| dc.description.abstract | In the global energy market, there is an increasing pressure on industries to digitise and develop efficient methods of simulating and controlling plants and components, one of the many reasons being the prediction that energy consumption will increase by 50% by 2050. To this end, one of the most cutting-edge technologies to emerge in recent years is the Digital Twin. This thesis describes the modelling and development project of a digital twin for gas turbines owned by Equinor, used to produce the energy required to operate offshore plants. The aim of the project is therefore to create a digital twin in Python to simulate and predict the behaviour of two gas turbines for offshore power generation, with the aim of implementing this model for predictive maintenance purposes. In the first phase of the project, the code for the computational model of the components of a gas turbine is developed using object-oriented programming in Python, designing specific methods and functions for the detailed description of the thermodynamic behaviour of the compressor, combustion chamber, turbine and air and fuel flows, using the NeqSim library for the determination of fluid properties. Subsequently, the computational model is validated through a comparative analysis with the Aspen HYSYS® software, developing a single-shaft gas turbine case study, under different operating conditions. Following the validation of the prediction and calculation methods for a generic gas turbine, the digital twin modeling is carried out under design conditions for the two real turbines under study: GE LM2500 and GE LM6000, through the use of Thermoflow©, which provides the main inputs to build the model in Python and the outputs to validate the digital twin itself, and GasTurb© to determine the polytropic and isoentropic efficiencies of the components in design conditions. Next, a model for the off-design behaviour at varying ambient temperature of the GE LM2500 turbine is developed, using the off-design model outlined in Thermoflow© as a reference. Finally, since one of the most important and significant applications of digital twins is predictive maintenance, a brief analysis of the indicator for detecting the degradation of gas turbine components is presented, with an initial and illustrative comparison of the results with field data, provided by Equinor. The results show that the digital twin developed in Python under design conditions for both turbines, and under off-design conditions for the GE LM2500 turbine, produce outputs that deviate from Thermoflow© by less than 1%, while also providing detailed data for individual components, such as temperatures and power required and generated at each turbine stage, fluid composition, etc. | |
