and analyze its applicability. The results show that the wind farm grouping model can effectively reduce the calculation dimension of large-scale wind farm power prediction. The prediction accuracy was improved significantly, in which SOM and spectral clustering models improved the prediction accuracy by 1.67% and 1.59, respectively, compared with those of ungrouped models. Spectral clustering with the highest computational efficiency is suitable for grouping modeling of large-scale wind farms and SOM with the highest prediction accuracy is suitable for medium-sized wind farms. Both of them can provide more accurate power prediction information for power system dispatching.
【学位授予单位】:华北电力大学
【学位级别】:硕士
【学位授予年份】:2015
【分类号】:TM614
【参考文献】
相关期刊论文 前7条
1 黄梅;万航羽;;在动态仿真中风电场模型的简化[J];电工技术学报;2009年09期
2 魏巍;王渝红;李兴源;行舟;李小虎;顾威;;大型风电场建模及接入甘肃嘉酒电网仿真分析[J];电力系统自动化;2009年18期
3 王文生;王进;王科文;;SOM神经网络和C-均值法在负荷分类中的应用[J];电力系统及其自动化学报;2011年04期
4 袁贵川;倪林;陈颖;;大型风电场等值建模与仿真分析[J];广东电力;2010年11期
5 蔡晓妍;戴冠中;杨黎斌;;谱聚类算法综述[J];计算机科学;2008年07期
6 朱连江;马炳先;赵学泉;;基于轮廓系数的聚类有效性分析[J];计算机应用;2010年S2期
7 万志华,欧阳为民,张平庸;一种基于划分的动态聚类算法[J];计算机工程与设计;2005年01期
相关硕士学位论文 前1条
1 阎洁;风电场功率预测不确定性分析方法及其应用研究[D];华北电力大学;2012年
,本文编号:1409016
本文链接:https://www.wllwen.com/kejilunwen/dianlilw/1409016.html
教材专著