Wake effect parameter calibration with large-scale field operational data using stochastic optimization
Wake effect parameter calibration with large-scale field operational data using stochastic optimization
复制标题
DOI：
10.1016/j.apenergy.2023.121426
复制
发表时间：
2023-10
期刊：
影响因子：
11.2
通讯作者：
Pranav Jain;S. Shashaani;E. Byon
中科院分区：
文献类型：
--
作者：
Pranav Jain;S. Shashaani;E. Byon
本研究旨在展示随机优化在工程尾流模型的有效和鲁棒参数校准中的应用。工程尾流模型通常采用尾流效应参数的标准值来预测功率，但最近的一些研究表明，这些值并不能准确预测。该方法利用实际风电场的运行数据对尾流效应参数进行估计，通过信赖区优化使尾流模型的预测误差最小化。为了进一步提高计算效率，我们实现了分层自适应采样。我们采用决策树对数据进行分层，并提出了两种使抽样预算适应构造层的方法：动态权值预算分配和固定权值预算分配。我们扩展了我们的分析，以确定湍流强度和尾迹衰减系数之间的函数关系。我们的实验表明，尾流参数或湍流强度与尾流衰减系数之间的函数关系可能需要根据特定风电场的运行数据进行调整（从假设的标准值），以更好地表征尾流效应。
This study aims to show the application of stochastic optimization for efficient and robust parameter calibration of engineering wake models. Standard values of the wake effect parameters are generally used to predict power using engineering wake models, but some recent studies have shown that these values do not result in accurate prediction. The proposed approach estimates the wake effect parameters using operational data available from actual wind farms to minimize the prediction error of the wake model by using trust-region optimization. To further improve computational efficiency, we implement stratified adaptive sampling. We employ decision trees to stratify the data and propose two ways of adapting the sampling budget to the constructed strata: budget allocation with dynamic weights and fixed weights. We extend our analysis to determine the functional relationship between the turbulence intensity and wake decay coefficient. Our experiments suggest that wake parameters or a functional relationship between turbulence intensity and wake decay coefficient may need adjustments (from assumed standard values) for a particular wind farm using its operational data to characterize the wake effect better.