Data-driven small-signal stability assessment of power systems
The widespread usage of renewable energy sources such as solar and wind power shows that mankind is building a green and low-carbon energy consumption system. Due to the intermittent nature of renewable energy sources (RES), the stability of the system is facing more serious challenges as increasing...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/152497 |
| _version_ | 1811696648327266304 |
|---|---|
| author | Wang, Xurui |
| author2 | Xu Yan |
| author_facet | Xu Yan Wang, Xurui |
| author_sort | Wang, Xurui |
| collection | NTU |
| description | The widespread usage of renewable energy sources such as solar and wind power shows that mankind is building a green and low-carbon energy consumption system. Due to the intermittent nature of renewable energy sources (RES), the stability of the system is facing more serious challenges as increasing integration of RESs to the modern power systems. The assumption that the system's operational point state will not change over relatively long time period is required for the use of traditional stability analysis methodologies. However, because intermittent renewable energy sources result in an ever-changing power flow pattern, conventional approaches are ineffective for evaluating system stability.
In recent years, the growth and iteration of artificial intelligence have been extremely rapid, and machine learning approaches have become widely utilized in power system prediction and stability evaluation. The machine learning technique works on the premise of using previous data identified in the system to train the algorithm model offline and then using the learned model for online prediction. Its benefits include high nonlinearity, adaptivity (adaptive to both seen and unseen scenario), and fast speed. With the outstanding performance of machine learning algorithms in solving regression problems of nonlinear systems, how to choose a suitable machine learning algorithm to apply to the power system to predict the stability of the power system has become a hot research topic.
This article first introduces the related concepts and significance of power system stability, especially small signal stability, and then focuses on comparing the principles of several popular machine learning algorithms, namely Extreme Learning Machine (ELM), Random Vector Functional Link (RVFL), and Artificial Neural Network (ANN). In this paper, a novel small-signal stability assessment method based on an ensemble learning method is proposed. Ensemble learning allows the algorithm to characterize the data from several viewpoints, allowing it to attain a greater accuracy of fit. The proposed methods are tested on an IEEE 39-bus 10-machine system.
Experiments show that the adjustment of the respective parameters of each algorithm has a great impact on its final performance. Furthermore, the ensemble learning method has varying degrees of influence on the various algorithms, having the most substantial effect on the optimization of the ELM algorithm. After experiments, the ensemble learning can improve the regression effect from 0.28 to 0.53 with the other parameters of ELM algorithm unchanged. Eventually, when the parameters of the ELM, RVFL, and ANN algorithms are selected as their optimal prediction performance, the accuracy of the ANN algorithm is the best. |
| first_indexed | 2024-10-01T07:42:42Z |
| format | Thesis-Master by Coursework |
| id | ntu-10356/152497 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T07:42:42Z |
| publishDate | 2021 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1524972023-07-04T17:01:09Z Data-driven small-signal stability assessment of power systems Wang, Xurui Xu Yan School of Electrical and Electronic Engineering xuyan@ntu.edu.sg Engineering::Electrical and electronic engineering::Electric power Engineering::Electrical and electronic engineering::Computer hardware, software and systems The widespread usage of renewable energy sources such as solar and wind power shows that mankind is building a green and low-carbon energy consumption system. Due to the intermittent nature of renewable energy sources (RES), the stability of the system is facing more serious challenges as increasing integration of RESs to the modern power systems. The assumption that the system's operational point state will not change over relatively long time period is required for the use of traditional stability analysis methodologies. However, because intermittent renewable energy sources result in an ever-changing power flow pattern, conventional approaches are ineffective for evaluating system stability. In recent years, the growth and iteration of artificial intelligence have been extremely rapid, and machine learning approaches have become widely utilized in power system prediction and stability evaluation. The machine learning technique works on the premise of using previous data identified in the system to train the algorithm model offline and then using the learned model for online prediction. Its benefits include high nonlinearity, adaptivity (adaptive to both seen and unseen scenario), and fast speed. With the outstanding performance of machine learning algorithms in solving regression problems of nonlinear systems, how to choose a suitable machine learning algorithm to apply to the power system to predict the stability of the power system has become a hot research topic. This article first introduces the related concepts and significance of power system stability, especially small signal stability, and then focuses on comparing the principles of several popular machine learning algorithms, namely Extreme Learning Machine (ELM), Random Vector Functional Link (RVFL), and Artificial Neural Network (ANN). In this paper, a novel small-signal stability assessment method based on an ensemble learning method is proposed. Ensemble learning allows the algorithm to characterize the data from several viewpoints, allowing it to attain a greater accuracy of fit. The proposed methods are tested on an IEEE 39-bus 10-machine system. Experiments show that the adjustment of the respective parameters of each algorithm has a great impact on its final performance. Furthermore, the ensemble learning method has varying degrees of influence on the various algorithms, having the most substantial effect on the optimization of the ELM algorithm. After experiments, the ensemble learning can improve the regression effect from 0.28 to 0.53 with the other parameters of ELM algorithm unchanged. Eventually, when the parameters of the ELM, RVFL, and ANN algorithms are selected as their optimal prediction performance, the accuracy of the ANN algorithm is the best. Master of Science (Power Engineering) 2021-08-23T06:43:36Z 2021-08-23T06:43:36Z 2021 Thesis-Master by Coursework Wang, X. (2021). Data-driven small-signal stability assessment of power systems. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/152497 https://hdl.handle.net/10356/152497 en application/pdf Nanyang Technological University |
| spellingShingle | Engineering::Electrical and electronic engineering::Electric power Engineering::Electrical and electronic engineering::Computer hardware, software and systems Wang, Xurui Data-driven small-signal stability assessment of power systems |
| title | Data-driven small-signal stability assessment of power systems |
| title_full | Data-driven small-signal stability assessment of power systems |
| title_fullStr | Data-driven small-signal stability assessment of power systems |
| title_full_unstemmed | Data-driven small-signal stability assessment of power systems |
| title_short | Data-driven small-signal stability assessment of power systems |
| title_sort | data driven small signal stability assessment of power systems |
| topic | Engineering::Electrical and electronic engineering::Electric power Engineering::Electrical and electronic engineering::Computer hardware, software and systems |
| url | https://hdl.handle.net/10356/152497 |
| work_keys_str_mv | AT wangxurui datadrivensmallsignalstabilityassessmentofpowersystems |