Effective monitoring of Pelton turbine based hydropower plants using data-driven approach
(2023) In International Journal of Electrical Power and Energy Systems 149.- Abstract
Hydropower is a renewable and reliable energy source that can be utilized to fulfil energy demands and promote energy sustainability. In hydropower plants, Pelton turbines are installed at high-head and low-discharge sites. In Pelton turbines, needle, nozzle, buckets, and splitter are most exposed to silt erosion. An eroded turbine takes more discharge than rated for generating the same amount of power. Therefore, to minimize the losses due to silt erosion, real-time condition-based monitoring of a hydropower plant is necessary. In this paper, correlations are developed using the historical data collected from the Urgam hydropower plant to predict the generated power. The artificial Neural Network (ANN) technique has been used to... (More)
Hydropower is a renewable and reliable energy source that can be utilized to fulfil energy demands and promote energy sustainability. In hydropower plants, Pelton turbines are installed at high-head and low-discharge sites. In Pelton turbines, needle, nozzle, buckets, and splitter are most exposed to silt erosion. An eroded turbine takes more discharge than rated for generating the same amount of power. Therefore, to minimize the losses due to silt erosion, real-time condition-based monitoring of a hydropower plant is necessary. In this paper, correlations are developed using the historical data collected from the Urgam hydropower plant to predict the generated power. The artificial Neural Network (ANN) technique has been used to develop correlation, and the performance of this model has been compared with the model developed using the Curve Fitting technique. Based on the comparison, it has been found that the ANN model performs better than the curve-fitting model in forecasting the generated power with an R2-value of 0.9522 and Mean Absolute Percentage Error (MAPE) of 2.40% at 3.1515% Root Mean Square Percentage Error (RMSPE). The developed model can be utilized to plan a maintenance schedule of machines based on real-time conditions. It can also help to decide which machine should be taken in maintenance first.
(Less)
- author
- Kumar, Krishna LU ; Saini, Gaurav ; Kumar, Aman ; Elavarasan, Rajvikram Madurai ; Said, Zafar and Terzija, Vladimir
- publishing date
- 2023-07
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Energy, Machine Learning, Maintenance, Monitoring, O&M, Pelton turbine
- in
- International Journal of Electrical Power and Energy Systems
- volume
- 149
- article number
- 109047
- publisher
- Elsevier
- external identifiers
-
- scopus:85149178002
- ISSN
- 0142-0615
- DOI
- 10.1016/j.ijepes.2023.109047
- language
- English
- LU publication?
- no
- additional info
- Publisher Copyright: © 2023 Elsevier Ltd
- id
- a2a54046-325f-49b0-9ff5-0eb5f9f92b7c
- date added to LUP
- 2024-04-24 11:49:18
- date last changed
- 2024-05-16 14:46:50
@article{a2a54046-325f-49b0-9ff5-0eb5f9f92b7c, abstract = {{<p>Hydropower is a renewable and reliable energy source that can be utilized to fulfil energy demands and promote energy sustainability. In hydropower plants, Pelton turbines are installed at high-head and low-discharge sites. In Pelton turbines, needle, nozzle, buckets, and splitter are most exposed to silt erosion. An eroded turbine takes more discharge than rated for generating the same amount of power. Therefore, to minimize the losses due to silt erosion, real-time condition-based monitoring of a hydropower plant is necessary. In this paper, correlations are developed using the historical data collected from the Urgam hydropower plant to predict the generated power. The artificial Neural Network (ANN) technique has been used to develop correlation, and the performance of this model has been compared with the model developed using the Curve Fitting technique. Based on the comparison, it has been found that the ANN model performs better than the curve-fitting model in forecasting the generated power with an R<sup>2</sup>-value of 0.9522 and Mean Absolute Percentage Error (MAPE) of 2.40% at 3.1515% Root Mean Square Percentage Error (RMSPE). The developed model can be utilized to plan a maintenance schedule of machines based on real-time conditions. It can also help to decide which machine should be taken in maintenance first.</p>}}, author = {{Kumar, Krishna and Saini, Gaurav and Kumar, Aman and Elavarasan, Rajvikram Madurai and Said, Zafar and Terzija, Vladimir}}, issn = {{0142-0615}}, keywords = {{Energy; Machine Learning; Maintenance; Monitoring; O&M; Pelton turbine}}, language = {{eng}}, publisher = {{Elsevier}}, series = {{International Journal of Electrical Power and Energy Systems}}, title = {{Effective monitoring of Pelton turbine based hydropower plants using data-driven approach}}, url = {{http://dx.doi.org/10.1016/j.ijepes.2023.109047}}, doi = {{10.1016/j.ijepes.2023.109047}}, volume = {{149}}, year = {{2023}}, }