Sustainable operations of a combined cycle power plant using artificial intelligence based power prediction

Combined Cycle Power Plants (CCPP) are an effective method for Power generation due to their high thermal efficiency, low fuel consumption, and low greenhouse emissions. However, investing millions into building a power plant without knowledge of the power generation capacity seems unproductive. With the help of AI, we have tried to eliminate this conundrum. The present study focuses on the prediction of power produced by a 747 MW Combined Cycle Power Plant (CCPP) using a Back Propagation Neural Network (BPNN) and compares its results with the actual data from CCPP. BPNN is a regression-based prediction technique that is utilized in this study to develop a predictive model and train it using the following input features: Ambient Temperature, Ambient Pressure, Mass Flow rate of fuel in Gas Turbine 1, and Mass Flow rate of fuel in Gas Turbine 2. The Predictive Model with 10 neurons in the hidden layer was found to be most effective with Mean Squared Error (MSE) value, for the validation dataset, of 0.0063237. CCPP is also analyzed through a thermodynamic model, developed using EES. A detailed energy analysis is carried out and the results were compared with predicted and actual data. It was found that the thermal efficiency and total power generation of actual, predicted, and simulated models were 27.541% & 667.32 MW, 28.238% & 683.48 MW and 28.201% & 683.16 MW, respectively. A parametric study was further carried out to investigate the significance of operating parameters on power output and it was concluded that the temperatures across the Gas turbines have a significant impact on the performance of CCPP. Finally, Methane was replaced by 3 different fuels, one by one, and the effect of each fuel was investigated thermodynamically. It was found that the Lower Heating Value (LHV) of fuel was an important parameter in achieving a higher power output. It can be summarized from this research work that predictive models do have accuracy and such data science techniques can be used as a substitute for extensive thermodynamic calculations.