PENGEMBANGAN DIGITAL TWIN PADA PEMBANGKIT LISTRIK TENAGA SURYA (PLTS) UNTUK MIKRO GRID CERDAS DENGAN METODE INFORMED MACHINE LEARNING
Rapid population growth is driving global demand for energy, driving increased clean energy generation to reduce dependence on fossil fuels and mitigate the impacts of climate change. In this transition, Micro Grid technology, with a combination of energy sources, including PLTS, is crucial for incr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/79325 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Rapid population growth is driving global demand for energy, driving increased clean energy generation to reduce dependence on fossil fuels and mitigate the impacts of climate change. In this transition, Micro Grid technology, with a combination of energy sources, including PLTS, is crucial for increasing network reliability. To support the design of PLTS systems and influence costs, a reliable energy production prediction model is needed. Transmission operators and energy suppliers also rely on this model to maintain grid balance and participate in energy markets.
However, this is also a problem in itself, where it is difficult to model PLTS production results accurately due to its very weather-dependent nature. For this reason, in this research an Informed Machine Learning model will be developed where there is a physical constraint in the form of Angle of Incidence to represent the position of the sun towards the PLTS into the DNN model followed by the addition of an Approximation Constraint to limit the modeling results to the desired range.
Through the development of Informed Machine Learning modeling, it provides better accuracy and model compliance with the physical rules that apply in PLTS power modeling. Based on model evaluation, the performance of the Informed Machine Learning development model was obtained with the performance metrics Root Mean Square Error (RMSE) 71.00 W and Mean Absolute Error (MAE) 49.11 W, without any violations of the rules of physics. This is different from before the addition of physical constraints and Approximation Constraints where the PLTS model had a performance metric of Mean Square Error (RMSE) of 114.83 W and Mean Absolute Error (MAE) of 91.90 followed by 811 violations of the rules of physics.
|
|---|