风电功率爬坡事件预测方法研究

发布时间：2018-06-10 10:32

本文选题：风电功率爬坡事件 + 预测　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：在含有高比例风电的电力系统中,风电功率爬坡事件(即风电场输出功率在短时间内的大幅度波动现象)对电网的冲击已不容忽视。它会直接导致电力系统发用电不平衡,威胁电力系统的安全稳定运行,甚至造成严重的电力系统停电事故,给社会经济造成很大的损失。掌握风电功率爬坡事件的发生规律,进而对其及时、准确的预测已成为风电并网过程中亟待解决的问题。论文从风电功率爬坡事件的分布特征、影响因素及预测方法三个方面对其展开了探究。主要工作和成果如下:(1)研究了风速的周期特性风速的特性决定风电功率的特性。基于小波分解方法对随机波动的风速时间序列进行了拆解,从周期的角度对其展开了分析,并提出了周期强度评价指标PI(Periodicity Intensity)和RPI(Relative Periodicity Intensity),对各周期分量的显著特征进行了定量描述,初步建立了风速波动性与周期性的内在联系。(2)分析了风电场输出功率爬坡事件的分布特征及影响因素分析了风电场输出功率爬坡事件的分布规律。结果表明:不同风电场输出功率爬坡事件的分布特点及主要影响因素有明显不同。针对这一问题,建立了一套具有普适性的风电功率爬坡事件影响因素分析方法,用于确定不同风电场的输出功率爬坡事件的主导因素,为其预测提供基础。(3)建立了基于正交实验与支持向量机的风电功率爬坡事件预测模型基于风电功率爬坡事件影响因素分析方法以及风电功率爬坡事件与各气象要素之间的联系,建立了基于正交实验与支持向量机的风电功率爬坡事件预测模型(OT-SVM)。该模型基于数值天气预报(NWP),通过引入正交实验环节为各风电场的预测模型选取最合适的气象要素输入量。经算例验证:OT-SVM模型能够有效提高预测精度,且具有普适性,能够充分考虑不同风电场输出功率爬坡事件发生特性的差异,针对每个风电场制定最适合的预测策略。(4)建立了基于小波分解与自回归滑动平均的风电功率爬坡事件预测模型基于对风电场历史输出功率的时间序列分析,建立了基于小波分解与自回归滑动平均的风电功率爬坡事件预测模型(WT-ARMA)。针对预测过程中“全面性”与“准确性”无法同时满足的问题,提出了“单支预测,统筹决策”的预测策略。经算例验证:WT-ARMA模型能够有效解决高捕获率与高准确率不可兼得的问题,实现了风电功率爬坡事件的全面且准确的预测。且该模型无需风速、风向等气象要素的数值天气预报值作为输入,有效克服了数值天气预报误差对风电功率爬坡事件预测结果的影响。
[Abstract]:In the power system with a high proportion of wind power, the impact of wind power climbing event (i.e. the large fluctuation of wind farm output power in a short period of time) on the power grid can not be ignored. It will directly lead to the imbalance of power system, threaten the safe and stable operation of power system, even cause serious power system blackout, and cause great losses to the society and economy. It is an urgent problem to grasp the occurrence law of wind power climbing event and to predict it timely and accurately in the process of wind power grid connection. In this paper, the distribution characteristics, influencing factors and prediction methods of wind and electric power climbing event are discussed. The main work and results are as follows: (1) the periodic characteristics of wind speed determine the characteristics of wind power. Based on the wavelet decomposition method, the wind speed time series of random wave is disassembled and analyzed from the point of view of period, and the index of periodic strength evaluation, Pi Periodicity Intensityand RPI / Relative Periodicity Intensityy, are put forward, and the significant characteristics of each cycle component are quantitatively described. The inherent relation between fluctuation and periodicity of wind speed is established. (2) the distribution characteristics of wind farm output power climbing events and the distribution law of wind farm output power climbing events are analyzed. The results show that the distribution characteristics and main influencing factors of the output power climbing events of different wind farms are obviously different. In order to solve this problem, a universal analysis method of wind and electric power climbing factors is established, which can be used to determine the dominant factors of wind farm output power climbing events. The prediction model of wind power climbing event based on orthogonal experiment and support vector machine is established. Based on the analysis method of factors influencing wind power climbing event and the relation between wind power climbing event and various meteorological elements, The prediction model of wind power climbing event based on orthogonal experiment and support vector machine is established. The model is based on the numerical weather forecast (NWP) and the orthogonal experiment is introduced to select the most suitable meteorological input for each wind farm prediction model. An example shows that the OT-SVM model can effectively improve the prediction accuracy and is universal, and can fully take into account the different characteristics of different wind farm output power climbing events. For each wind farm, the most suitable prediction strategy is made. (4) the wind power climbing event prediction model based on wavelet decomposition and autoregressive sliding average is established based on the time series analysis of the historical output power of the wind farm. The prediction model of wind power climbing event based on wavelet decomposition and autoregressive moving average is established. Aiming at the problem that "comprehensiveness" and "accuracy" can not be satisfied simultaneously in the process of prediction, this paper puts forward the forecasting strategy of "single prediction and overall decision making". It is verified by an example that the problem of high capture rate and high accuracy can be solved effectively by using the 1: WT-ARMA model, and the comprehensive and accurate prediction of wind power climbing event is realized. The model does not need the numerical weather forecast value of wind speed, wind direction and other meteorological elements as input, which effectively overcomes the influence of numerical weather forecast error on the forecast result of wind power climbing event.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM614

【参考文献】