A Pattern‐Based Machine Learning Model for Imputing Missing Records in Coastal Wind Observation Networks
ABSTRACT Promoting green energy is essential for environmental sustainability, with wind energy playing a crucial role in this effort. While the Taiwan Strait has long been developed as a prime wind farm location, the search for new sites has led the government to focus on northern Taiwan, where the...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-05-01
|
| Series: | Meteorological Applications |
| Subjects: | |
| Online Access: | https://doi.org/10.1002/met.70050 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | ABSTRACT Promoting green energy is essential for environmental sustainability, with wind energy playing a crucial role in this effort. While the Taiwan Strait has long been developed as a prime wind farm location, the search for new sites has led the government to focus on northern Taiwan, where the Northeast Monsoon prevails during winter. Since 2022, new meteorological stations have been established to monitor wind potential in this region. However, missing wind data from these stations can undermine the accuracy of wind assessments. To address this, we develop an imputation model using the Weighted K‐Nearest Neighbors (WKNN) algorithm. This study focuses on seven meteorological stations near National Taiwan Ocean University (NTOU), located along the northeastern coast of Taiwan, including six on Taiwan proper and one on a nearby offshore islet, each recording wind speed and direction hourly. Complete data points, where all stations have recorded data simultaneously, are compiled into a reference database. When data from a particular station is missing, several complete data points from the database are used to estimate the missing values through weighted averaging. Calibration, validation, and testing procedures confirm that the model reliably estimates missing data, even when only four of the seven stations are operational. |
|---|---|
| ISSN: | 1350-4827 1469-8080 |