Software-in-the-loop applications for improved physical model tests of ocean renewable energy devices using artificial intelligence

Student thesis: Doctoral Thesis

Abstract

Experimental research in laboratory is a necessary and useful method to explore the full potential of a device. Because it does not only require much less money than the prototype at sea test, it also provides more reliable results compared to numerical simulations. Hence, it is significantly vital to make accurate model tests of the concerned ocean renewable energy (ORE) devices possible. For this reason, this study for a PhD degree has been finished and a thesis, therefore, is produced. There is a need for a method to provide linear or nonlinear real-time power-take-off forces to the wave energy converting mechanism in the water during the experiment. More urgently, it is essential to overcome the discrepancy caused by following Froude-scaling law and Reynold-scaling law in the test of a model-scaled FOWT. Two applications for WECs and FOWTs are proposed separately, to meet the challenges.Following the conceptual design of the software-in-the-loop (SIL) application for a WEC, an innovative generic platform, which can explicitly provide a real-time PTO damping force in terms of either linear or non-linear (at different scales) is developed and characterised by 1349 drop tests. Subsequent physical model tests of a OWSC WEC device are carried out. The power efficiency of the OWSC WEC device under different PTO strategies is then estimated based on the analysis of experimental results. The best linear damping in regular waves is driven by gaining 80 in the control function, while 160 for nonlinear PTO damping. Furthermore, it is revealed that nonlinear PTOs have no distinct advantage in the amount of electricity output, but can lead to better stability and broader damping range. Following the conceptual design of an AI-based hybrid testing application for a FOWT system, a prediction module of the rotor thrust is needed to be estimated and optimised in the first place. For this reason, a considerable amount of simulations under various conditions are carried out by fully-coupled computation software, and the results obtained are used to train an artificial intelligence structure. Then a prediction module which depends on five inputs, and gives one output rotor thrust, is estimated mathematically. The mathematical module is converted to the control function in the program in a controller to execute it in real-time tests. Therefore, the AI machine is sometimes referred to as the SIL application for FOWTs, which consists of a prediction module obtained by AI training, a controller, and the program in the controller. The AI machine is the key component to implement the AI-based real-time hybrid model (AIReaTHM) testing methodology.As one of the highlights in the present study, the AIReaTHM testing rig is developed, and bench tests are carried out with a manoeuvrable motion simulator. The comprehensive testing results are analysed for three purposes: 1, validating the AIReaTHM testing methodology. 2, assessing the influences of wind speed, wind turbulence intensity, wave spectrum, input hydrodynamic motions on rotor thrust are reflected by the SIL application.3, evaluating the systematic uncertainty in the testing rig, which is to be compensated by further improving the testing system. The effect of the surge frequency, wave spectrum and wind models have on the targeted thrust is discussed. The time delay in the testing system is identified as within 0.1s, and the overall uncertainty from the testing rig is 5-15KN (the minimum rotor thrust is 508KN, hence the uncertainty is 0.98%-2.95% in percentage) when compared to the AI prediction.The testing rig developed is further applied to a 1:73 model of a Hywind floating wind turbine. 4 testing campaigns are carried out, and 303 independent tests are conducted. Testing results with the real-time rotor thrust provided by the AI-based software-in-the-loop application are compared with the other
Date of Award28 Jul 2020
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde
SupervisorDavid Clelland (Supervisor), Longbin Tao (Supervisor) & Sandy Day (Supervisor)

Cite this