Geothermal power plant system performance prediction using artificial neural networks

Abstract

This paper describes an application of a feedforward backpropagation artificial neural network (ANN), to predict Geothermal Power Plant (GPP) Unit 4 Kamojang performance under a broad range of operating conditions. The ANN model was used for predicting specific steam consumption (SSC) of the plant, using 10 input parameters such as steam input parameters, turbine-generator parameters, condenser parameters, cooling tower parameters, and ambient parameters. The ANN model was trained with 2 different data combination. The first model was trained using commissioning data and after plant major overhaul data (ANN model 1) and the other model was trained only with after plant major overhaul data (ANN model 2). 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. The ANN model was tested using normal operation data taken during Feb-April 2015. During the testing stage, even though both ANN models performance yield moderate results, ANN model 2 shows a similar correlation (same positive or negative gradient) with the plant real time data. The difference between ANN model 2 SSC prediction and plant actual real time shows a significant difference. The experiment shows that there is 24 T/h of steam flow or equals to 3.4 to 4 MW (using SSC range 6–7 T/MWh) difference between venturi steam flow reading and ANN prediction. Combined with good and sufficient training data, and an independent measurement of steam flow for validation, the neural network approach can be utilized to develop a good performance program that able to identify the degradation of the plant or instruments (in this case is the steam flow instrument, venturi tube). If the data from the instrument reading show noticeable shift from ANN predicted value, then it can be a good sign to perform thorough analysis on the plant to prevent losses, especially in steam sales contract scheme.