风电及资源相关的特性分析和建模工作是众多研究的基础。中国风电集群开发、远距离输送以及并网发电的发展模式，使其呈现与欧美不同的特点，难以直接借鉴国外已有的特性分析结论和经验。因此，首先需要构建一套适合集群风电发展、集中并网接入的中国风电特性分析指标和方法，以及满足风电运行调度需求的风资源快速模拟方法。基于上述背景和目标，本文主要从风电统计特性的综合分析评价方法、大范围风资源的时空分布物理特性分析及面向工程应用的建模方法两个方面展开研究。本文首先从风电与电网关系出发，结合风电自身特性，建立了逐层分类的多时空尺度风电统计特性评价指标体系。该指标体系涵盖风电自然特性及其与电网交互作用特性两大方面，在指标系统性、完整性方面有较大提升，可有效反应风电并网后可能对系统造成的影响，为系统调度、规划提供多角度的参考信息。之后论文综合考虑统计建模和数值模拟的不同特点，提出了可满足电力系统工程应用需求的大范围风资源空间分布物理特性的精细化分析建模方法。针对风资源的空间分布特性，研究首先在常规模型和分析方法基础上，引入最短距离聚类方法对传统反距离加权插值模型加以改进，对资源的空间分布可以进行更为准确的刻画；之后，本研究引入EOF分解方法以挖掘空间场典型特性，最终提出了以EOF分解为基础的标杆风机选择及发电量估算方法，为风电场运行中的相关应用提供更为科学的理论指导；最后，考虑到研究中采用的统计类建模方法在刻画地形对资源作用规律方面的不足，本文进一步引入质量守恒风场诊断模式，并通过对序列进行不同的预处理，有效说明了主导物理因素对资源基本作用特点，也为今后工程化的资源模拟工作提供了更多的物理规律参考。论文最后基于风资源的时序变化特征，分别采用直接和间接方法建立了资源超短期预测模型，提出了针对多数据源的数据筛选标准及预处理方法。直接建模过程中，本文以区域风资源超短期预测为目标，基于临近区域资源的相关特性，构建了多输入降维预测模型，提出的降维处理方法以及引入的数据筛选依据对实际工程应用中多数据源的建模工作具有重要指导意义。而在间接建模过程中，文章则将单序列预测模型和改进插值模型加以综合，实现了对任意区域的资源预测，并进一步通过与直接建模方法的比较分析这两类预测模型的不同特点以及应用场合。
The characteristic analysis and modeling of wind power and wind resource is the basis of different studies. The development patterns of wind power in China, which are cluster exploitation, long-range transportation and grid integration, make the characteristics of China wind power quite different from those in Europe and America. So it is not appropriate to directly refer to existing foreign conclusions and lessons. As a result, building a wind power characteristic evaluation index system which is suitable for the cluster development and centralized integration characteristics of wind power in China, and proposing a fast wind resource simulation method to meet the needs of wind power operation and dispatching is a first step. Based on the background and target, this paper mainly studies from two aspects: comprehensive evaluation method of wind power statistical characteristics and the analysis and engineered modeling method of wind resource spatial and temporal physical characteristics.Firstly, proceded from the relations between wind power and the grid, also combined with the characteristics of wind power itself, the paper builds an evaluation index system of multi-temporal and spatial wind power statistical properties. The index system could be divided into two categories which reflect the wind power natural features and the interaction with grid respectively. It is a more systemic and complete index system which could effectively reflect the effects wind power may bring to the grid after integration. Analysis using such indexes could offer muti-angle reference information in system dispatching and planning. Then the paper comprehensively considers the different characteristics of statistical and numerical modeling methods, and proposes a refined analyzing and modeling method on the large scale wind resource spatial and temporal physical characteristics, which could meet the power system engineering application requirements. Focusing on the spatial distribution characteristic of wind resource and based on the conventional model and analytical method, the research uses the shortest disrance clustering method to improve the traditional inverse distance weighted model, and the spatial distribution of wind resource can be more accurately described. Then, the research introduces a EOF decomposition method to discover the typical characteristics of resource spacial field, and finally proposes a EOF decomposition based benchmarking turbine selection and energy production estimation methods, which can provide a more scientific and theoretical guidance for the application in the wind farm operation. At last, taking into account the shortage of statistical modeling methods in describing terrain effects on wind resource, the paper further employs the mass conservation diagnostic mode. And by different pretreatment to the sequences, the study explains the regular influence pattern of dominant physical factors on resources, offering more physical reference for future engineered resource modeling work. Finally, based on the timing variation characteristics of wind resource, the paper builds the resource ultra-short-term forecasting model both by direct and indirect methods, and proposes a data screening criteria and a pretreatment method for multiple data sources. With the target of ultra-short-term forecasting of regional wind resources, and based on the relevant resource characteristics of the adjacent area, the paper builds a multi-input dimensionality reduction forecasting model. The proposed dimensionality reduction method and introduced data screening criteria is of great guiding significance to the multi-input model in practical application. In the indirect modeling process, the study comprehensively uses the single sequence forecast model and the improved interpolation model to forecast the wind resources for any region, and further analyzes the different characteristics and applications of these two models by comparation.
