Hazard considerations in the vulnerability assessment of offshore wind farms in seismic zones

2.1 Fault rupture effects on floating wind turbines

Deterministic hazard analysis is an approach to determining approximate fault displacement based on empirical relationships. Wells and Coppersmith⁵ is a well-known correlation for this method. The formulation depends on the magnitude of an earthquake event and the type of fault rupture (e.g., Strike-Slip, Reverse, Normal). These input parameters are surface rupture length, subsurface rupture length, downdip rupture width, rupture area, maximum displacement, and average displacement. Two methods are available to carry out seismic hazard assessments. The most straightforward and most economical is Deterministic Seismic Hazard Analysis (DSHA), and the other is Probabilistic Seismic Hazard Analysis (PSHA). The DSHA approach considers a specific earthquake scenario. However, the PSHA approach adds the probability to address the uncertainties in earthquake events. The main inputs studied in this approach are scale, location, distance from the epicentre for several events in the past.⁶ Therefore, DSHA is essentially a specific case of PSHA. Different types of fault ruptures are shown in Figure 3.

Fault rupture in floating foundations could cause rupture and destabilization of the structure. Figure 4 shows two main floating systems classified based on mooring lines. In one case, the mooring line is taut, which restricts the movement, and the system is known as TLP (tension leg platform). The mooring system that allows excursion, known as the catenary mooring system (See Figure 4A). The fault rupture between the supports could cause loosening of cables and a minor tilt in the tower system. However, in a pre-stressed mooring system, for example, TLP, any amount of fault rupture between the supports could lead to a considerable amount of tilt, as shown in Figure 4B. The tilt significance is proportional to the displacement at the fault rupture location or even due to the response of earthquake motion at the surface level.^{3, 8}

2.2 Liquefaction and possible hazards

Due to the cyclic loading applied to the soil particle, the excess pore pressure will cause a reduction in the effective stress. This excess pore pressure loosens the contact between the soil particles, and as a result, a fluid-like behavior occurs in saturated soil, also known as liquefaction.⁹ An increase in the pore pressure results in effective stress reduction, and therefore, the stiffness of the soil also decreases. Therefore, this results in different types of failure mechanisms such as overall settlement, excessive rotation, and increased period of the system. Further details can be found in Amani et al.¹⁰ Lombardi and Bhattacharya¹¹ and Bhattacharya et al.^{6, 12}

2.3 Blade collision during a seismic event

One of the other possible hazards during the cyclic loading event (particularly in the seismic event) is the crash of the blade with the tower. Improving the wind turbine's rated power requires a higher wind capture area and longer blades. Longer blades during the seismic events are more likely to cause higher deformation and, therefore, collision. Besides, the importance of other parameters such as the tower specification, RNA mass, tsunami loads, and unexpected ground accelerations is not negligible. This hazard is demonstrated schematically in Figure 5.

2.4 Electrical cables failure

Cable (inter-array and export) design of OWTs in seismic regions will need some extra considerations. A study conducted by KIS-ORCA¹³ concludes that seismic activities, landslides, and current abrasion in water depths exceeding 1000 m could be associated with cable failures. For water depths less than 200 m, cable failures mainly occur due to anchoring and fishing activities. Statistics revealed that 70% of cable failures are due to human-related activities (i.e., fishing and anchoring), and the remaining is due to natural calamities, that is, earthquakes. Therefore, it is highly likely that for the current installation depths (which is less than 200 m), the risk of cables failure due to only seismic effects (such as shaking, i.e., inertia) is low. However, since OWTs are relatively expensive assets, cable failure (array cables) could cut power. Therefore, no risk should be left unconsidered.

2.5 Discussion on potential hazards for some upcoming wind farm sites

4.1 Plotting the moment (M_R) and lateral (H_R) resisting capacity curve in liquefiable and non-liquefiable soil

Offshore wind farms are currently being constructed in potentially liquefiable soil. Moreover, young offshore deposits are particularly vulnerable to soil liquefaction during strong shaking. Soil liquefaction occurs in stages where the supporting ground behaves as a heavy fluid, resulting in lateral and vertical resistance loss. Further, upward artesian flow due to excess pore pressure generation can result in high buoyant forces, resulting in cable floatation if not accounted for in the design. Aleem et al.²⁷ suggest a methodology to predict the load utilization ratio of monopiles supporting OWTs for non-liquefiable soils. This paper presents a method as an extension to adopt this method on liquefiable soils.

Figure 10 presents a simplified sketch to plot the capacity of the foundation in two stages: no liquefaction (pre-liquefaction) and maximum liquefaction (post-liquefaction). The plot also includes the design load cases. The effect of soil liquefaction is the loss of the foundation's lateral and moment-resisting capacity. Such a simplified approach estimates the capacity of pre- and post-liquefaction properties to aid preliminary sizing of offshore wind farm foundations. Once resistance (capacity) is estimated, a linear line presents the failure envelope under ULS condition for the conisdered foundation.

During earthquakes, soil deposits often liquefy top-down. The upper layers lose strength, and the liquefaction front progressively travels to deeper layers, as highlighted by Scott.²⁸ Therefore, as shown in Figure 11, it is expected that with progressive liquefaction, the foundation capacity will reduce in stages.

