Forecasting the energy output from a combined cycle thermal power plant using deep learning models

The energy output from a combined cycle power plant (CCPP) varying with the operating thermal parameters like ambient pressure, vacuum, relative humidity, and relative temperature is modelling using different approaches. The huge data obtained from the experimental readings is found to be highly non-linear using the data visualization technique. The energy output from the CCPP reduces linearly with the temperature and non-linearly with pressure. A mathematical model is developed for the predictions of the energy output. Modelling using sequential API and functional API based artificial neural network (SANN and FANN) having single hidden layer is carried out. Finally, energy output modelling using sequential API and functional API based deep ANN (SDNN and FDNN) is also performed. The residuals of the predicted and experimental observations indicate that the error is acceptable and it lies uniformly above and below the regression line. The R-squared value of the mathematical model is 0.93 and 0.94 during training and testing. The obtained R-squared value of the ANN and DNN using sequential and functional API is 0.94. The training and testing of all the models are successful and these models have shown a great compatibility in predicting the energy output of a CCPP. The ANN model with single layer and deep layer has no difference in accuracy hence the former one is recommended as it is computationally less expensive.

Automated summary

The study focused on modeling the energy output of a CCPP, finding that it is linearly related to temperature and non-linearly related to pressure. The mathematical model showed high accuracy in predictions, with neural network models also demonstrating reliability in testing.