摘要
传统的概率分布模型在不确定性建模方面主要依赖于模型的选择,但假定的单一模型往往无法准确刻画随机量的复杂变化规律。除此之外,参数概率模型对随机量之间时间、空间相关性的描述不够,而基于连接函数等相关性建模方法在描述多个随机量的相关性方面又过于复杂,给模型实际应用增加了难度。通过采用一种数据驱动的不确定性建模方法,提出了一种数据驱动下的两阶段机组组合模型,基于非参数狄利克雷过程高斯混合模型(DPGMM)和变分贝叶斯推断(VBI)方法来描述风电、光伏和负荷的不确定性,考虑了多风电场和各节点负荷间的相关性,并采用传统数学优化方法对所构建的机组组合模型进行求解。最后,在IEEE-30节点算例系统上进行了仿真验证,结果表明,采用DPGMM模型能够较好地拟合随机量的概率分布并描述随机量之间的相关性。
The traditional probability distribution model mainly relies on the choice of model in uncertainty modelling,but the assumed single model often cannot accurately describe the complex variations of random quantities.In addition,the parametric probability model does not adequately describe the temporal and spatial correlations between random variables,while the correlation modelling methods like Copula functions are too complicated in describing the correlations of multiple random variables,which adds difficulty to the practical applications of the model.Taking a data-driven uncertainty modelling method,a data-driven two-stage unit commitment model is proposed.The non-parametric Dirichlet process Gaussian mixture model(DPGMM)and the variational Bayesian inference(VBI)method are used to describe the uncertainty of wind power,photovoltaics and load.Taking the correlations between multiple wind farms and the load of each node into consideration,the traditional mathematical optimization method is applied to solve the established unit commitment model.Finally,the simulation verification is carried out on the IEEE-30 node test system.The results show that the DPGMM model can fit the probability distribution of random quantities and describe the correlation between them.
作者
石立宝
翟放
SHI Libao;ZHAI Fang(National Key Laboratory of Power Systems in Shenzhen,Shenzhen International Graduate School,Tsinghua University,Shenzhen 518055,China)
出处
《综合智慧能源》
CAS
2022年第1期18-25,共8页
Integrated Intelligent Energy
基金
国家自然科学基金项目(51777103)。
