In this paper, software-based monitoring system of mooring tension in Floating Offshore Wind Turbine (FOWT) is studied. Long-short-term memory (LSTM) model, which is one of the neural networks (NN), is used as a prediction model to monitor the top tension of the mooring line. Training and test data are collected thorough numerical simulation of OC3 Hywind spar with NREL 5MW baseline turbine. Before constructing the prediction model, feature importance is investigated by the recursive feature reduction method using cross-validation and XGboost. Moreover, the prediction accuracy of the top tension is also investigated to confirm the robustness in the environment with wind speed change. As results of this research, it is confirmed that the optimal feature combination can be narrowed down to some extent by means of the proposed method. By constructing a model that predicts the variation of the top tension and adding multiple wind speed conditions to the training data, the prediction model is confirmed to have robustness to changes in the wind speed.
In floating wind turbine, the maintenance cost accounts for 35.4% of levelized cost of energy of FOWT developments that is shown in Figure 1 (Stehly, et al., 2022). This is because it includes various systems that are not easy to repair, such as subsea power cables and mooring systems. Monitoring systems will lead to optimization of the maintenance frequencies and the reduction of the operational costs. There are existing options such as load cells and inclinometers for the monitoring. However, they have problems with installation costs and reliability. Table 1 shows the typical method of the monitoring systems. These methods were taken from the reference (Ford, et al., 2020), and only the overview of position was included in the author's research.
Only catenary theory can be used for estimation of mooring tension by the inclinometer. The theory cannot consider the dynamic load of the mooring line because it represents the statical shape of whole line by using the angle of the top chain. Based on the disadvantages of these existing methods, the purpose of this study is to estimate the dynamic tension of the mooring line from the floating motion data. Using the results of this research, it is possible to estimate the residual fatigue life of the mooring system, which can lead to optimization of maintenance frequency and period.
