含风电场的机组组合二阶段随机模型及其改进算法

提出一种含风电场的机组组合二阶段随机规划模型,将风电功率作为随机变量处理,目标函数包含常规机组发电成本和切负荷惩罚费用,由于风电功率存在多种可能的情景,后一种费用采用期望值形式,同时提出一种求解二阶段模型的SAA-自适应多切割L形算法,具体为首先基于抽样平均逼近(SAA)理论,将随机模型转换成确定性模型,然后提出一种自适应多切割L形算法求解。求解中引入全局辅助变量实现迭代过程中历史最优切割信息的保存,并设置主模型约束条件数上限保证模型始终具有较小的规模。与传统单切割和多切割L形算法相比,所提出算法的迭代次数介于两者之间,但计算时间要少于两者。最后通过3机、10机和100机算例在不同数量的风电情景下仿真计算,结果表明本文模型可以有效处理风电随机性,SAA-自适应多切割L形算法在样本数量较大的情况下保持了良好的收敛性和可靠性。

Two-Stage Stochastic Model of Unit Commitment with Wind Farm and an Improved Algorithm

This paper introduces two-stage stochastic model of unit commitment with wind farms. The objective cost of the model is divided into generating cost of thermal units and load shedding penalty cost. Due to the randomness of wind power, the latter cost is in the form of expectation. At the same time, a SAA-adaptive multi-cut L-shaped algorithm is proposed, where the sample average approximation (SAA) theory translates the proposed model into a certain one and the Adaptive multi-cut L-shaped algorithm solves the model. A kind of global assist variables is employed to save history optimal cuts and set upper limit of the main model’s constraint number. The iteration number of the proposed one is between the single-cut and multi-cut L-Shaped methods, while the computing time is the least. Finally, 3-uint, 10-unit and 100 unit systems are simulated with different sample numbers. The results verify the convergence and validity of the proposed model, and show the correctness of dealing with uncertainty of wind power with more samples.