Isotonic Quantile Regression Averaging for uncertainty quantification of electricity price forecasts
Arkadiusz Lipiecki and Bartosz Uniejewski
量化预测模型的不确定性对于评估和减轻与数据驱动决策相关的风险至关重要,特别是在电力市场等动荡领域。 机器学习方法可以提供高度准确的电价预测,对于告知市场参与者的决策至关重要。 然而,这些模型往往缺乏不确定性估计,这限制了决策者避免不必要风险的能力。 在本文中,我们提出了一种新的方法来从点预测集合中生成概率预测,称为同位素分位数回归平均(iQRA)。 在Quantile Regression Averaging(QRA)的既定框架的基础上,我们引入了随机顺序约束,以提高预测准确性,可靠性和计算成本。 在一项对德国电力市场的广泛预测研究中,我们表明iQRA在可靠性和清晰度方面始终优于最先进的后处理方法。 它在多个置信水平之间产生经过良好校准的预测间隔,为所有基准方法提供卓越的可靠性,特别是基于覆盖的构象预测。 此外,同位素正则化降低了分位数回归问题的复杂性,并为变量选择提供了一种无参数的方法。
Quantifying the uncertainty of forecasting models is essential to assess and mitigate the risks associated with data-driven decisions, especially in volatile domains such as electricity markets. Machine learning methods can provide highly accurate electricity price forecasts, critical for informing the decisions of market participants. However, these models often lack uncertainty estimates, which limits the ability of decision makers to avoid unnecessary risks. In this paper, we propose a novel ...