Method for a wind farm

Abstract

The invention relates to a method, in particular a computer-implemented method, for a wind farm, comprising: determining an actual remaining useful life of a first wind turbine facility of the wind farm on the basis of a model of at least one reference wind turbine facility and on the basis of at least one first operating dataset of the first wind turbine facility; comparing the determined actual remaining useful life of the first wind turbine facility with a specified remaining useful life of the first wind turbine facility; and carrying out a performance adaptation of a predefined performance framework of the first wind turbine facility for a defined future period of time such that the actual remaining useful life of the first wind turbine facility and the specified remaining useful life of the first wind turbine facility are approximated if a predefinable deviation is identified between the actual remaining useful life of the first wind turbine facility and the specified remaining useful life of the first wind turbine facility.

Description

[0001] Procedure for a wind farm

[0002] The invention relates to a method, in particular a computer-implemented method, for a wind farm. Furthermore, the invention relates to a computing device and a wind farm.

[0003] Electrical energy is increasingly being generated using renewable energy sources these days. For example, (onshore and / or offshore) wind farms are installed with at least one wind turbine. A wind farm may also have additional structures, such as a substation, a measuring station, etc. A wind turbine is designed to convert wind energy, i.e., the kinetic energy of moving air, into electrical energy.

[0004] A wind turbine typically consists of a foundation structure, a tower, and a nacelle mounted on the tower. Wind energy is converted into electrical energy by means of a rotor, a drive train, a generator, and the like.

[0005] To ensure safe operation, a wind turbine is generally equipped with a SCADA (Supervisary Control and Data Acquisition) system that maps the operational management and safety concept, can record operating data, and is used for control purposes. Structural Health Monitoring (SHM) systems can also be used to monitor the foundation structures. SCADA and SHM systems can be integrated and supply the operating data sets. If several wind turbines are operated in a wind farm, the SCADA systems of the wind turbines are generally linked to a higher-level wind farm control and operations management system. After installation and during operation, a wind turbine is exposed to high (mechanical) loads. These loads lead to a finite service life for each wind turbine. This is primarily due to fatigue phenomena on certain (irreparable orA wind turbine consists of a number of non-replaceable components. These components are also referred to as load-bearing structural components (e.g., tower, nacelle, foundation, etc.).

[0006] In practice, such structural components of a wind turbine cannot be replaced or repaired with spare parts. This means, in particular, that replacing or repairing such a component is not carried out in practice for cost reasons, compared to a new installation of the wind turbine, even if it would be theoretically possible.

[0007] Due to the natural fluctuation of wind speeds throughout the year, wind turbines are designed for statistically representative wind speeds defined in so-called type or wind classes. The key design wind characteristics for wind turbines are the average wind speed and turbulence intensity. During operation, these wind characteristics are subject to seasonal and even multi-year fluctuations.

[0008] Today, wind turbines, especially offshore wind turbines, are designed for a defined or specified service life (e.g., 20 years). In other words, a wind turbine is constructed in such a way that the technical or actual service life of the wind turbine essentially corresponds to the specified service life (usually exceeding it for safety reasons).

[0009] To further guarantee that the technical service life actually corresponds to the specified service life (i.e., that no damage to the structural components occurs before the end of the specified service life), a performance framework is also specified for each wind turbine. The performance framework for each wind turbine specifies, in particular, at least one (maximum permissible) performance parameter limit that must not be exceeded during operation of the wind turbine.

[0010] In the current state of the art, the design of a wind turbine is generally based on the assumption that the turbine will operate continuously throughout its specified service life with at least one performance parameter limit defined in the performance framework (e.g., a specified nominal power). This assumes that the wind characteristics will not change during the operating phase and will remain within the design wind characteristics. However, it has been observed during operation that strong and weak wind years can occur, with correspondingly different energy yields. Even on a multi-year average, the design wind characteristics are generally not achieved, as the wind conditions at the respective wind farm locations are within or below the type or wind class.

[0011] It has been recognized that in practice, the technical lifetime of a wind turbine typically exceeds its specified lifetime. For example, the actual frequency of loads on a wind turbine is lower in practice. Due to maintenance work or specifications from a grid operator of an external power grid to which the wind farm is connected, or periods of low wind speeds, there may be periods in which a wind turbine is not operating within the range of at least one performance parameter limit, but may even be idle.

