风电监测信息海量数据挖掘与特性信息提取

发布时间：2018-04-15 00:06

本文选题：数据挖掘技术 + 风电　；参考：《山东大学》2015年硕士论文

【摘要】：风电已经成为发展最迅速、技术最为成熟的新型清洁能源,我国非常重视风电的发展,目前风电装机容量已经达到了世界第一位。但是由于风电出力具有很强的随机性和波动性,风电的大规模并网对电力系统调度运行的调频、调峰、调度等多个方面产生了很大影响,增加了电网调度运行的难度。因此,只有从多个时间和空间尺度对风电出力进行分析,全面掌握风电的出力特性,才能为风电并网后电力系统各个方面的分析研究提供参考和关键数据支撑,更好地进行风电大规模并网消纳。本文针对现有数据挖掘和特性信息提取技术在处理海量风电出力数据提取风电出力特性时的缺点和不足进行了补充和完善,主要研究了数据归约、数据聚类、数据统计、相关性分析四个方面的数据挖掘和特性信息提取技术,从而建立了较为完整的应用在风电出力特性分析中的海量数据数据挖掘和特性信息提取技术体系,通过上述技术实现了对风电海量数据的特性信息提取和深度挖掘。本文以冀北电网风力发电的海量监测数据为依据,采用多时空尺度数据挖掘和特性信息提取技术,对各个风电场出力数据进行了详细的整理和统计,提出了多个风电出力特性分析评价指标,并按照风电出力的随机性、波动性、相关性3种特性将评价指标进行归类,从而形成了用于评价风电出力特性规律的多时空尺度指标体系,大大提高了现有评价指标的精确性和完整性。时间尺度上包括了分钟级、小时级、日级等不同时间尺度,空间上从单个风场、风电场群延伸到整个冀北风电基地。风电出力随机特性分析部分统计了风速和风电出力的概率分布并建立了相应的概率密度分布函数。统计了风速和风电出力的预测误差分布,分析了现有风电出力预测误差分布模型的缺点和不足并利用非参数估计法进行了改进,建立了风电出力预测误差的分区分布函数,进而结合风电出力点预测值得到了风电出力预测波动区间。风电出力波动特性分析部分从不同的时间尺度出发统计分析了风电出力单步变化率、风电高风险爬坡事件、风电出力极值、风电峰谷差贡献率以及风电中长期出力典型日模式,可以为大规模风电并网后的电力系统调频、调峰和风险评估分析等多个领域提供参考。风电出力相关特性分析部分分为单个风电场出力自相关和多个风电场出力互相关两类展开分析。利用自相关系数和风电出力区间转换概率统计分析了单个风电场出力的自相关特性。利用互相关系数分析了位于空间不同区域的多个风电场出力时间序列间以及风电场与风电基地出力之间的互相关特性。利用风电场之间的出力变化率,出力标准差和出力同时率分析了风电场的空间集群效应。
[Abstract]:Wind power has become the most rapid development, the most mature technology of the new clean energy, our country attaches great importance to the development of wind power, wind power installed capacity has reached the first in the world.However, due to the strong randomness and volatility of wind power generation, the large-scale grid connection of wind power has a great influence on the frequency modulation, peak-shaving, dispatching and other aspects of power system dispatching, which increases the difficulty of power grid dispatching.Therefore, only by analyzing wind power from multiple time and space scales, and fully master the characteristics of wind power, can we provide reference and key data support for the analysis and study of all aspects of power system after wind power grid connection.Better wind power large-scale grid absorption.In this paper, the shortcomings and shortcomings of the existing data mining and characteristic information extraction techniques in dealing with massive wind power output data extraction are supplemented and improved. Data reduction, data clustering and data statistics are mainly studied in this paper.Correlation analysis of four aspects of data mining and feature information extraction technology, so as to establish a more complete application in wind power performance analysis of mass data mining and feature information extraction technology system,The characteristic information extraction and depth mining of wind power magnanimity data are realized by the above technology.Based on the massive monitoring data of wind power generation in the north Hebei power grid, the data of each wind farm are sorted out and counted in detail by using the techniques of multi-space-time scale data mining and characteristic information extraction.Several evaluation indexes of wind power output characteristics are put forward, and the evaluation indexes are classified according to the randomness, fluctuation and relativity of wind power output.Therefore, a multi-space-time scale index system for evaluating the characteristics of wind power output is formed, which greatly improves the accuracy and integrity of the existing evaluation indexes.The time scale includes minute scale, hour scale, day scale and so on. In space, the wind farm group extends from a single wind field to the whole wind power base in the north of Hebei Province.The probability distribution of wind speed and output force is analyzed and the probability density distribution function is established.The prediction error distribution of wind speed and wind power output is analyzed, the shortcomings and shortcomings of the existing wind power output prediction error distribution model are analyzed, and the non-parametric estimation method is used to improve the prediction error distribution function of wind power output prediction error.Furthermore, the fluctuation interval of wind power output prediction is obtained by combining the predicted value of wind power output point.The characteristics of wind power output fluctuation are analyzed in terms of different time scales, such as wind power output single step change rate, wind power high risk climbing event, wind power output extreme value, wind power peak and valley difference contribution rate and typical day model of wind power output.It can be used as a reference for power system frequency modulation, peak-shaving and risk assessment analysis after large-scale wind power grid connection.The analysis of wind power output correlation is divided into two types: single wind farm output autocorrelation and multiple wind farm output correlation analysis.The autocorrelation characteristics of a single wind farm are analyzed by means of autocorrelation coefficient and transition probability of wind power output interval.The cross-correlation characteristics of multiple wind farm output time series located in different regions of space and between wind farm and wind power base are analyzed by using the correlation number.The spatial cluster effect of wind farm is analyzed by using the variation rate of output force, the standard deviation of output force and the simultaneous rate of output force between wind farms.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM614

【参考文献】