基于频繁模式挖掘的风电爬坡事件统计特性建模及预测

Frequent Pattern Mining Based Modeling and Forecasting for Statistical Characteristics of Wind Power Ramp Events

风电爬坡事件的统计特性建模和精准预测有利于电网的安全稳定运行。文中首先通过参数分辨率自适应算法对大型历史风电数据库进行爬坡事件检测,得到风电爬坡事件的历史学习集。对该学习集进行数据挖掘,建立了单个爬坡事件的起点、终点、持续时间以及爬坡间隔的多属性联合统计特性模型,并得到爬坡事件的基本模式。通过关联规则算法建立了多个相邻爬坡事件之间的自相关性统计特性模型。在此基础上,提出了爬坡事件序列预测算法的基本概念和模型。算例结果表明,所提算法能够更为直观地描述爬坡事件的统计特性,且基于事件序列的预测算法能够较好地进行日前的爬坡预测。

The statistical characteristic modeling and accurate forecasting of wind power ramp events are conducive to the safe and stable operation of the power grid. In this paper, first of all, the parameter and resolution adaptive algorithm is used to detect the ramp events in a large-scale historical wind power database, and the history learning set of wind power ramp events is obtained.Data mining is carried out on this learning set to establish a model with multi-attribute joint statistical characteristics of the starting point, ending point, duration, interval ramp for a ramp event, and the basic mode of ramp events is got. The modeling of autocorrelation statistical characteristics between multiple adjacent ramp events is established by using the association rule algorithm. On this basis, the basic concept and model of the forecasting algorithm of ramp event sequence are proposed. The case results show that the statistical characteristics of ramp events can be described more intuitively by the proposed algorithm, and the events sequence based forecasting algorithm can provide better performance for the day-ahead ramp forecasting.