Author: Chris Nelson, PhD, Data Scientist
We analyse this rich data stream using our extensive range of rigorous algorithms to identify locations along subsea cables where potentially hazardous risks are present. These risks include cable exposure, reductions in the strength of the armour, and even the onset of fatigue caused by high localised strain. The primary driver behind this analysis is that, by helping the offshore community identify these risks early enough that remedial action can occur, we significantly reduce the impact of costly subsea cable failures for stakeholders across the industry.
The beauty of this rich data stream is that, not only is it sensitive to risks felt by the cable along its length, but it is also responsive to a range of external stimuli that can be investigated without compromising our primary aim, or your operation. All of the computations and information organisation are accomplished using Indeximate’s Scattersphere®. An example of an external stimulus that we discussed in a recent blog are vessel signatures which are easily detected along cables and can now be monitored routinely using our advanced algorithms. Another example, explored here, are tonal frequencies associated with turbines in an offshore wind farm.
The stability of offshore wind turbines relies on avoidance of resonance. Resonance occurs in physical systems when an interacting element (whether environmental or engineered) vibrates with the same frequency as the natural frequency of the structure, potentially leading to a runaway increase in amplitude. Numerous videos of such resonance on bridges can be found online, sometimes with catastrophic endings. If similar resonance were to occur in an offshore wind turbine monopile, the long tower that supports the nacelle and blades, then the whole structure could topple into the sea.
The main risk of resonance in monopiles comes from the rotation of the turbine blades. When in operation, the blades complete a full rotation at the rotor frequency, which is often labelled as ‘1P’. However, as there are three blades that each rotate past the monopile on each turn of the rotor, the monopile also ‘feels’ a blade passing frequency, or ‘3P’. The monopile natural frequency must, therefore, remain firmly different to 1P and 3P. At installation this can be guaranteed, but as the dynamic marine environment scours the seabed near the turbines everything can change, bringing the monopile natural frequency closer to resonance.
A Traditional Approach to Monitoring Monopiles
The natural frequency of an individual monopile is traditionally measured using accelerometers. Accelerometers may be mounted onto a monopile tower and can measure the movement of the structure with great detail. Although this can be an effective method for turbine monitoring, some limitations are present. Examples of limitations include weather constraints on the collection of reliable data, the length of time it takes to extract results from the raw data, and monetary costs associated with buying the necessary equipment in the first place. Indeed, given the high value of the monetary costs associated with the equipment, it is typically impractical to apply accelerometers to a whole windfarm meaning only a handful of monopiles may be monitored. After this, mathematical models must be applied to extrapolate these results to the rest of the turbines, which increases uncertainty.
The Indeximate Approach
Unlike traditional methods, Indeximate’s approach does not require any specialised equipment to be attached to the turbines to identify their natural frequencies. By necessity, each turbine is connected to the offshore substation by a cable, with this cable often being strapped directly to the monopile for several 10s of meters. This means that every movement of the monopile, including at its natural frequency, is also felt by the cable. Hence, by monitoring the optical fibres within these cables using our data stream, we directly measure the natural frequency of every single turbine – at all times, with high precision, and without compromise to our primary cable health monitoring activities. In Figure 1, we plot a multi-channel spectrum plot sampled around a single wind turbine in an offshore wind farm, where the x-axis is distance along a cable and the y-axis is frequency. The vertical stripe at the centre of the image locates the turbine, whilst the thin red bar reveals the natural frequency of this monopile. The broad band of power (yellow) across the entire plot, below the natural frequency of the monopile, corresponds to the wave frequency on the sea.
Figure 1: A multi-channel spectrum plot around a single offshore wind turbine during a period of calm weather when no generation was occurring. The wind turbine can be identified by the central vertical band of increased (yellow) power. The thin horizontal band of further increased power (red) clearly identifies the natural frequency of the monopile
Of course, measuring the natural frequency of each monopile is useful and valuable by itself, but to contextualise this in terms of resonance, one would also need to be able to measure the rotor and blade passing frequencies. With our data stream, this is also possible. In Figure 1, we plot a period when the wind was low, and the rotor was not turning to best display the natural frequency of the monopile. In Figure 2, on the other hand, we plot a more confined spectral range (that does not include the natural frequency of the monopile) for a time when the turbine was in operation. Here, the rotor frequency, 1P, is clearly apparent as the sharp horizontal stripe of increased power (red).
Figure 2: Like Figure 1, but for a time when generation was occurring. The thin horizontal band of increased power (red) in this plot clearly identifies the rotor frequency, 1P, of the turbine.
The clear signals associated with the monopile’s natural frequency and the rotor frequency, means that both can be monitored accurately in real-time. In Figure 3, we plot an example of how our algorithms can be used to identify the change of the natural frequency from the target frequency at installation (green line) for ten representative turbines. These measured frequencies can be monitored to ensure they do not become too close to the rotor or blade passing frequencies. Thresholds can be set for frequencies at which remedial action may be required (red horizontal line). Clearly, these turbines remain relatively stable.
Figure 3: An example of automated natural frequency identification using our algorithms. The green horizontal line indicates the desired installation frequency, whilst the red line plots the frequency at which remedial action may be required.
If real time structural health monitoring of turbines would help your windfarm operations, don’t hesitate to get in touch via https://indeximate.com/contact-us/
