Village power consumption and green energy prediction with deep learning | |
| Author | Sitakarn Na Nagara |
| Call Number | AIT RSPR no.DSAI-22-02 |
| Subject(s) | Electric power distribution--Thailand--Mathematical models Deep learning (Machine learning) |
| Note | A research study submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Data Science and Artificial Intelligence, School of Environment, Resources and Development |
| Publisher | Asian Institute of Technology |
| Series Statement | Research studies project report ; no. DSAI-22-02 |
| Abstract | Nowadays, electricity consumption forecasting has become an integral part of the power generation system and power grid planning. Mae Sariang is a district in the north located in a high valley where there is not enough electricity to meet the demand. This leads to problems such as power outages, power failures, etc. In this research, we aimed to develop an electricity consumption prediction model using deep learning. For the dataset, weather datasets are retrieved from Weather Research and Forecasting (WRF) dataset. This dataset is trained with the machine learning model and are compared to evaluate the quality of the dataset. For electricity usage data, the load profile data is supported by the Provincial Electricity Authority (PEA). In addition, in terms of the prediction model, we aimed to use XGBoost, Artificial Neural Network (ANN), Long-Short Term Memory (LSTM), Convolution Neural Network – 1 Dimension (CNN-1D) and Convolution Neural Network – Long-Short term Memory (CNN-LSTM). After analyzed the data and conducted the experiment by training the model, we found that the feature that impacts all feeders is summer, as previously research. Furthermore, hours impact energy consumption since we may divide daily hours into working hours and off-hours, which both have different electricity usage. The following revelation was that each feeder contained unique information. They are unrelated since each feeder transmits power to a different location. Therefore, the variables that affect the electricity consumption of each feeder are different but not very different. Lastly, the important finding for machine learning model training is that XGBoost model gave an impressive evaluation score, around 96 - 98% of R-square for test set. It indicates that this model can forecast the power usage efficiently. For deep learning model, CNN-LSTM gave the highest evaluation score, around 90 - 94 % of R-square for test set. Interestingly, another result we found was that the machine learning model was very efficient compared to the deep learning model. Under the same conditions, the machine learning model's R-square (XGBoost) could reach 98%, while the deep learning model (CNN-LSTM) had about 94 % of the R-square. |
| Year | 2022 |
| Corresponding Series Added Entry | Asian Institute of Technology. Research studies project report ; no. DSAI-22-02 |
| Type | Research Study Project Report (RSPR) |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Data Science and Artificial Intelligence (DSAI) |
| Chairperson(s) | Mongkol Ekpanyapong; |
| Examination Committee(s) | Dailey, Matthew N.;Singh, Jai Govind; |
| Scholarship Donor(s) | Royal Thai Government Fellowship; |
| Degree | Research Studies Project Report (M. Eng.) - Asian Institute of Technology, 2022 |