An Optimized ANN Measure-Correlate-Predict Method for Long-term Wind Prediction in Malaysia
The major issues on the wind measurement campaign are the data measured in a short period and the occurrence of missing data due to the failure of the measurement instrument. Meanwhile, Measure-Correlate-Predict (MCP) method had widely been used to predict the long-term condition and missing data at...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Conference Paper |
| Language: | English |
| Published: |
2020
|
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The major issues on the wind measurement campaign are the data measured in a short period and the occurrence of missing data due to the failure of the measurement instrument. Meanwhile, Measure-Correlate-Predict (MCP) method had widely been used to predict the long-term condition and missing data at the measurement site based on nearest Malaysian Meteorological Department (MMD), Meteorological Aerodrome Report (METAR) and extended Climate Forecast System Reanalysis (ECFSR) data. In this research, the long-term wind data at selected potential sites in Malaysia were predicted by optimized Artificial Neural Networks (ANNs). The Genetic Algorithm (GA) was applied to optimize the ANN. Five different ANN MCP models had been designed based on different types of reference data and different temporal scales to predict wind data at three target sites. Weibull frequency distributions and RMSE examined predicted wind data. The prediction of ANN had been improved in between 20.562% to 113.573% by GA optimization. The best R-value obtained from optimization were affected the Weibull shape and scale of predicted data. At last, the result revealed that the optimized ANN model could predict the long-term data for the target site with better accuracy. © 2018 Asian Institute of Technology. |
|---|