Generator capacity predictor models using logistic regression and artificial neural network at Pt Saka Indonesia Pangkah limited

The gas turbine generator at PT Saka Indonesia Pangkah Limited is an essential part of providing energy used in various processes. The reliability and efficiency of gas turbine generators are essential, so it is necessary to design a prediction system through a multi-model approach. This study uses logistic regression and artificial neural network methods to build a predictor model. The model is designed by eight data inputs, namely air supply pressure, enclosure temperature, pressure compressor discharge, actual fuel flow, generator total, turbine air inlet temperature, gas fuel temperature, turbine air inlet differential pressure, and output data is generator capacity. The correlation value for each data was calculated by the Pearson method. Data pairs were trained and tested by logistic regression and artificial neural network methods with training, testing, and evaluation data ratio of 70:20:10. The test results show the highest correlation value of 0.99 at turbine air inlet temperature to generator capacity. The results of the model performance show that the logistic regression method is better than the ANN method with MAE value of 6.299, MSE of 68.240, R2 of 0.900, and EVS of 0.902.