แจ้งเอกสารไม่ครบถ้วน, ไม่ตรงกับชื่อเรื่อง หรือมีข้อผิดพลาดเกี่ยวกับเอกสาร ติดต่อที่นี่ ==>
หากไม่มีอีเมลผู้รับให้กรอก thailis-noc@uni.net.th
หากไม่มีอีเมลผู้รับให้กรอก thailis-noc@uni.net.th
Thanyaporn Harncharnchai. Prediction daily electricity consumption in Thailand using multiple linear regression, artificial neural network, support vector machine, and hybrid models. Master's Degree(Logistics and Supply Chain Systems Engineering). Thammasat University. Thammasat University Library. : Thammasat University, 2021.
Prediction daily electricity consumption in Thailand using multiple linear regression, artificial neural network, support vector machine, and hybrid models
Name: Thanyaporn Harncharnchai
keyword:
Daily electricity consumption
Abstract:
This article involves forecasting daily electricity consumption in Thailand. Electricity consumption data are provided by the Electricity Generating Authority of Thailand, the leading power utility state enterprise under the Ministry of Energy. Five forecasting techniques, including multiple linear regression (MLR), artificial neural network (ANN), support vector machine (SVM), hybrid models, and ensemble models, are implemented. The article proposes a hyperparameter tuning technique, called sequential grid search, which is based on the widely used grid search, for ANN and hybrid models. Auxiliary variables and indicator variables that can improve the models forecasting performance are included. The performance measurement of this research is mean absolute percentage error (MAPE). The results of each model indicate as follows: For MLR: The prediction model is multiple linear regression. MAPEs of MLR without and with indicator variables are 2.02% and 1.95%, respectively. For EM1&EM2: These two models for prediction are based on ANN. After the hyperparameter tuning by sequential grid search is performed, the model size or node which is one of ANN architecture is considered. The considered model size in this research is one to fifteen sizes. For EM1, the best five settings which have the lowest MAPE is chosen among all model sizes. On the other hand, EM2 chooses the best five settings for each model size. After that, the forecast values of the best five settings are averaged in the ensemble model. Finally, the improving performance of the model is done by adding indicator variables. MAPEs of testing data set without and with indicator variables are 1.64% and 1.56% for EM1, and 1.74% and 1.71% for EM2. For EM3: SVM is the forecasting model. After the hyperparameter tuning by grid search is done, the best five settings which have the lowest MAPE of the validating data set are chosen. Then, the ensemble model is performed by averaging the forecast values of the best five settings. Eventually, the indicator variables are added. MAPEs of testing data set without and with indicator variables are 2.05% and 2.08%, respectively. For HM1, HM2, and HM3: The hybrid models for prediction is MLR combining with ANN or SVM. The residuals from the MLR model are used as the dependent variables which are predicted by ANN or SVM. The same best five settings from EM1, EM2, and EM3 models are used to predict the residuals denoted as HM1, HM2, and HM3, respectively. Next, the ensemble model is performed by averaging the forecast residuals of the best five settings. Then, the forecast values are computed through the hybrid model algorithm. Ultimately, indicator variables are added to improve the performance of the model. MAPEs of testing data set without and with indicator variables are 1.77% and 1.58% for HM1, 1.76% and 1.58% for HM2, and 1.83% and 1.85% for HM3. Based on the mean absolution percentage error, the best model, which has the test set MAPE of 1.56%, is the ensemble model of ANNs obtained from the proposed sequential grid search, which also includes additional indicator variables for some national holidays
Thammasat University. Thammasat University Library
Address:
BANGKOK
Email:
preserv@tu.ac.th
Role:
advisor
Created:
2021
Modified:
2023-04-19
Issued:
2023-04-19
วิทยานิพนธ์/Thesis
application/pdf
eng
DegreeName: Master of Engineering
Level: Master's Degree
Descipline: Logistics and Supply Chain Systems Engineering
Grantor: Thammasat University
©copyrights Thammasat University
RightsAccess:
ใช้เวลา
0.029483 วินาที
Thanyaporn Harncharnchai
| Title | Contributor | Type |
|---|---|---|
| Prediction daily electricity consumption in Thailand using multiple linear regression, artificial neural network, support vector machine, and hybrid models
มหาวิทยาลัยธรรมศาสตร์ Thanyaporn Harncharnchai | Jirachai Buddhakulsomsiri | วิทยานิพนธ์/Thesis |
Jirachai Buddhakulsomsiri