| Abstract |
This paper describes an application of a feedforward backpropagation artificial neural network (ANN), to predict Geothermal Power Plant (GPP) Unit 3 Ulubelu exergy efficiency. Exergy analysis was chosen since it considered as the most practical thermodynamics method for the system’s energy evaluation. From the exergy analysis, merging both exergy efficiency and exergy destruction highlights the energy inefficiencies within a system and provides useful information to the managers and decision makers for prioritizing the improvement potentials. However, inaccuracy of measuring equipment or instrument is inevitable especially when the equipment has been in operation since quite some time. Thus, the exergy analysis results can be compromised and inaccurate. The ANN method with backpropagation had been investigated to overcome this problem. The ANN model with 6-4-1 structure has been developed and investigated. The input layer for the ANN model was designed using 6 parameters such as steam inlet pressure, steam inlet temperature, steam flow (inlet), ambient temperature, electricity generation (gross), and condenser pressure as the input, and system’s exergy efficiency as the output. The data utilized for the training of the ANN was taken from commissioning period (early operation of the plant), after First Year Inspection (Turnaround) of the plant, and present data. The predictive capability of the model was evaluated in terms of correlation coefficient (R), mean squared error (MSE), and mean absolute percentage error (MAPE) between the ANN model data prediction and plant real-time data. It was found that the ANN model can predict turbine and overall plant exergy efficiency with good result. However, the exergy efficiency for hot well pump HWP, condenser, and cooling tower (CT) was found to be in poor performance, while the GES’s exergy efficiency shows a moderate result. This experiment has shown that the ANN method can be used and developed to predict the exergy efficiency of the geothermal power plant. Nevertheless, several improvements required for future experiment to achieve a better performance of the ANN. The selection of input parameter, the type of ANN’s algorithm used for training, the number of hidden layers used in the model, the number of outputs, or data filtering mechanism are among the required improvement to achieve better prediction results. The ANN provides an effective analyzing and diagnosing tool to understand and simulate the non-linear behavior of the plant and can also be used as a valuable performance assessment tool for plant operators and decision makers. |