[0012] Nevertheless, the current state of the art stipulates that a wind turbine is always deinstalled after reaching its specified service life. Only in a few exceptional cases can continued operation be implemented for a limited period of time while complying with the performance parameter limits. The disadvantage of this is that the electrical power or electrical energy that can generally be generated by a wind turbine during its technical service life is not achieved within the specified service life of the wind turbine. In other words, the energy yield of a wind turbine during its specified service life is generally reduced compared to the energy yield that can generally be achieved by this wind turbine.

[0013] A wind farm usually consists of several wind turbines, each with a uniform technical lifetime. During the planning phase, the operating life of the wind farm is equated with the technical lifetime of the type of wind turbine used. However, it has been found that during the operational phase, the various wind turbines experience different operating conditions, e.g. wind, external throttling by grid operators, downtimes for maintenance, which lead to different ageing of the individual wind turbines in the wind farm. When the technical lifetime is reached, the respective wind turbine is usually taken out of service if the remaining service life assessment no longer permits safe continued operation. The infrastructure required for operation, e.g. ships, control rooms, operating teams, maintenance contracts, and the associated costs remain for a now reduced number of wind turbines.This reduces the economic viability of the wind farm and increases the specific electricity generation costs. The ideal economic scenario is when all wind turbines in a wind farm reach their technical lifetime at the same time.

[0014] The invention is therefore based on the object of creating a way to at least reduce the disadvantages of the prior art and, in particular, to increase the achievable energy yield of a wind turbine during the specified operating period of this wind turbine. Furthermore, the method is to be used to adjust the lifetime consumption of all wind turbines in a wind farm during the entire operating phase in such a way that the operating period of the entire wind farm corresponds as closely as possible to the respective lifetime consumption of the individual wind turbines. According to a first aspect of the invention, the object is achieved by a method according to claim 1. The method is, in particular, a computer-implemented method. The method is for use in a wind farm, in particular in an offshore wind farm. The method comprises:

[0015] Determining an actual remaining useful life of a first wind turbine of the wind farm, based on a model of at least one reference wind turbine (of the wind farm) and on at least one first operating data set of the first wind turbine,

[0016] Comparing the determined actual remaining useful life of the first wind turbine with a predetermined remaining useful life of the first wind turbine, and

[0017] Carrying out a performance adjustment of a predefined performance framework of the first wind turbine for a defined future period such that an alignment of the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine is effected upon detection of a predefined deviation between the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine.

[0018] A further aspect of the invention is a computing device comprising at least one data memory containing computer program code and at least one processor, wherein the data memory and the processor are configured such that the computing device is caused to execute the method described above.

[0019] According to the invention, the method represents, in particular, an extension of the operational management and safety concept of the wind turbine. The method is preferably implemented either individually on each wind turbine or at the level of the wind farm control and operational management.

[0020] By providing, in contrast to the prior art, a method according to the invention in which the actual remaining service life is first determined for at least one (installed) first wind turbine, then this actual remaining service life is compared with the predetermined remaining service life of this wind turbine and then, depending on the comparison result, at least one performance parameter value of a predetermined performance framework is adjusted, the disadvantages of the prior art are at least reduced and, in particular, the achievable energy yield during the predetermined service life or operating period of this wind turbine is increased.In particular, it has been recognized that by changing the performance framework (which is not conceivable in the state of the art), for example the specified nominal power of the wind turbine, the actual remaining useful life of this wind turbine can be adapted to the specified remaining useful life of this wind turbine.

[0021] Changing the performance framework of a wind turbine can include adjusting the orientation of the nacelle to the wind direction, in particular to adapt the actual remaining service life to the specified remaining service life. This can be particularly relevant in so-called spin operation in order to increase aeroelastic damping and thus reduce service life-consuming structural vibrations due to resonance phenomena. In addition, adjusting the orientation of the nacelle to the wind direction can be used in such a way that the actual remaining service life of this wind turbine is reduced, but in favor of better airflow conditions for a neighboring wind turbine in the wind farm, in order to increase its energy yield and adjust its actual remaining service life to the specified remaining service life. According to the invention, an actual or technical remaining service life (orThe remaining actual or technical service life of a first wind turbine. The actual remaining service life of the wind turbine can also be referred to as the "Remaining Useful Life" (RUL). The determination is based on a provided model of at least one reference wind turbine. In other words, the model is based specifically on at least one reference wind turbine.

[0022] At least one first operating data set of this wind turbine can be used as at least one input parameter of such a model for (directly or indirectly) determining the actual remaining useful life of a wind turbine, in particular a plurality of first operating data sets.

[0023] The at least one first operating data set may have been acquired and, in particular, recorded during the operation of the first wind turbine (before the time at which the actual remaining useful life of this wind turbine is determined).

