Alexandros Iliopoulos

Full load estimation of an offshore wind turbine based on SCADA and accelerometer data

As offshore wind farms (OWFs) grow older, the optimal use of the actual fatigue lifetime of an offshore wind turbine (OWT) and predominantly its foundation will get more important. In case of OWTs, both quasi-static wind/thrust loads and dynamic loads, as induced by turbulence, waves and the turbines dynamics, contribute to its fatigue life progression. To estimate the remaining useful life of an OWT, the stresses acting on the fatigue critical locations within the structure should be monitored continuously. Unfortunately, in case of the most common monopile foundations these locations are often situated below sea-level and near the mud line and thus difficult or even impossible to access for existing OWTs. Actual strain measurements taken at accessible locations above the sea level show a correlation between thrust load and several SCADA parameters. Therefore a model is created to estimate the thrust load using SCADA data and strain measurements. Afterwards the thrust load acting on the OWT is estimated using the created model and SCADA data only. From this model the quasi static loads on the foundation can be estimated over the lifetime of the OWT. To estimate the contribution of the dynamic loads a modal decomposition and expansion based virtual sensing technique is applied. This method only uses acceleration measurements recorded at accessible locations on the tower. Superimposing both contributions leads to a so-called multi-band virtual sensing. The result is a method that allows to estimate the strain history at any location on the foundation and thus the full load, being a combination of both quasi-static and dynamic loads, acting on the entire structure. This approach is validated using data from an operating Belgian OWF. An initial good match between measured and predicted strains for a short period of time proofs the concept.

Continuous strain prediction for fatigue assessment of an offshore wind turbine using Kalman filtering techniques

Offshore wind turbines are exposed to continuous wind and wave excitation. The continuous monitoring of high periodic strains at critical locations is important to assess the remaining lifetime of the structure. Some of the critical locations are not accessible for direct strain measurements, e.g. at the mud-line, 30 meter below the water level. Response estimation techniques can then be used to estimate the response at unmeasured locations from a limited set of response measurements and a system model. This paper shows the application of a Kalman filtering algorithm for the estimation of strains in the tower of an offshore monopile wind turbine in the Belgian North Sea. The algorithm makes use of a model of the structure and a limited number of response measurements for the prediction of the strain responses. It is shown that the Kalman filter algorithm is able to account for the different types of excitation acting on the structure in operational conditions, in this way yielding accurate strain estimates that can be used for continuous fatigue assessment of the wind turbine.

Full-Field Strain Prediction Applied to an Offshore Wind Turbine

Fatigue life is a design driver for the foundations of offshore wind turbines (OWT’s). A full-scope structural health monitoring strategy for OWT’s needs to consider the continuous monitoring of the consumption of fatigue life as an essential part. To do so, the actual stress distribution along the entire length of the structure and predominantly at the fatigue hotspots needs to be known. However installation of strain sensors at these hotspots is not always feasible since these hotspots are mainly situated beneath the water level (e.g., mudline). In practice this implies the installation of strain gauges on the monopile prior to pile driving and difficulty in maintaining these submerged sensors throughout the operational life of the turbine. Therefore, an effective and robust implemented technique using the more reliable accelerometers and very limited strain sensors at few easily accessible locations integrated within a new analytical structural dynamic approach is preferred. In this paper, a novel multi-band implementation of the well-known modal expansion approach, a.k.a. full-field strain prediction, is introduced. This technique utilizes the limited set of response data derived during a monitoring campaign and a calibrated Finite Element Model (FEM) to reconstruct the full field response of the structure. The obtained virtual responses are compared with measurements from an ongoing measurement campaign on an offshore wind turbine.

Long-term Prediction of Dynamic Responses on an Offshore Wind Turbine Using a Virtual Sensor Approach

Since fatigue life is a design driver for the foundations, the continuous monitoring for life-time assessment of an offshore wind turbine during its wide range of operational states can serve as a valuable tool for maintenance, end-of-life decisions and feedback into design for optimization of future substructures. For the offshore wind turbine, though, practical limitations prohibit to mount sensors at stress (and fatigue) hotspots. E.g. for a monopile foundation, the most popular design, the stress hot spot is at the mudline below the water level. Installing a measurement system at the mudline is unfavourable in terms of cost and maintenance. This limitation is overcome by reconstructing the full-field response of the structure based on the limited number of accelerometers and a calibrated Finite Element Model of the system. A reduced-order model that exploits the limited information obtained by the acceleration sensor data and adaptively incorporates them to permit adaptation to system changes is utilised for optimal generation of virtual dynamic strains. The model uses a multi band modal decomposition and expansion approach for reconstructing the responses at all degrees of freedom of the finite element model. The paper will demonstrate the possibility to estimate dynamic strains from acceleration measurements based on the aforementioned methodology. These virtual dynamic strains will then be evaluated and validated based on long term actual strain measurements obtained from a monitoring campaign on an offshore wind turbine on a monopile foundation. This new structural health monitoring approach has the ability to interrogate an entire structure and accurately assess fatigue life consumption and remaining useful life at the true fatigue hot spots. doi: 10.12783/SHM2015/346

Prediction of dynamic strains on a monopile offshore wind turbine using virtual sensors

The monitoring of the condition of the offshore wind turbine during its operational states offers the possibility of performing accurate assessments of the remaining life-time as well as supporting maintenance decisions during its entire life. The efficacy of structural monitoring in the case of the offshore wind turbine, though, is undermined by the practical limitations connected to the measurement system in terms of cost, weight and feasibility of sensor mounting (e.g. at muddline level 30m below the water level). This limitation is overcome by reconstructing the full-field response of the structure based on the limited number of measured accelerations and a calibrated Finite Element Model of the system. A modal decomposition and expansion approach is used for reconstructing the responses at all degrees of freedom of the finite element model. The paper will demonstrate the possibility to predict dynamic strains from acceleration measurements based on the aforementioned methodology. These virtual dynamic strains will then be evaluated and validated based on actual strain measurements obtained from a monitoring campaign on an offshore Vestas V90 3 MW wind turbine on a monopile foundation.

Online Input and State Estimation in Structural Dynamics

This paper presents two applications of joint input-state estimation in structural dynamics. The considered joint input-state estimation algorithm relies upon a limited set of response measurements and a system model, and can be applied for online input and state estimation on structures. In the first case, the algorithm is applied for force identification on a footbridge. The second case shows an application where strains in the tower of an offshore monopile wind turbine are estimated. In both cases, real measured data obtained from in situ measurements are used for the estimation. The dynamic system model, used in the estimation, is for both case studies obtained from a finite element model of the structure. The quality of the force and response estimates is assessed by comparison with the corresponding measured quantities.