Enhancing Predictive Modeling in Power Plants: A Comparative Analysis of ANFIS and ANN for Electrical Energy Output Estimation

The main goal of this research work is to evaluate two widely known machine learning models as the potential predictors of electrical power output in a gas turbine and steam turbine combined cycle power plant. These four main input parameters-adjusted ambient temperature ( AT ), adjusted ambient pressure ( AP ), adjusted relative humidity (RH) and adjusted exhaust vacuum ( V )--are the basis for training and testing the models given in this study. Two methods, Grid-Partitioning and Subtractive Clustering, were used to develop the ANFIS model. The structure of the ANN is a two-layer feedforward network with sigmoid neurons in the hidden layer and linear neurons in the output layer. It is interesting to note that the ANN model was trained using the Levenberg-Marquardt optimization algorithm. The performance of ANFIS and the ANN models are measured using both Mean Squared Error (MSE) and Correlation (R) metrics. Both of them are preferable to each other in predicting the performance of power plants because the results of these tests fully support the conclusion. ANFIS model uses the fast and reliable Levenberg-Marquardt training method. A strong R value of 0.9711, for example, and a minimum MSE value of 16.5887 testify to this advantage. On topics concerning energy plants, whole industries seeking final solutions have a responsibility to study the findings of the report with due care. CCGP operators can make the energy plant function more efficiently in the future by making the power management models more reliable and accurate. A study which constantly tests combined cycle power plant performance in terms of sustainability, operations for an entire year.