[0024] From an operating data set of an operating period, the wind characteristics relevant to the service life can be derived and used to determine the remaining service life of this wind turbine.

[0025] An operating data record contains at least one operation-related parameter value of a (e.g., first) wind turbine. Preferably, an operating data record contains a plurality of these operating parameters (values). An operating parameter refers, in particular, to an operating parameter of the wind turbine recorded during operation of the wind turbine (for the operation of the wind turbine) that relates to the operation of the wind turbine. Preferably, a provided operating data record of the (e.g., first) wind turbine is a SCADA operating data record of the (e.g., first) wind turbine or an operating data record of the (e.g., first) wind turbine with at least one SCADA operating parameter of the (e.g., first) wind turbine.

[0026] In variants of the invention, at least one first load data set can preferably be used as a further input parameter of the model in addition to the at least one first operating data set. A load data set, in particular in the form of a fatigue load data set, can contain at least one structural and / or mechanical load parameter or load parameter value of a (previously described) structural component of a (e.g., first and / or reference) wind turbine.

[0027] A load data set can be generated in particular by an SHM system.

[0028] In particular, during the operation of a wind turbine, load parameter values ​​of at least one load parameter of this wind turbine and operating parameter values ​​of at least one operating parameter of this wind turbine can be detected (and in particular recorded) simultaneously.

[0029] A plurality of time-dependent operating data records can then be formed from the recorded operating parameter values ​​of the at least one operating parameter, in particular in a conventional manner. Each operating data record can in particular be assigned an operating data period (e.g., a timestamp) that relates to the measurement period of the measurement or recording of the at least one operating parameter. For example, an operating data period can comprise a start time (formed in particular by the date and time (e.g., 1 / 1 / 21, 1:00 a.m.)), a time length (preferably between 5 and 30 minutes, in particular (in a conventional manner) 10 minutes)), and an end time (formed in particular by the date and time (e.g., 1 / 1 / 21, 1:10 a.m.)). A plurality of time-dependent load data records can then be formed from the recorded load parameter values ​​of the at least one load parameter.In order to enable a (time-) synchronized assignment of operating data sets to load data sets, the creation of load data sets can preferably be based on the creation of the operating data sets (e.g., already specified by a SCADA system) or correspond to it.

[0030] In particular, each load data record can be assigned a load data duration (e.g., a timestamp) that relates to the measurement period of the measurement or recording of the at least one load parameter. This load data duration can be selected according to the operating data duration. For example, a load data duration can comprise a start time (formed in particular by the date and time (e.g., 1 / 1 / 21, 1:00 a.m.)), a time length (preferably between 5 and 30 minutes, in particular (conventionally) 10 minutes), and an end time (formed in particular by the date and time (e.g., 1 / 1 / 21, 1:10 a.m.)).

[0031] By synchronously or synchronously assigning load data sets to operational data sets, particularly depending on the assigned load data duration or operational data duration, input data sets for the model can be created. In other words, a temporally synchronous or synchronous assignment of temporally synchronous data sets, i.e., data sets with the same time dependency (e.g., the same time stamp), to each other can be achieved.

[0032] For example, an operating data set with a start time tAi (e.g., 1 / 1 / 21, 1:00 a.m.), a time length TLI (e.g., 10 min), and an end time tEi (e.g., 1 / 1 / 21, 1:10 a.m.) is assigned to a load data set with a corresponding start time tAi (e.g., 1 / 1 / 21, 1:00 a.m.), a corresponding time length TLI (e.g., 10 min), and a corresponding end time tEi (e.g., 1 / 1 / 21, 1:10 a.m.). Through the synchronous assignment, an input data set is created that (always) contains (exactly) one load data set (or the corresponding parameter values) and (exactly) one operating data set (or the corresponding parameter values), or is formed from them.

[0033] As will be described below, the model may preferably be a machine learning model (also referred to as a machine learning model or machine learning model). In variants of the invention, the model used to determine the actual remaining useful life may also be a different model.

[0034] The RUL can, for example (as an alternative to a machine learning model), be determined by comparing the fatigue damage (see current turbine condition) introduced into the structure of the wind turbine, i.e., the at least one structural component of the wind turbine, up to the time of assessment with the permissible damage (e.g., defined by the at least one turbine condition limit value of the at least one structural component). It is understood that the permissible damage may include appropriate safety factors. In other words, the actual turbine condition of the at least one structural component can be determined in comparison with a permissible turbine condition, i.e., compared with at least one specific turbine condition limit value.

