Advanced machine learning approach for restoring smart grids
In this project, we explore the option of using Gaussian Process Regression in voltage control of power systems. The objective is to use machine learning data driven methods in favour of conventional methods which are iterative mathematical methods. The main advantage of data driven method is less c...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/158203 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | In this project, we explore the option of using Gaussian Process Regression in voltage control of power systems. The objective is to use machine learning data driven methods in favour of conventional methods which are iterative mathematical methods. The main advantage of data driven method is less computationally expensive and faster.
As there are many machine learning methods and even within Gaussian Process Regression, there are many ways and methods to achieve the purpose. This project attempts to answer which kernel would be most suitable to solve power flow equations.
The approach taken was to train the models with varying sample sizes and different kernels on IEEE 33 bus system. The error will then be used to compare against the methods and determine the most suitable kernel for solving power flow.
The results shown in this report indicated that the Exponential and Matern kernels have performed well due to their low Mean Absolute Error results in the tests performed. The default SquaredExponential kernel only performed well when trained with the least data sets in this project, 50. |
|---|