算法比赛 | KDD Cup2022 风力发电预测比赛总结

郭好奇同学

发布于 2022-11-15 09:19:51

3.2K0

发布于 2022-11-15 09:19:51

文章被收录于专栏：好奇心Log

赛题介绍

空间动态风力发电预测（Spatial Dynamic Wind Power Forecasting）对风能的利用具有实际意义，参与者应准确估计风电场的风能供应。

自 1997 年以来，KDD 杯一直是与 ACM SIGKDD 知识发现和数据挖掘会议一起举办的首屈一指的年度数据挖掘竞赛。今年的KDD杯挑战任务提出了有趣的技术挑战，对风能的利用具有现实意义。在这里，我们提出了一个空间动态风能预测挑战，以促进数据驱动的风能预测机器学习方法的进步。

赛题背景

风电预测（Wind Power Forecasting, WPF）旨在准确估计风电场在不同时间尺度上的风能供应。风电是一种清洁安全的可再生能源，但不能持续生产，导致高波动性。

这种可变性可能对将风力发电并入电网系统提出重大挑战。为了保持发电和消费之间的平衡，风电的波动需要从其他可能无法在短时间内获得的电力替代（例如，通常至少需要 6 个小时才能点燃一个燃煤电厂）。

因此WPF被广泛认为是风电并网运行中最关键的问题之一。数据挖掘和机器学习社区中出现了关于风力发电预测问题的研究爆炸式增长。然而，如何处理好 WPF 问题仍然具有挑战性，因为始终需要高预测精度来确保电网稳定性和供电安全。

赛题任务

我们提供了来自 Longyuan Power Group Corp. Ltd: SDWPF的独特的空间动态风力发电预测数据集，其中包括风力涡轮机的空间分布，以及时间、天气和涡轮机内部状态等动态背景因素。然而大多数现有的数据集和竞赛将 WPF 视为时间序列预测问题，而不知道风力涡轮机的位置和上下文信息。

数据集来自风电场风力涡轮机监控和数据采集系统的上下文监测数据。SCADA 数据每 10 分钟从龙源电力集团拥有的风电场（由 134 台风力涡轮机组成）中的每个风力涡轮机进行采样。此外，将发布风电场中所有风力涡轮机的相对位置，以表征风力涡轮机的空间分布。

优胜选手分享

比赛官网：https://aistudio.baidu.com/aistudio/competition/detail/152/0/introduction

答辩视频：https://www.bilibili.com/video/BV1Y14y1x7XN/

Regular Track

Complementary Fusion of Deep Spatio-Temporal Network and Tree Model for Wind Power Forecasting

Team member: Linsen Li, Qichen Sun, Dongdong Geng, Chunfei Jian, Dongen Wu; Advisor: Shiliang Pu
Team name: HIK; Leaderboard score: -44.91708;

paper: https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_0518.pdf
slides: https://baidukddcup2022.github.io/slides/hik.pdf

Solution to Spatial Dynamic Wind Power Forecasting for KDD Cup 2022

Team member: Hanhan Liu
Team name: trymore; Leaderboard score: -44.9234
  
paper: https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_1286.pdf
code: https://github.com/hanhanliu/wpf22
slides: https://baidukddcup2022.github.io/slides/trymore.pdf

Application of BERT in Wind Power Forecasting-Teletraan’s Solution in Baidu KDD Cup 2022

Team member: Longxing Tan, Hongying Yue
Team name: Teletraan; Leaderboard score: -45.09478;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_2696.pdf
code : https://github.com/LongxingTan/KDDCup2022-Baidu
slides : https://baidukddcup2022.github.io/slides/Teletraan.pdf

Team zhangshijin WPFormer: A Spatio-Temporal Graph Transformer with Auto-Correlation for Wind Power Prediction

Team member: Xuefeng Liang, Qingshui Gu, Su Qiao, Zhuwang Lv, Xin Song
Team name: zhanshijin; Leaderboard score: -45.13867;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_3454.pdf
code : https://github.com/David1X/Data_Competitions/tree/release-0.1/KDD2022_WPF
slides : https://baidukddcup2022.github.io/slides/zhanshijin.pdf

EasyST: Modeling Spatial-Temporal Correlations and Uncertainty for Dynamic Wind Power Forecasting via PaddlePaddle

Team member: Yiji Zhao, Haomin Wen, Junhong Lou, Jinji Fu, Jianbin Zheng; Advisor: Youfang Lin
Team name: EasyST; Leaderboard score: -45.17326;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_4203.pdf
code : https://aistudio.baidu.com/aistudio/projectdetail/4378775
slides : https://baidukddcup2022.github.io/slides/EasyST.pdf

Wind Power Forecasting with Deep Learning: Team didadida_hualahuala

Team member: Marcus Kalander, Zhongwen Rao, Chengzhi Zhang
Team name:didadida_hualahuala; Leaderboard score: -45.18139;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_5582.pdf
code : https://github.com/shaido987/KDD_wind_power_forecast
slides : https://baidukddcup2022.github.io/slides/didadida_hualahuala.pdf

KDD CUP 2022 Wind Power Forecasting Team 88VIP Solution

Team member: Fangquan Lin, Wei Jiang, Hanwei Zhang; Advisor: Cheng Yang
Team name: 88VIP; Leaderboard score: -45.21301;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_1833.pdf
code : https://github.com/linfangquan/kddcup2022

dataZhi: A multi-scale fusion method for wind power forecasting with spatiotemporal attention networks

Team member: Hongzhi Luan; Advisor: Junxiong Hou
Team name: dataZhi; Leaderboard score: -45.23701;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_9920.pdf
code : https://github.com/luanhzh/wpf_fusion