[0035] The damage to a wind turbine can (as an alternative to the preferred machine learning model) be calculated from the wind turbine loads or stresses using a computer system based on a suitable damage model for the wind turbine. In particular, a model can be used for each component or structural component under consideration. This allows continuous measurement of component loads on all wind turbines over the entire service life of the respective wind turbine, and then the prediction can be made based on the damage model.

[0036] Furthermore (as an alternative to the preferred machine learning model), simulation models can be used to determine, in particular estimate, the RUL of a wind turbine. So-called aeroelastic simulation models can be used to estimate the component loads or component stresses of a wind turbine. A simulation model, in particular, represents a specific turbine type. The simulation models are then calibrated and validated using available operating data from a wind turbine, such as the power curve, thrust curve, foundation loads under design conditions, etc. The method can, in particular, include a calibrated aeroelastic simulation model that determines the actual remaining service life using the operating data set.

[0037] One way to determine the remaining service life of a wind turbine is to classify the operating data sets from the SCADA and SHM systems with regard to wind speeds and other operating parameters, convert them into collectives and compare them with the design characteristics in order to derive the remaining service life.

[0038] After determining the actual remaining service life of the first wind turbine, the invention compares this determined actual remaining service life of the first wind turbine with a predetermined remaining service life of the first wind turbine. A wind turbine is designed, in particular, for a defined or (fixed) predetermined service life (e.g., between 15 and 30 years, for example, 20 years). The predetermined remaining service life results from this predetermined service life and the time at which the actual remaining service life of the first wind turbine is determined. As already described, a performance framework is also specified to guarantee that the technical service life actually corresponds to the predetermined service life (i.e., that no damage to the structural components occurs before the end of the predetermined service life).In the prior art, this is particularly fixed and specified for the entire specified service life of the wind turbine. The performance framework for a wind turbine specifies, in particular, at least one maximum permissible performance parameter limit that must not be exceeded during operation of the wind turbine. In other words, the first wind turbine is controlled or regulated in such a way that exceeding the at least one maximum permissible performance parameter limit is prevented. For example, at least the (permissible) rated power of a wind turbine can be specified in the corresponding performance framework. Furthermore, the performance framework can contain a maximum permissible shutdown wind speed. The shutdown wind speed can be increased, i.e. operating periods at higher shutdown wind speeds lead to accelerated aging if these wind speeds occur during the operating phase.At these wind speeds, it would also be conceivable to operate production with reduced electrical output.

[0039] According to the invention, the performance framework is adjusted or changed depending on the comparison result. Thus, according to the invention, a performance adjustment of the aforementioned predefined performance framework of the first wind turbine is carried out for a defined future period in such a way that an alignment of the actual remaining service life of the first wind turbine and the predetermined remaining service life of the first wind turbine is achieved (only) if a predefined deviation is detected between the actual remaining service life of the first wind turbine and the predetermined remaining service life of the first wind turbine.For example, the adjustment may include changing at least one performance parameter, in particular the at least one maximum permissible performance parameter limit, of the performance framework, for example by a predetermined value and / or depending on the degree of deviation determined in the comparison between the actual remaining useful life and the predetermined remaining useful life. For example, it may be provided that the higher the determined degree of deviation, the greater the adjustment or the adjustment value of the at least one performance parameter of the performance framework.

[0040] A performance adjustment is made at least when a predefined, especially predefined (minimum) deviation has been identified. This ensures, in particular, that an adjustment is not always made for even minor deviations.

[0041] The performance adjustment is carried out for a future period. This means, in particular, that the change to the at least one performance parameter remains in place at least for the specified future period. After the expiration of the future period, the adjustment can, for example, be automatically reset, i.e., in particular, the at least one adjusted performance parameter limit value can be reset to the originally specified performance parameter limit value. As will be described below, in a preferred embodiment, upon expiration of the future period, a renewed determination of the actual remaining service life then existing at this determination time can be carried out and a subsequent comparison can be carried out in order to (re)evaluate whether a reset or a further adjustment of the performance framework should be carried out.

[0042] The future period can, for example, be between 1 month and 5 years, preferably between 6 months and 2 years, for example essentially one (operating) year. According to a further embodiment of the method according to the invention, the first wind turbine can be operated based on the adjusted power framework during the defined future period. In particular, a control device, such as the aforementioned computing device, can control or regulate the first wind turbine such that the at least one permissible power parameter limit value of the adjusted power framework is maintained during the defined future period.

