基于切片采样的风力发电并网系统概率潮流计算

Probabilistic Load Flow Calculation Based on Slice Sampling for Wind Farms Integration System

在基于马尔科夫链蒙特卡洛(MCMC)模拟法的概率潮流计算方法中,被广泛应用的Gibbs采样算法需要进行大量复杂的迭代运算才能得到较精确的计算结果。针对该算法的缺陷,提出基于切片采样(slice sampling)算法的MCMC方法,并应用于风力发电并网系统概率潮流计算中。首先,采用加权高斯混合分布(WGMD)对风电场出力进行建模;然后,通过切片采样算法对风电场出力的概率分布进行采样,从而构建出风电场出力的样本空间;最后,对样本空间中的每组采样点进行潮流计算,并在含有风电模型的IEEE 39节点系统中与Gibbs采样算法得到的结果进行比较。结果表明:切片采样算法能够显著提高传统MCMC方法的计算准确度;同时,在与Gibbs算法采样迭代次数相同的情况下,切片采样算法所生成的马尔科夫链可以更快、更稳定地收敛于平稳分布。

Gibbs sampling algorithm that is widely used in Markov Chain Monte Carlo（ MCMC） simulation method suffers from complicated sampling iterations when accurate results from probabilistic load flow is required. According to the defect,an improved MCMC method based on Slice sampling is proposed in this paper and is integrated into probabilistic load flow algorithm for wind farms integration system. The probabilistic model of wind farm outputs is firstly constructed by weighted Gaussian mixture distribution（ WGMD）. Then,the sample space of wind farm outputs is obtained by Slice sampling from the WGMD of wind farm outputs. Finally,the samples from the sample space of wind farm outputs are calculated by load flow and the results of these two sampling methods are compared in IEEE 39-bus system. It is shown that the proposed method can distinctly improve the calculation accuracy of MCMC method. Additionally,the Markov Chain generated by Slice sampling can reach a stationary distribution more quickly and stably than Gibbs sampling with the same iterations.