AIStudio2338769Team: Long-Short Term Forecasting for Active Power of a Wind Farm

Team member: Wenwei Wang
Team name: AIStudio2338769Team; Leaderboard score: -45.27256;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_8243.pdf
code : https://github.com/wenwei-pku/kddcup2022

Multi-Stage Robust Wind Power Forecasting

Team member: Chenxu Wang, Jinda Lu, Yuan Gao
Team name: SlienceGTeam; Leaderboard score: -45.32777;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_0931.pdf
code : https://github.com/injadlu/KDDCUP2022

BUAA_BIGSCity: Spatial-Temporal Graph Neural Network for Wind Power Forecasting in Baidu KDD CUP 2022

Team member: Jiawei Jiang, Chengkai Han; Advisor: Jingyuan Wang
Team name: BUAA_BIGSCity; Leaderboard score: -45.36026;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_9863.pdf
code : https://github.com/BUAABIGSCity/KDDCUP2022

Hybrid Model: Deep learning GRU neural network and K-nearest neighbors for Wind Power Forecasting

Team member: Fernando Sebastián Huerta, Manuel Ángel Suárez Álvarez, Daniel Velez Serrano, Eugenio Neira Bustamante, Alejandro Carrasco Sanchez
Team name: datateam-UCM; Leaderboard score: -45.56335;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_0115.pdf
code : https://github.com/ManuelAngel99/KDD_CUP_2022
slides : https://baidukddcup2022.github.io/slides/datateam_UCM.pdf

PaddlePaddle Track

Solution to Spatial Dynamic Wind Power Forecasting for KDD Cup 2022

Team member: Hanhan Liu
Team name: trymore; Leaderboard score: -44.9234
  
paper: https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_1286.pdf
code: https://github.com/hanhanliu/wpf22
slides: https://baidukddcup2022.github.io/slides/trymore.pdf

Team zhangshijin WPFormer: A Spatio-Temporal Graph Transformer with Auto-Correlation for Wind Power Prediction

Team member: Xuefeng Liang, Qingshui Gu, Su Qiao, Zhuwang Lv, Xin Song
Team name: zhanshijin; Leaderboard score: -45.13867;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_3454.pdf
code : https://github.com/David1X/Data_Competitions/tree/release-0.1/KDD2022_WPF
slides : https://baidukddcup2022.github.io/slides/zhanshijin.pdf

EasyST: Modeling Spatial-Temporal Correlations and Uncertainty for Dynamic Wind Power Forecasting via PaddlePaddle

Team member: Yiji Zhao, Haomin Wen, Junhong Lou, Jinji Fu, Jianbin Zheng; Advisor: Youfang Lin
Team name: EasyST; Leaderboard score: -45.17326; 

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_4203.pdf
code : https://aistudio.baidu.com/aistudio/projectdetail/4378775
slides : https://baidukddcup2022.github.io/slides/EasyST.pdf

Spatial wind power forecasting using a GRU-based model: WindTeam CSU123

Team member: Zhi Liu, Min Li, He Wei, Baichuan Yang, Advisor: Min Li
Team name: WindTeam CSU123; Leaderboard score: -45.6186;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_0328.pdf
code : https://github.com/LiuZhihhxx/KDD2022

Spatial Dynamic Wind power forecasting using lightGBM and multi-variate LSTM with hierarchical coherence constraints(Team name: Dynamo)

Team member: Hongfeng Ai, Wenqi Wu and Chaodong Zhang
Team name: Dynamo; Leaderboard score: -45.64764;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_6393.pdf
code : https://github.com/cdzhang/wind_power_forecast
slides : https://baidukddcup2022.github.io/slides/dynamo.pdf

DLinear Makes Efficient Long-term Predictions

Team member: Chaoqun Su
Team name: yura; Leaderboard score: -46.16117;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_8925.pdf
code : https://github.com/ChaoqunSu/kddcup

A combination of Spatial-Temporal Graph Transformer Model and LSTM Model (Team: noritoshiTeam)

Team member: TAMURA Noritoshi
Team name: noritoshiTeam; Leaderboard score: -46.17403;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_9718.pdf
code : https://github.com/noritoshitamura/noritoshi_kddcup2022

A spatial-temporal ensemble deep learning framework for wind power forecasting (Team QDU)

Team member: Zhiruo Li, Jieqi Xing, Shunyao Wu; Advisor: Shunyao Wu
Team name: QDU; Leaderboard score: -46.26986;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_0841.pdf
code : https://github.com/hansu1017/SDWPF-Baidu-KDD-Cup-2022

Two Strategies to Reduce the Negative Effects of Abnormal and Missing Values for Wind Power Forecasting

Team member: Ruizhi Zhang, Zeyu Long, Yusu Mao; Advisor: Nengjun Zhu
Team name: 123123; Leaderboard score: -46.33641;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_0353.pdf
code : https://github.com/9898zy998/wind-power/tree/main

IFBD: Graph Convoluational networks with Transformer for Long Sequence Predictions

Team member: Xiaotian Yu
Team name: IFBD; Leaderboard score: -46.34794;

paper : https://baidukddcup2022.github.io/papers/Baidu_KDD_Cup_2022_Workshop_paper_5732.pdf

本文参与腾讯云自媒体同步曝光计划，分享自微信公众号。

原始发表：2022-09-15，如有侵权请联系 cloudcommunity@tencent.com 删除

数据结构

wpf

数据挖掘

深度学习

本文分享自好奇心Log 微信公众号，前往查看

如有侵权，请联系 cloudcommunity@tencent.com 删除。

本文参与腾讯云自媒体同步曝光计划，欢迎热爱写作的你一起参与！

登录后参与评论

0 条评论

热度