讲座题目：Quantilesfor Curve-to-Curve Regression and Probabilistic Forecasting for DailyElectricity Load Curves 主讲人：姚琦伟教授 英国伦敦经济与政治科学学院 讲座时间：2020年9月8日（周二）下午16：00-18：00 参与方式： 本场报告将通过腾讯会议举办， 会议 ID：418188 690 会议直播:https://meeting.tencent.com/l/vST3VwyiD4TL 点击链接入会，或添加至会议列表： https://meeting.tencent.com/s/qoi2ao8hCsJi 主讲人简介： 姚琦伟教授是国际知名的统计学家，一直从事统计学的教学和科研工作，主要研究领域为：时间序列分析、时空过程分析、金融计量经济学。他在非线性和高维时间序列方面的研究国际领先。姚琦伟教授迄今已发表学术论文80多篇， 并获得EPSRC, BBSRC等英国国家基金会支持的多项研究基金项目。其专著《非线性时间序列：非参数及参数方法》（与范剑青合著）于2003年由Springer 出版，《计量金融简要》（与范剑青合著）于2017年由剑桥出版社出版。姚琦伟教授已担任包括Annals of Statistics，Journal of the American Statistics Association, Journal of the RoyalStatistical Society (Series B) 等多个顶级杂志副主编，并曾任 Statistica Sinica的联合主编。姚琦伟教授还曾为巴克莱银行，法国电力公司以及Winton资本等多家企业提供咨询。 讲座摘要： Probabilisticforecasting of electricity load curves is of fundamental importance foreffective scheduling and decision making in the increasingly volatile andcompetitive energy markets. We propose a novel approach to constructprobabilistic predictors for curves (PPC), which leads to a natural and newdefinition of quantiles in the context of curve-to-curve linear regression.There are three types of PPC: a predict set, a predictive band and a predictivequantile, and all of them are defined at a pre-specified nominal probabilitylevel. In the simulation study, the PPC achieve promising coverageprobabilities under a variety of data generating mechanisms. Whenapplying to one day ahead forecasting for the French daily electricity loadcurves, PPC outperform several state-of-the-art predictive methods in terms offorecasting accuracy, coverage rate and average length of the predictivebands. For example, PPC achieve up to 2.8-fold of the coverage rate withmuch smaller average length of the predictive bands. The predictive quantilecurves provide insightful information which is highly relevant to hedging risksin electricity supply management. (Jointwork with Xiuqin Xu, Ying Chen and Yannig Goude)
