Wind Power Forecasting | Marcus Kalander

Challenge Overview

The KDD Cup 2022 competition aimed to predict the wind power output of a wind farm across various time horizons. Accurate wind power forecasting is crucial for integrating wind energy into grid systems, given the inherent variability of wind. Wind power forecasting has been widely recognized as one of the most critical issues in wind power integration and operation.

Key details of the competition include:

The wind farm comprises 134 wind turbines.
Each turbine provides a time series with 10-minute intervals and 10 features.
The task requires predicting wind power from 0 to 48 hours into the future (288 steps) at each timestep.
The spatial locations of the wind turbines are provided.
Approximately 2,500 teams participated in the competition.

Immediate forecast.

One-day future forecast.

Two-day future forecast.

Feature importance across different forecast horizons visualized using SHAP values.

Proposed Solution

A prediction for a single wind turbine at time t.

Our proposed solution, detailed in (Kalander et al., 2022), combines two innovative models. A brief overview is also available in a YouTube presentation. The approach integrates:

Modified DLinear (MDLinear): An enhanced adaptation of the DLinear model (Zeng et al., 2023).
Extreme Temporal Gated Network (XTGN): A novel architecture built by stacking gated temporal convolutional networks (TCNs) (Dauphin et al., 2017; Lea et al., 2016) and incorporating nearest-neighbor information diffusion.

Both models employ a masked loss function to handle missing, unknown, or anomalous values effectively. Our solution achieved a remarkable 6th place out of approximately 2,500 competing teams.

References

2023

Are Transformers Effective for Time Series Forecasting?

Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu

In , 2023

2022

KDD Cup

Wind Power Forecasting with Deep Learning: Team didadida_hualahuala

Marcus Kalander, Zhongwen Rao, and Chengzhi Zhang

KDD 2022, 2022

Abs HTML Video Code

Wind energy is an effective supplement to traditional energy sources. However, wind power generation is strongly weather-dependent and thus not only unsustainable in terms of production, but also highly volatile. This complex variability poses a huge challenge to the smooth operation of the grid system. In this scenario, a unique Spatial Dynamic Wind Power Forecasting dataset from Longyuan Power Group Corp. Ltd (SDWPF) is used for modern wind turbine power forecasting. In this work, we propose a framework for accurate wind power generation forecasting (WPF) based on deep learning. To obtain a more generalizable learner, we use ensemble learning to build our framework, which consists of the following two parts: a) a modified version of the DLinear model, which uses time series decomposition and linear layers for information aggregation, and b) an extreme time-gated network that adaptively captures finegrained information and re-aggregates information by spatial location in the inference stage to obtain higher forecasting accuracy. The results indicate that our proposed combined model framework can capture long-term time series information well. Our code is available at https://github. com/shaido987/KDD_wind_power_forecast.

2017

Language Modeling with Gated Convolutional Networks

Yann N. Dauphin, Angela Fan, Michael Auli, and David Grangier

2017

2016

Temporal Convolutional Networks for Action Segmentation and Detection

Colin Lea, Michael D. Flynn, Rene Vidal, Austin Reiter, and Gregory D. Hager

2016