[0043] According to a preferred embodiment of the method according to the invention, performing a power adjustment of a predefined power framework of the first wind turbine for a defined future period may include

[0044] Increasing at least one permissible performance parameter limit value (by a predefinable value, for example X percent, e.g. between 0.1 and 10%) of the performance framework if it is determined that the actual remaining service life of the first wind turbine is greater than the specified remaining service life of the first wind turbine by a predefinable deviation.

[0045] By increasing at least one permissible performance parameter limit of the performance framework, the energy yield can be increased in a particularly simple manner and at the same time it can be ensured that the specified remaining service life is not exceeded. Preferably, the (maximum permissible) rated power can be increased (as at least one permissible performance parameter limit), for example between 0.1% and 10%, preferably between 1% and 5%, for example by 2%. Alternatively, the shutdown wind speed can be increased slightly, e.g. from 25 m / s to 26 m / s, or instead of shutdown, operation could continue with reduced power, e.g. 50% of the rated power. Furthermore, the direction of flow of the rotor can be changed by changing the yaw angle such that the wake flow of a first wind turbine influences a second wind turbine that is at least partially in the wake to a lesser extent.

[0046] According to a further embodiment of the method according to the invention, performing a power adjustment of a predefined power framework of the first wind turbine for a defined future period may comprise:

[0047] Reducing at least one permissible performance parameter limit value (by a predefinable value, for example X percent) of the performance framework if it is determined that the actual remaining service life of the first wind turbine is shorter than the specified remaining service life of the first wind turbine by a predefinable deviation.

[0048] By reducing at least one permissible performance parameter limit of the performance framework, lifetime consumption can be reduced particularly easily, thus ensuring that the specified remaining service life is not undercut. Preferably, the (maximum permissible) rated power can be reduced (as at least one permissible performance parameter limit), for example, between 0.1% and 10%, preferably between 1% and 5%, for example, by 2%.

[0049] As already described, the described method steps of determining the actual remaining useful life, comparing the determined actual remaining useful life and performing the power adjustment can preferably be carried out several times (during the specified service life of a wind turbine), for example regularly (e.g. after each expiry of the defined future period).

[0050] According to a preferred embodiment of the method according to the invention, at least upon expiration of the defined future period, the method may comprise: re-determining the actual remaining useful life of the first wind turbine based on the model of the reference wind turbine and at least one second operating data set of the first wind turbine recorded during the defined future period, comparing the re-determined actual remaining useful life of the first wind turbine with the predetermined remaining useful life of the first wind turbine, and

[0051] Carrying out a renewed performance adjustment of a predefined performance framework of the first wind turbine for a further defined future period in such a way that an alignment of the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine is brought about, upon a renewed determination of a predefined deviation between the newly determined actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine and / or upon a renewed determination of a predefined correspondence between the newly determined actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine.

[0052] In particular, throughout the entire service life of a wind turbine, it can be determined regularly whether a power adjustment should be performed based on the respective comparison result. Depending on the comparison result, at least one permissible power parameter limit of the power framework can be reset, remain unchanged, or adjusted again, for example, increased or decreased again.

[0053] As already described, the first wind turbine is, in particular, a wind turbine of a wind farm. A wind farm can, in particular, comprise a plurality of (first) wind turbines. Preferably, the energy yield can be optimized not only with respect to one of these wind turbines, but (additionally) with respect to the entire wind farm. According to a preferred embodiment of the method according to the invention, the method can further comprise:

[0054] Carrying out the procedure (orat least the aforementioned method steps of determining the actual remaining useful life, comparing the determined actual remaining useful life and carrying out the power adjustment) for a plurality of wind turbines (in particular all wind turbines) of the wind farm, and carrying out a respective power adjustment of a predefined respective power framework of the respective wind turbine (of the plurality of wind turbines) for the defined future period, such that an alignment of the actual respective remaining useful life of the respective wind turbine and the respective remaining useful life of the respective wind turbine and an alignment of the respective remaining useful lives to one another is brought about, upon detection of at least one predefinable deviation between the respective actual remaining useful life of the respective wind turbine and the respective predetermined remaining useful life of the respective wind turbine.

[0055] In other words, an adjustment can be made in particular such that preferably all wind turbines each have an actual service life that is essentially identical to one another. This can ensure that, if possible, all wind turbines in a wind farm reach the end of their service life at essentially the same time. Particularly in the case of an offshore wind farm, it has been determined that continuing to operate only a portion of the wind turbines is disadvantageous due to cost. By now implementing a performance optimization for the entire wind farm in addition to an individual performance optimization, the energy yield of the entire wind farm can be increased even further. According to a further preferred embodiment of the method according to the invention, the method can further comprise:

[0056] Extending the specified remaining useful life of the first wind turbine if it is determined that the actual remaining useful life of the first wind turbine is greater than the specified remaining useful life of the first wind turbine by a predefined deviation.

