Predictive techniques applied to geothermal power plants data

Abstract

An extensive operational database is usually present in any power plant and geothermal power plants are no exception, due to the amount of information that is constantly collected from sensors and measurement parameters during the normal operation. As time goes on power plants start becoming a unique structure due to the different components in the plant and also the added efficiencies that keep changing over the any component lifetime. Thermodynamic models while always reliable tend to be less accurate over time, this research tries a different approach on predicting a component operation. The idea behind this research is to predict a component output (a turbine in this case) using a series on models based on all the data collected relevant to that particular component. It is shown that a certain data processing need to be done in order to start the analysis, this data processing is mostly to adapt the algorithms to the data analyzed, otherwise the process becomes straightforward. An event prediction model based on geothermal field reports is also considered to try to determine what causes anomalous operation in the power plant