Geothermal power plant performance estimation using ANFIS-PCA and ANFIS-GA

Monitoring the performance of the power plant is important to see the overall system efficiency. The performance in geothermal power plants can be viewed from the Specific Steam Consumption (SSC) value. This research aims to estimate the SSC using an artificial intelligence approach based on the Adaptive Neuro Fuzzy Inference System (ANFIS). The ANFIS’s input variables consist of 10 variables originating from the geothermal power generation sub-system, namely the steam supply and venting system (SSVS), the turbine-generator system (TGS), the steam return and condensate system (SCRS), the gas removal system (GRS), and a cooling water system (CWS). In this study, principal component analysis (PCA) and genetic algorithm (GA) are used to minimize the estimation error value and to analyze variables affecting the SSC. The evaluations of the ANFIS-PCA and ANFIS-GA models used are RMSE, MAE, and MAPE. In this study, the ANFIS-GA and ANFIS-PCA algorithms produce the same and better estimation performance than without selecting variables. The ANFIS-PCA, ANFIS-GA, and ANFIS are successfully implemented for SSC performance estimation. The RMSE evaluation showed a value of 0.0298 for ANFIS-GA and ANFIS-PCA and 0.0351 for ANFIS without variable selection.