基于日内时段划分的风电出力序列模拟方法

To address the limitation of Markov chain Monte Carlo (MCMC) in preserving anti-peak-shaving characteristics, this paper proposes a time-segmented wind power simulation method. It clusters daily output curves by month using K-means, segments days into equal intervals, constructs state transition matrices per segment, and enhances them with randomness and fluctuation components. Case studies show superior performance in PDF, autocorrelation, and anti-peak-shaving retention.

为了克服马尔可夫-蒙特卡洛方法(Markov chain Monte Carlo,MCMC)在生成风电出力序列时无法较好保留反调峰特性的局限性,提出了一种基于日内时段划分的风电出力时序模拟方法.首先,对每个月的日出力曲线进行K-means聚类;其次,将一天划分为等长的若干时间段,将同类自然日的相同时间段拼接为新序列,根据出力值划分包含等样本数量的状态区间,并求新序列的状态转移矩阵;最后,对状态转移矩阵添加随机性和波动分量模拟波动性,再逐时段生成日曲线.通过案例对比证明,该方法不仅在基本的概率密度和自相关函数指标上优于传统的K-means MCMC法,而且更好地保留了反调峰特性,验证了该方法的有效性.论文研究可为电力系统规划中的风电数据样本生成提供参考.