[0057] In particular, the specified remaining service life of a wind turbine can be adjusted to the actual remaining service life. For example, the specified remaining service life can be increased or reduced accordingly. This may be particularly preferred if the respective actual remaining service life of at least almost all wind turbines in a wind farm is essentially the same. Then, in particular, the specified remaining service life of the wind farm can be adjusted to the actual remaining service life of the wind farm (which can essentially correspond to the almost identical actual remaining service lives of the respective wind turbines), in particular extended accordingly.

[0058] This independently inventive aspect can be carried out in particular without the step of carrying out a power adjustment of a predefined power framework of the first wind turbine for a defined future period, such that an alignment of the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine is effected upon detection of a predefinable deviation between the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine.

[0059] In variants of the invention, the method may further comprise:

[0060] Reducing the specified remaining useful life of the first wind turbine if it is determined that the actual remaining useful life of the first wind turbine is shorter than the specified remaining useful life of the first wind turbine by a predefined deviation.

[0061] According to a particularly preferred embodiment of the method according to the invention, determining the actual remaining service life of a first wind turbine may comprise using a machine learning model of the reference wind turbine. Training the machine learning model of the reference wind turbine may comprise:

[0062] Providing a variety of operating data sets of the reference

[0063] wind turbine,

[0064] Providing a plurality of load data sets of the reference wind turbine, wherein a load data set is based on at least one load parameter measured on the reference wind turbine, and

[0065] Generating a plurality of wind turbine training datasets for training a machine learning model by synchronously mapping a respective operational dataset to a respective load dataset.

[0066] (- Providing damage statistics for identical wind turbines in other wind farms

[0067] Providing damage statistics for identical wind turbines in the wind farm).

[0068] In particular, the method may include a method for training a machine learning model. In other words, a machine learning model can be trained. The trained machine learning model can then be used, in particular indirectly, to determine the remaining service life of a wind turbine.

[0069] The machine learning model (also referred to as a machine learning model or machine learning model) can preferably be and / or comprise an artificial neural network. An artificial neural network comprises artificial neurons. Such a neuron is in particular configured to receive at least one input from at least two other artificial neurons. The at least two inputs can each be obtained via at least one connection with a specific (trainable) weighting.

[0070] An artificial neuron can further be configured to integrate the at least one input into at least one output, for example, by summing at least two received inputs and applying a function, such as a sigmoid transformation, to the sum to generate an output. The output of an artificial neuron, in particular multiplied by a connection weight of a further connection to another neuron, can form an input for this further neuron.

[0071] In particular, for the at least one reference wind turbine, a plurality of operating data sets of this at least one reference wind turbine and a plurality of (mechanical and / or structural) load data sets of this at least one reference wind turbine are provided.

[0072] As already described, a current system condition of a structural component can be determined, in particular estimated, from load parameters or the load parameter values ​​measured on a wind turbine during a measurement period.

[0073] It has been recognized that there is a fundamental relationship between the load parameter values ​​and the operating data values. In particular, it has been recognized that using special training data sets containing both load parameter values ​​and operating data values ​​from a (reference) wind turbine, a machine learning model can be trained in such a way that this relationship(s) can be detected and mapped into the machine learning model. In particular, a machine learning model can recognize patterns between load parameter values ​​and operating data values. These patterns can be mapped into the machine learning model.

[0074] This subsequently makes it possible to determine, in particular estimate, the actual remaining service life of a first (and each subsequent) wind turbine (at least indirectly) using (exclusively) the operating data of this first (or each subsequent) wind turbine. Thus, at least one operating data set, preferably a plurality of operating data sets, of a first wind turbine can be input into the machine learning model, i.e., in particular, made available as input(s). As output, at least one turbine condition data set can be provided by the machine learning model, which may already contain an (estimated) remaining service life or at least one turbine condition indicator (e.g., at least one fatigue load indicator) from which the remaining service life can be determined, in particular, estimated in a conventional manner.

[0075] To create a plurality of training data sets, a plurality of operating data sets and a plurality of load data sets of a reference wind turbine can be provided.

[0076] A reference wind turbine refers, in particular, to an operating wind turbine, particularly within the wind farm that also includes the first wind turbine, whose operating parameter values ​​and load parameter values ​​are recorded to create training data sets. A reference wind turbine can be selected in any manner from a plurality of (first) wind turbines within a wind farm.

