精确的短期风速预测对可靠安全的电力系统运行很重要。传统的预测方法没有考虑空间相邻风电场的信息。然而,多个风电场的风速在时间和空间上是相关的。该文给出了一个采用贝叶斯克里金卡尔曼模型的短期风速预测方法。由主克里金函数构成的空域结构使用贝叶斯层次结构进行建模,同时应用状态空间模型对时域动态性进行建模。采用计算速度更有效的变分贝叶斯方法来逼近推断和学习模型参数。在公开的多风电场数据集上评估提前1h的风速预测性能,与持续预测算法进行比较的结果显示了该文提出的方法在均方根误差评价指标上的改善。
Accurate short-term wind speed forecasting is critical to reliable and secure power system operations.Traditional forecasting approaches did not take account of the spatial neighborhoods information.However,wind speeds in multiple wind farms are correlated both in time and space.This paper introduced a short-term wind speed forecasting approach using Bayesian KrigedKalman model in which spatial structure with principal Kriging functions was modeled by a Bayesian hierarchical structure and time dynamic component was modeled by a state space process.Variational Bayesian method was applied to inference approximately and learn parameters of the model,which was a computationally efficient deterministic approach.One hour ahead forecasting was evaluated on publicly available wind data of multiple wind farms.The results were compared to persistence forecasting approach to show its improvement in terms of root mean square errors.