Loading...

Exploring the Possibility of Improving the Performance of a Steam Turbine Applying the Data Mining Approach on the Real Data Collected from the Related Power Cycle

Atarodi, Mahmood | 2017

369 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 50588 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Darbandi, Masoud
  7. Abstract:
  8. The steam turbines share a significant part of Iran's electricity production. Evidently improving the performance of these steam turbines will improve the performance of power plants and enhance the capability of power production in Iran. Since the behavior of steam turbines is highly nonlinear and that it is affected by various factors, it is difficult to determine the optimum operating conditions for each turbine in specific unit of a power plant. In addition, it is very difficult to model the turbine behavior based on the mathematical equations. Fortunately a large amount of information is collected and stored annually by the data-processing systems in each power plant. These collected data can well represent the actual behavior of a steam turbine. Using these collected data and the data mining method, one can model the real behavior of a specific turbine. In this study, we use a neural network approach to simulate behavior of a hybrid power cycle with 100 MW production and investigate the possibility of improving the performance of the steam cycle part. Different types of neural networks are used to perform the required modeling. We eventually arrive to a neural network with 2 hidden layers whose neurons numbers are 20. In this modeling, the steam cycle output power and its efficiency are calculated using the temperature, the flow, the pressure of the high pressure turbine, the flow of the intermediate pressure turbine and the turbine back pressure data. The average error for estimating the steam cycle power is about 0.1030 megawatts and it is about 0.043% for the efficiency modeling. After ensuring the real turbine behavior modeling performed, it is investigate the possibility of improving the performance of this cycle. Based on the calculations in the present study, it is anticipated that there will be a possibility to increase the steam power cycle efficiency as much as 2%
  9. Keywords:
  10. Data Mining ; Efficiency ; Neural Network ; Data Acquisition ; Power Output ; Turbine Performance ; Performance Improvement

 Digital Object List

 Bookmark

No TOC