Dissertations for Engineering Management (EM)

Permanent URI for this collection

https://bspace.buid.ac.ae/handle/1234/1075

Browse

Browsing Dissertations for Engineering Management (EM) by Subject "ambient weather parameters"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    The Influence of Ambient Weather Parameters on the Prediction of an Electrical Power Production of a Combined Cycle Power Plant in the UAE
    (The British University in Dubai (BUiD), 2022-10) ALABDOULI, HAJER ALI
    To improve the utilisation of power plants and enhance production, this study is devoted to predicting the baseload electrical power production of a combined cycle power plant in the UAE. The data for this study was taken from plant sensors over a period of one month (September 2021) from specific sensors installed in the power plant, and provided the data for input features that correspond to affect and change the electrical power production. In the UAE, the hot summer climate and ambient weather conditions adversely affect the performance of gas turbines (GT) and have an influence on steam turbines too. Accordingly, this paper studies four input variables: ambient temperature (ranges from 25.29°C to 36.5°C), relative humidity (ranges from 35.47% to 90.28%), atmospheric pressure (ranges from 0.99 bar to 1.01 bar) and exhaust steam vacuum (ranges from 0.057 bar to 0.126 bar). All influence the target variable (power production), which ranges from 506.32MW to 864.44MW. The change in the exhaust vacuum pressure in the steam turbine is affected by the change in ambient temperature, relative humidity, and atmospheric pressure in the gas turbine. The analysis includes applying machine learning methods such as linear regression and artificial neural networks (ANNs) to develop a predictive power production model using different interactive computer programs such as Minitab, RStudio and Microsoft Excel. The linear regression model R-sq value was found to be 53.49%. Consequently, Minitab software is found to be a slightly more accurate statistical package compared to RStudio. In addition, the best data subset is found to be for week 1, with R-sq value of 82.16%. Moreover, the power linear regression model is ascertained to be more accurate than the ANN power predictive model, with a mean absolute deviation of 46.385, symmetric mean absolute per cent error of 6.719 and residual standard error of 57.392 (Minitab outputs).