[0077] An operating data record of the reference wind turbine can contain at least one operation-related parameter value of the reference wind turbine. An operating data record preferably contains a plurality of these operating parameters (values). As described, an operating parameter refers in particular to an operating parameter of the wind turbine recorded during the operation of a wind turbine (for the operation of the wind turbine) that relates to the operation of the wind turbine. A provided operating data record of the reference wind turbine is preferably a SCADA or SHM operating data record of the reference wind turbine or an operating data record of the reference wind turbine with at least one SCADA or SHM operating parameter of the reference wind turbine.

[0078] A load data set of the reference wind turbine, in particular in the form of a fatigue load data set, can contain at least one structural and / or mechanical load parameter or load parameter value of a (previously described) structural component of the reference wind turbine.

[0079] In particular, during the operation of a reference wind turbine, load parameter values ​​of at least one load parameter of the reference wind turbine and operating parameter values ​​of at least one operating parameter of the reference wind turbine can be recorded simultaneously.

[0080] A plurality of time-dependent operating data sets can then be formed from the recorded operating parameter values ​​of the at least one operating parameter, in particular in the manner described above.

[0081] A plurality of time-dependent load data sets can then be created from the recorded load parameter values ​​of the at least one load parameter, in particular in the manner described above. In order to enable, in particular, a (temporally) synchronized assignment of operating data sets to load data sets, the creation of load data sets can preferably be based on or correspond to the creation of the operating data sets (e.g., already specified by a SCADA system).

[0082] To generate wind power training datasets, a synchronous or synchronized assignment of load datasets to operational datasets can be performed, in particular depending on the respective assigned load data duration or operational data duration. In other words, a temporally synchronous or synchronized assignment of temporally synchronous datasets, i.e., datasets with the same time dependency (e.g., the same timestamp), to each other can be performed.

[0083] Through the synchronous assignment, in particular, a wind power training data set can be created which (always) contains (exactly) one load data set (or the corresponding parameter values) and (exactly) one operating data set (or the corresponding parameter values) or is created from them.

[0084] According to one embodiment of the method according to the invention, the machine learning model can be trained (in a conventional manner) with the generated wind power training data sets during a training period. In other words, the generated wind power training data sets are provided to the machine learning model, in particular for training purposes.

[0085] According to a further embodiment of the method according to the invention, a portion (for example, between 10% and 30%, preferably between 15% and 25%) of the generated wind power training data sets can be used as validation data sets during a training period. In particular, the training success of a machine learning model can be checked, i.e., in particular, validated, using validation data sets. The validation data sets can preferably be randomly selected from the generated wind power training data sets. According to a preferred embodiment of the method according to the invention, a provided operating data set can contain at least one operating parameter (or a recorded operating parameter value) of a (e.g., first or reference) wind turbine. The at least one operating parameter can be selected from the group comprising:

[0086] Tower top acceleration (also called tower top acceleration), pitch angle (also called pitch angle),

[0087] Pitch speed (in particular the speed at which the blade is adjusted)

[0088] Rotor speed (also called rotor speed), Electrical power, especially electrical active power (also called active power),

[0089] Nacelle windspeed(s) (also called nacelle windspeed(s)).

[0090] Other operating data includes, for example, wind speed frequency and turbulence intensity, power production, temperatures, vibrations, and electrical currents. Alternatively or additionally, operating data can include SHM (Structural Health Monitoring) data, such as structural stresses, natural frequencies, and, in particular, changes in crack propagation rates on components, e.g., based on observed crack propagation in past years of operation or simulated damage patterns, statistics and repairs of service-relevant components, e.g., complex failures in rotor blade laminates that are not accessible through computational analysis, occurrence, frequency, and duration of operating events such as emergency shutdowns, cold starts, structural resonance effects, e.g., vortex-induced vibrations, grid losses, curtailments by grid operators, wind farm operators, or power plant controllers.

[0091] The operating data set can comprise only one of the aforementioned operating parameters, preferably two of the aforementioned operating parameters, particularly preferably all of these operating parameters. It is understood that additional operating parameters not explicitly mentioned here can also be provided alternatively or additionally. In particular, at least one sensor module, in particular a plurality of (different) sensor modules, can be provided to measure the at least one operating parameter or its operating parameter values ​​(preferably continuously). The measured operating parameter values ​​and / or signals of an operating parameter can then be provided to the SCADA system, for example, via a measuring amplifier.

