A Comparative Study on Various Measure-Correlate-Predict Techniques in Jeju Island
- Abstract
- For wind farm development the wind condition for the measurement period of one year is often considered. However in view of the turbine life cycle it is more important to consider the long-term wind conditions rather than those of one year or less. The Measure-Correlate-Predict (MCP) technique is a statistical method to predict the long-term wind resource at a target site, where the onsite short-term data are available, with a reference long-term data usually measured at meteorological observatories. Many MCP methods have been proposed and applied in different cases in accordance with the terrain types, data measurement period, etc, but there are still no fixed rules to apply these methods.
This study has been conducted to find a better MCP method that is in particular suitable for Jeju island which has a different terrain shapes strongly characterized by the Halla mountain centered on the island.
The three different methods such as linear regression, matrix, joint probabilistic methods were selected to compare their performance in different terrain types. The Hangwon site nearby coastal area was chosen as simple terrain type, and the Hoichun and Susan sites were selected as the mountainous-complex terrain types.
These methods have been applied in two different ways. First the linear regression method was applied to check the usefulness of it. Secondly, on the basis of the first test results, two other methods including the linear regression method were applied to each terrain type. The predicted results were compared with measured wind data including weibull parameters. The results showed the advantages of each MCP model for prediction of long-term wind conditions at target sites. It also demonstrated the limitation of these models for the different terrain types. It was shown that matrix and joint probabilistic methods were more useful in mountainous terrain than linear regression method because these methods are considering wind direction.
- Issued Date
- 2010
- Awarded Date
- 2010. 2
- Type
- Dissertation
- Alternative Author(s)
- Kwak, Gyeong Il
- Affiliation
- 제주대학교 대학원
- Department
- 대학원 기계공학과
- Advisor
- 허종철
- Table Of Contents
- 1. Introduction 1
1.1 Background 1
1.2 Objectives 5
2. A Review of MCP Techniques 6
2.1 Linear regression method 6
2.2 Matrix method 7
2.3 Joint probabilistic method 8
3. Site Description and Data sets 10
3.1 Site Description 10
3.1.1 Overview of Jeju island 10
3.1.2 Characteristics of each site 11
3.1.2.1 Hangwon site 11
3.1.2.2 Hoichun site 12
3.1.2.3 Susan site 14
3.2 Data sets 15
3.2.1 Mast data 15
3.2.1.1 Hangwon Mast 16
3.2.1.2 Hoichun Mast 17
3.2.1.3 Susan Mast 18
3.2.2 Reference data 19
4. Application of each method 22
4.1 Evaluation of usefulness of linear regression method 22
4.1.1 Data set 22
4.1.2 Correlation and long-term prediction 24
4.1.3 Predictive accuracy on both sites 25
4.1.3.1 Hangwon site 25
4.1.3.2 Hoichun site 27
4.2 Comparison of different MCP methods 30
4.2.1 Wind data sets 31
4.2.2 Correlation coefficient 32
4.2.3 Evaluation 33
4.2.3.1 Wind Veering of each site 33
4.2.3.2 Wind speed and direction of each site 36
4.2.3.3 Weibull parameters 40
5. Conclusions and future work 43
5.1 Conclusions 43
5.2 Future work 44
References 45
Appendices 49
Appendix A Scatter plot and Joint probability distribution of Hangwon 50
Appendix B Scatter plot and Joint probability distribution of Hoichun 51
Appendix C Scatter plot and Joint probability distribution of Susan 52
Appendix D Wind directional frequency of each site by three different methods 53
Appendix E Directional Weibull parameters of each site 54
- Degree
- Master
- Publisher
- 제주대학교 대학원
- Citation
- 곽경일. (2010). A Comparative Study on Various Measure-Correlate-Predict Techniques in Jeju Island
- Appears in Collections:
- Faculty of Applied Energy System > Mechanical Enginering
- 파일 목록
-
-
Download
A Comparative Study on Various Measure-Correlate-Predict Techniques in Jeju Island.pdf
기타 데이터 / 8.91 MB / Adobe PDF
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.