The text below is a brief description of the project topics supervised by professor Lars Imsland. Contact lars.imsland@itk.ntnu.no for further information. Most topics can be continued in a master thesis project if the candidate and supervisor(s) so wishes. Some of the descriptions are rather advanced/involved for a single project, and should be considered descriptions for a combined project/master project.

Related to a cooperation project we have with two top Brazilian universities (in Florianapolis and Rio de Janeiro), we seek 1-2 adventurous students to do a combined project/master. The specialization project (fall) will be in Norway, while the master thesis work (spring) will be done in Brazil. The topic will be related to machine learning and control for production optimization in offshore oil and gas production. Some financial support for the stay in Brazil is available.

Drillbotics (se drillbotics.com) is an international student competition in autonomous drilling. A student team (3-6 people) needs to design a robotic drilling rig that would drill through a given test rock with a given accuracy and at the least time. The rock may contain unknown layers and can cause significant operational challenges. NTNU has participated in the competition before, resulting in 2nd place in 2017 and 1st place in 2018. Similar to previous years, we are collecting a team of cybernetics and petroleum (drilling) students to compete in 2020/2021 competition (against USA, Canada, Germany, Norway and other countries). This is a very practical project, with lots of challenges on all aspects including control, measurements, modelling, etc. According to the previous teams, this was the most interesting part of their education at NTNU. The winners will be invited to present at an international drilling conference organized by Society of Petroleum Engineers.

Offshore wind on the Norwegian Continental Shelf is foreseen a massive capacity expansion in the coming years. This capacity expansion may provide value opportunities to supply offshore petroleum installations with renewable electricity generation and thereby reduce CO2 emissions associated with Norwegian petroleum activity. Yet, integration of offshore wind is likely to be performed in combination with existing or newbuilt gas-turbines, possibly together with steam-bottoming cycle (GTCC) (exhaust-gas heat recovery), which have been shown to be a promising technology also for offshore installations to reduce platform CO2 emission rates. Integration of intermittent renewables and offshore GT(CC)s requires good control strategies to enable high energy efficiency, good complementarity and stable electric-power supply. To this end, model predictive control (MPC) may be a preferred control solution. This project will seek to develop adequate reduced-order models of offshore GTs, including GTCC together with predictive models for electricity supply from offshore wind mills. Furthermore, these models will be to study the potential for MPC of the electricity supply on an offshore platform.

The successful outcome of the project may be a prospect for improved integration of renewables and GTs on platforms at the NCS, contributing to reduced CO2 emissions from offshore operations.

Co-supervisor: Brage Rustad Knudsen, SINTEF Energy.

A typical plant in the process industry is able to generate hundreds of thousands of data in real time. A lot of effort has been put lately on how to explore this vast amount of data to further develop process optimization. In model-based process optimization, hybrid models have been gaining attention lately. These kind of models are generated through the combination of knowledge-based and data-driven-based models. In this project, the student will work with a parallel pumping system. This virtual plant will be simulated with first-principle models where friction and head plays an important role. A hybrid model should be developed based on the data extracted from the virtual plant and on a simplified version of the first-principle model. Some previous knowledge in machine learning is desired.

Co-supervisor: PhD-student Otavio Ivo, NTNU

The petroleum industry has for many years used powerful mechanistic simulators to predict the multiphase flow rates from production wells. However, such simulators are complex to develop and maintain, and is often too computationally expensive to use in real-time optimization of the process. In recent years, a data-driven, AI modeling approach has become increasingly popular, particularly artificial neural networks, due to increased availability of process data. Yet, the prior knowledge of the process that the simulators represent is valuable and should not be discarded. The gray-box approach combines artificial intelligence and mechanistic modeling, utilizing the best of both worlds to enhance the predictive performance and lessen the computational demand of the model.

Co-supervisor: PhD-student Mathilde Hotvedt, NTNU and SolutionSeeker

Heat pumps are promising technologies to provide renewable heating to buildings. The focus of this project is on the control of heat pumps for residential heating purpose. Traditional heat pumps use hydrofluorocarbons and chlorofluorocarbons, which seriously damage the ozone layer. Compared with traditional refrigerants, carbon dioxide (CO2) with the advantage of non-flammability and environmental friendliness is advanced to be used in the heat pump. The well-functioned control methods for CO2 heat pumps are still lacking for improving the energy efficiency of the entire system. The problem in improving the energy efficiency of the CO2 heat pumps is in improving the control between the heat pump and the belonging heating system. Traditional real-time optimal control method (TROCM) is hard to be realized in real applications due to the large computational load. To deal with this issue, a multiplexed real-time optimal control method (MROCM) that might effectively reduce the computational load is needed. The aim of this project is to develop a MROCM for CO2 heat pumps. This project is an important part of Marie Skodowska-Curie Actions Individual Fellowship (MSCA-IF) project by the European Commission’s Horizon 2020 research and innovation program for the project ROCOCO2HP. The student will be guided and obtain necessary help and supervision from the MSCA-IF Postdoctoral Fellow. Within this thesis, the student will first develop a reliable simulation platform with local control loops of the CO2 heat pump in MATLAB/Simulink. At the Department of Energy and Process Engineering, Thermosys 5 toolbox for MATLAB/Simulink is purchased. This toolbox provides necessary components to model heat pumps and belonging heating system. After the simulation model is developed, the TROCM will be developed in this simulation platform. The model-based predictive control method will be used in the TROCM, in which a reliable system model of CO2 heat pump will be used. Finally, the developed MROCM should be realized in this simulation platform. It is desirable that the student continue with the master assignment on the same topic.

Co-supervisors: Natasa Nord and Yantong Li, EPT, NTNU

DNV GL Fuel Fighter is a project with the goal of making the world's most energy efficient electic car to compete in the Shell Eco Marathon. A big part of the performance of the car is the driving strategy. This includes when to accelerate and brake, how fast to drive and finding the most efficient way to go around corners. You will be working closely with the person developing our telemetry system and our powertrain. We recommend this as a full year project with 5th year report and master thesis, and ending with the competition in London in the summer of 2021.

See https://www.fuelfighter.no/#/thesis, and for more information about Fuel Fighter, see https://www.fuelfighter.no/#/.

An optimization algorithm for nLink’s mobile drilling robot. Based on a floor plan and a list of tasks with coordinates, the robot should follow an optimal path to cover the whole area in the shortest amount of time. The algorithm should optimize: * Task clustering / selection of positions * Shortest path between positions * Floor plan including obstacles * Total station placements give the reference layout

Co-supervisor: NN, nLink

Many engineering problems can be formulated as static optimization problems, and these are usually efficiently solved with standard algorithms like steepest descent, quasi-Newton, etc. However, if the model of our problem is bad, the output of these algorithms will likely be not be optimal for the real problem. In this task, we will study optimization algorithms that incorporate measurements to reduce the effect of model errors.