Session: 06-06 Micro-Gas Turbine: Technologies and Applications

Paper Number: 124906

124906 - The Use of Artificial Neural Networks With a Systems Engineering Approach for Data Enhancement in Humidified Micro-Gas Turbine Application 

Globally the energy sector is undergoing a rapid transformation driven by innovation based on decarbonization, decentralization, and digitalization. These collectively encompass transitioning towards a future with distributed power generation systems having an increased share of renewable energy, artificial intelligence, and digital solutions. Micro-gas turbines are a competitive technology in distributed power generation owing to benefits like fuel flexibility, reduced operation and maintenance costs, emissions, noise, and vibrations. However, the economic performance of the micro-gas turbine stumbles when not operating in combined heat and power mode, requiring cycle modifications to increase electrical efficiency.

The humidified micro-gas turbine is one such modification that has shown great potential. For it to become a viable alternative in digitalized future energy systems with ever-growing amounts of data, there need to be reliable and deployable advanced tools for operation and maintenance applications. This cannot be realized without having access to relevant data with sufficient quality, which makes data quality management techniques an essential aspect of such tools. Therefore, this study investigates the use of a special type of artificial neural network called an auto-encoder network for enhancing the sensor data that has been obtained from a humidified micro-gas turbine. The objective is to preprocess sensor information by denoising it for any subsequent data-driven insight and decision. To achieve this, a comparison of different network models has been conducted to predict the denoised values of 15 sensor inputs from the experimental setup at Vrije Universiteit Brussel. From this, the auto-encoder network model that can best produce satisfactory denoised results across all provided parameters has been identified, validated, and tested. Furthermore, a systems engineering framework has been adopted for information modeling, leading to improved categorization of system elements and the effective transmittance of goals, objectives, and results amongst all stakeholders. The framework sets a foundation for employing model-based systems engineering concepts in developing digital solutions.

In conclusion, the developed model has successfully enhanced sensor measurements of the humidified micro-gas turbine. These results have proven the functionality of integrating artificial intelligence models with a systems engineering approach for data quality management. The outcomes from this work can be utilized for data enhancement in condition monitoring and predictive maintenance applications for the humidified micro-gas turbine. This will improve the system’s reliability and promote its utilization in energy systems of the future.

Presenting Author: Ahmad Jamil University of Stavanger

Presenting Author Biography: Ahmad Jamil is an energy engineer from Pakistan, currently working as a PhD Research Fellow at the Department of Energy and Petroleum Engineering, University of Stavanger, Norway. His research is focused on developing a digital solution for micro-gas turbine based distributed energy systems. Ahmad holds a master's degree in thermal energy engineering and a bachelor's degree in mechanical engineering. His main research interests include the technical, economic, and policy aspects of the energy sector, with a focus on thermal power, digitalization, decarbonization, and systems engineering. His previous research works include combined cycle gas turbine power plants, organic Rankine cycle-based waste heat recovery, end-of-life waste from renewables, and energy system planning. Ahmad has also worked as a Lab Engineer and Lecturer of thermo-fluids at Air University, Pakistan.

Authors:

Ahmad Jamil University of Stavanger
Tina Dinh Schneider Electric
Ward De Paepe University of Mons
Homam Nikpey Somehsaraei University of Stavanger