Handling arch effects in wind speed data using state space approach model
In general, Malaysia experiences low wind speed, but some areas in this country experience strong wind in certain periods of time within a year. In line with the necessity to enhance the utilization of indigenous renewable energy resources in order to contribute towards national electricity supply,...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/77773/1/AaishahRadziahJamaludinMFS2017.pdf http://eprints.utm.my/id/eprint/77773/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:105132 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In general, Malaysia experiences low wind speed, but some areas in this country experience strong wind in certain periods of time within a year. In line with the necessity to enhance the utilization of indigenous renewable energy resources in order to contribute towards national electricity supply, the study on the potential of the wind speed as a new source of renewable energy is significantly crucial. For that reason, this study aims to model and forecast wind speed data in 10 stations all over Peninsular Malaysia by using three different methods; Autoregressive Moving Average (ARMA), hybrid model (ARMA with Generalized Autoregressive Conditional Heteroscedasticity (ARMA-GARCH)) and Dynamic Linear models (DLM). ARMA was used as the benchmark in identifying an adequate linear model. The Autoregressive Conditional Heteroscedasticity (ARCH) effect in the residuals data from the developed conventional model was determined. The presence of ARCH shows that the model is not appropriate to be treated as a linear model. Therefore, to overcome this problem, ARMA model was hybridized with GARCH model. However, there is still some remaining ARCH exists in the residuals data for several datasets. Thus, a new approach namely DLM was introduced in order to treat the shortcoming. At the end of the research, a comparative study was made. It was discovered that in most cases, DLM outperforms than other models. DLM is found to be flexible in treating the dynamical fluctuation of the data and superior in terms of predictive accuracy with just a small error when compared with other methods |
|---|