基于运行数据的概率聚类风电场分群研究

Probabilistic Clustering for Wind Farms Based on the Operation Data Analysis

风电机组在实际运行时,受尾流效应等因素影响,运行状态并不相同。为提高风电场模型的精度,解决运行工况对分群结果的影响问题,提出了一种概率聚类分群算法,并从地理位置分布和仿真结果两个角度验证了其合理性。该方法在传统的K-means算法基础上,综合考虑各种风速工况发生的比例,以概率最大的机组分群为最终结果,并用轮廓值函数加以验证。此算法得到的分群结果可应用于各种风速的情况,对机组分布不规则的风电场有很好的效果,为以后风场建模的使用提供了极大的方便。

In the actual operation, running state of wind turbines was not the same, as influenced by some factors such as wake effect. In order to improve the precision of wind farms model and solve the problem of operation condition onthe result of clustering, a probabilistic clustering algorithm was proposed, and its rationality was verified from two aspects of geography distribution and the simulation results. The method was based on the traditional K-means algorithmand considered proportions of all kinds of working wind speed conditions. The final result was selected depending on the largest probability of groups and verified with contour value function. The algorithm of clustering applied to fullrange of wind speed has great effect on the irregular distribution of wind farm and provides a great convenience for the use of wind field modeling.