Therefore, considering the design life of the wind farm and tolerable risk, a probabilistic assessment can be performed for liquefaction triggering potential, thereby reducing the requirement for expensive ground remediation measures. An additional hazard due to loss of foundation capacity is the accumulated tilt of the system. As shown in Figure 12, pre-liquefaction, the ground offers sufficient resistance (capacity) to prevent excessive tilt of the superstructure. However, strong shaking can result in high inertial demands at the hub level, resulting in significant moments on the mudline. These demands can lead to tilting of the foundation, which can be exacerbated by soil liquefaction and additional long-term loading if not corrected.

Two methods could be used to conduct this assessment: (i) complete removal of the p–y springs in the liquefiable soil region and (ii) consideration of p–y springs for the liquefiable region. Several techniques currently exist in the literature to model the p–y curve for liquefied soil, including p multiplier reduced p–y curves,²⁹ C_u-factor method,³⁰ hybrid curves,³¹ and zero-strength p–y curves.³²

Strain hardening p–y curves, as a new method, consider the effects of soil dilation during the liquefaction process as mentioned in Lombardi et al.³³ or Dash et al.,³⁴ also known as the hyperelastic model. Therefore, the selection of inappropriate p–y springs could lead to overestimation or underestimation in design based on the problem analyzed. A typical p–y spring for the liquefied soil using the p multiplier (8%), reduced API p–y, and a hyperelastic p–y curve is shown in Figure 13.

These p–y curves could be modeled as springs to carry out a soil-structure interaction analysis and assess the wind turbine structure performance in seismic regions.

4.2 Examples of the moment and lateral resisting capacity curve in liquefiable soil

A sandy soil profile is considered in this example. The moment and lateral resisting capacity values were estimated using the methodology presented in Aleem et al.²⁷ For the estimation of the liquefied soil the p–y springs were removed to represent the loss of the soil strength in that regions. For an accurate representation of p–y springs the readers are referred to Lombardi et al.¹¹ or Dash et al.³⁴—the depth of liquefied soil presented by the removal of p–y springs. API³⁵ p–y springs are used to represent the non-liquefiable depth. The results shown in Figure 14 show a nonlinear reduction for the moment and lateral resisting capacity.

5.1 Load scenarios

This study considers two prominent design load cases based on typical values. These loads consist of wind, wave, earthquake, and tsunami loads.

5.2 Parametric study

The parametric study often helps to determine the level of design optimization. The sensitivity of the original wind turbine is dependent on various parameters studied. The various parameters are the diameter of the pile and the length of the pile in liquefied soil deposit. In Study 1, the pile length (L_P) varied versus the liquefied depth (L_L). Furthermore, in Study 2, the diameter of monopile (D_P) varied versus the liquefied depth (L_L). Figure 16 shows a comparison between these two studies.

The analysis considered load combination (1) with a 1% damping ratio for the structural time history analysis. Then, as part of parametric Study 1, the length of the pile (L_P) varied. Moreover, in Study 2, the diameter of the pile (D_P) also varied. Figure 17 outlines the combined outputs from parametric Studies 1 and 2 against the liquefied depth of the considered soil.

The fixed based frequency of the studied wind farm was estimated as 0.43 Hz. In this parametric study the 1^st mode frequency (Natural frequency) of the tower-pile system was estimated. Figure 17A compares the ratio between the whole structures (foundation and tower) 1^st mode frequency versus the tower fixed based frequency ratio. This demonstrates the structural stiffness, with respect to the dimensional variation. On increasing the length of the pile, the foundation frequency remained unchanged. However, increases or decreases in the pile diameter affected the pile frequency. Therefore, a change of the foundation length has a minor impact on the system frequency compared with a change in the foundation diameter.

Figure 17B shows the pile's bending strain (%) under varied dimensions. Length variation shows a minor impact on the bending strain value. The alteration of dimeters has the most impact on the structure's bending strain (%). Bending strain increases when the pile diameter reduces and reduces when the pile diameter increases.

Figure 17C,D shows the pile head displacement against the typical allowable pile head displacement (200 mm. 0.75 degrees). Therefore, this study shows that changes in the diameter and length could impact the foundation displacement and rotation.

Figure 17E shows RNA acceleration versus the considered PGA. In general, RNA acceleration is very low compared to the assigned PGA load to the system. This study shows that changes in both length and pile dimension could have a minor impact on the outcome of the results. The results indicated that RNA acceleration is 0.09 to 0.12 times the given PGA defined in this study.

In addition to Figure 17, Tables 2 and 3 show the detailed results.

In addition to the above analysis, the resilience of OWTs to Tsunami height run-up was assessed by varying the tsunami height and force in liquefied soil using load combination (2). The typical loads are assigned based on Figure 9, assuming a hydrostatic load scenario. The rise of tsunami load and height in this analysis showed a nonlinearity in the assessed parameters (e.g., pile head displacement, rotation, and the bending moment). Tsunami load was estimated based on Figure 9 for different local tsunami depths. The results are shown in Figure 18.

The output data of tsunami results shown in Figure 18 are presented in Table 4.