基于混沌理论的风电功率实时预测研究

发布时间：2018-01-13 03:34

本文关键词：基于混沌理论的风电功率实时预测研究　出处：《东北电力大学》2017年硕士论文　论文类型：学位论文

【摘要】：近十几年,我国风能的开发利用处于快速发展阶段,风电装机容量以及风电并网情况增长较快,但是因为近地风具有间歇特征,风电功率在一定程度上具有随机特征,风电并网后,风电功率的大幅度波动变化会对电力系统正常运行、合理分配调度等方面造成影响。掌握风电功率的时空分布规律、对风电功率进行较为准确的预测以及对预测误差进行分析研究,这对风能大规模的开发利用具有重要意义。本文以东北风电场的有功功率数据为例,分析研究风电功率的时空分布特征。为了客观认识风电功率的波动变化特征、研究风电功率混沌特征的时空分布特点,本文提出了衡量风电功率混沌特征的量化指标—滚动最大Lyapunov指数,并对混沌特征的时间和空间分布特征进行分析验证;利用自相关系数图和周期图验证了风电功率序列中存在周期性分量这一特征,然后利用傅里叶变换提取周期分量,并且利用递归图法验证了剩余分量具有混沌特征;利用集合经验模态分解对风电功率时间序列进行降噪处理,然后分析降噪后的时间序列的长程相关性和分形特征。针对风电功率预测的研究,本文提出了基于混沌理论的三种不同的预测方法:基于局域一阶加权法的风电功率超短期预测、校正的Lyapunov指数多步预测模型以及实时提取周期分量的组合预测模型。其中,基于局域一阶加权法的风电功率超短期预测模型是以距离作为邻近相点的选择判据构建预测模型;校正的Lyapunov指数多步预测模型则以Lyapunov指数预测模型为基础,并对滚动预测时的预测值进行校正;实时提取周期分量的组合预测把序列分解为周期分量和剩余分量,然后把两个分量各自的预测值合并后加入到原序列中,针对新的风电功率序列再次分解和预测。针对风电功率预测误差的研究,本文以混沌理论为基础,基于东北风电场的有功功率数据,分析了风电功率在实时预测时的单步预测误差的概率分布,研究了风电功率预测误差与预测步数的关系、预测误差与风电场出力情况的关系以及预测误差与装机容量之间的关系。针对风电功率的多步预测,建立了基于VB语言的风电场有功功率预测系统。该预测系统可直接从实时监测系统中读取功率信息,其建模简单、运算速度较快,能够满足在线使用的要求。该系统适用于超短期风电功率预测,尤其适用于历史数据量较少、气象信息不足、仅有风电功率序列的风电场。
[Abstract]:In recent ten years, the development and utilization of wind energy in China is in a rapid development stage, wind power installed capacity and wind power grid has increased rapidly, but because of the intermittent characteristics of near-ground wind. Wind power has random characteristics to some extent. After wind power is connected to grid, the large fluctuation of wind power will affect the normal operation of power system. The reasonable allocation of dispatch and other aspects of the impact. Grasp the distribution of wind power in time and space, wind power more accurate prediction and analysis of the prediction error. This is of great significance to the large-scale development and utilization of wind energy. This paper takes the active power data of northeast wind farm as an example. The temporal and spatial distribution of wind power is analyzed and studied. In order to understand the fluctuation and variation of wind power, the temporal and spatial distribution characteristics of wind power chaos are studied. In this paper, the rolling maximum Lyapunov exponent, a quantitative index to measure the chaotic characteristics of wind power, is proposed, and the temporal and spatial distribution characteristics of the chaotic features are analyzed and verified. The existence of periodic components in wind power series is verified by autocorrelation and period diagrams, and then the periodic components are extracted by Fourier transform. The remaining components are proved to be chaotic by recursive graph method. The wind power time series is de-noised by means of set empirical mode decomposition, and the long range correlation and fractal characteristics of the time series after noise reduction are analyzed. The prediction of wind power is studied. In this paper, three different prediction methods based on chaos theory are proposed: wind power ultra-short-term prediction based on local first-order weighting method. The corrected multistep prediction model of Lyapunov exponent and the combination prediction model of extracting periodic components in real time. Based on the local first-order weighting method, the wind power ultra-short-term prediction model is constructed using distance as the selection criterion of adjacent phase points. The corrected Lyapunov exponent multistep prediction model is based on the Lyapunov exponent prediction model, and the prediction value of rolling prediction is corrected. The combined prediction of extracting periodic components in real time decomposes the sequence into periodic components and residual components, and then combines the predicted values of the two components and adds them to the original sequence. For the new wind power series decomposition and prediction. For wind power prediction error research, this paper based on chaos theory, based on the northeast wind farm active power data. The probability distribution of single-step prediction error of wind power in real time prediction is analyzed, and the relationship between wind power prediction error and the number of prediction steps is studied. The relationship between the prediction error and the wind farm output and the relationship between the prediction error and the installed capacity. The active power prediction system of wind farm based on VB language is established, which can directly read the power information from the real-time monitoring system. The system can meet the requirements of on-line use. The system is suitable for the prediction of ultra-short-term wind power, especially for wind farms with less historical data, insufficient meteorological information and only wind power series.
【学位授予单位】：东北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM614

【参考文献】