[0092] Preferably, an operating data record has an operating data duration as previously described.

[0093] According to a further embodiment of the method according to the invention, an operating data record of a (e.g. first or reference) wind turbine can comprise, as an operating parameter value of an operating parameter, at least one operating parameter value which is selected from the group comprising: maximum operating parameter value recorded during (the time length) of the operating data period of the operating data record, minimum operating parameter value recorded during (the time length) of the operating data period of the operating data record, operating parameter mean value determined from the operating parameter values ​​recorded during (the time length) of the operating data period of the operating data record, standard deviation determined from the operating parameter values ​​recorded during (the time length) of the operating data period of the operating data record.

[0094] An operating data set can comprise two of the aforementioned operating parameter values ​​for the at least one operating parameter, preferably all of these operating parameter values. This allows higher-quality training data sets to be created. A preferred example of an operating data set for a (e.g., first or reference) wind turbine is given in Table 1.

[0095] Table 1 The preferred operating parameters described above refer in particular to onshore wind turbines.

[0096] It has been recognized that, for an offshore wind turbine, an operating data set should include at least one additional (offshore-related) operating parameter, or a further operating data set should be provided with at least one additional (offshore-related) operating parameter. In particular, it has been recognized that, for an offshore wind turbine, an operating data set (or another offshore operating data set that can be additionally assigned according to the described synchronous assignment) can contain at least one water status parameter. A water status parameter refers, in particular, to a parameter that indicates the status of the body of water surrounding the offshore wind turbine (in particular, the sea).

[0097] The at least one water body condition parameter is preferably a hydrodynamic parameter. The at least one water body condition parameter can be a wave parameter (e.g., wave height, wave direction, and / or wave frequency) and / or a flow parameter (e.g., flow velocity and / or flow direction).

[0098] An example of an (offshore) operating data set of a (e.g. first or reference) wind turbine with offshore-related operating parameters is given in Table 2.

[0099] Table 2

[0100] It is understood that Table 1 and Table 2 can be combined with each other. According to a further preferred embodiment of the method according to the invention, preprocessing of the provided operating parameter values ​​and / or load parameter values ​​and / or wind power training data parameter values ​​can be performed. In particular, the quality of the training data sets can be improved if the quality of the aforementioned values ​​is improved (e.g., erroneous values ​​due to measurement errors are identified in advance and then, in particular, remain unconsidered or are replaced with more suitable values ​​using suitable means).

[0101] Pre-processing may in particular include at least one of the following measures:

[0102] Removal of invalid operating parameter values ​​and / or load parameter values ​​and / or wind turbine training data parameter values ​​(or corresponding outliers). This is preferably done via direct indicators of wind turbine health (wind turbine health signal in SCADA and / or SHM, if available) and / or indirectly via the power signal in SCADA in conjunction with the power curve specification.

[0103] Extracting a suitable set of features / parameter values ​​by taking into account any collinearity between the features / parameter values. This can be done by calculating the correlations between the parameters and comparing them with a permissible upper bound. Alternatively, all features / parameter values ​​(except those that are highly collinear) can be retained and a machine learning method with implicit feature selection can be used (e.g., Bayesian LASSO (cf., e.g., Trevor Park & ​​George Casella (2008), The Bayesian Lasso, Journal of the American Statistical Association) or MARS). This can be done, for example, using a regression technique such as multivariate regression (cf., e.g., Friedman, JH (1991) “Multivariate Adaptive Regression Splines”).

[0104] Scaling of the operating parameter values ​​and / or load parameter value and / or wind power training data parameter values, e.g. standardization of the operating parameter values ​​with zero mean and, for example, constant variance (usually equal to 1).

[0105] Other transformations are also conceivable. The goal is always that the variables are approximately standard normally distributed after the transformation.

[0106] According to a further preferred embodiment of the method according to the invention, a provided load data set can be based on at least one measured load parameter of the reference wind turbine. Based means in particular that the load data set does not directly comprise the (measured) load parameter values ​​of the at least one load parameter, but rather fatigue load indicators or indicator values ​​derived or determined therefrom. The at least one load parameter of a (e.g. first or reference) wind turbine can preferably be selected from the group comprising: blade root load parameter, rotor load parameter, tower load parameter, tower torsion parameter, tower top moment parameter.

[0107] The load dataset can be based on only one of the aforementioned load parameters. The load dataset can be based on two of the aforementioned load parameters, preferably on all of these load parameters.

[0108] The at least one blade root loading parameter can in particular be a blade bending parameter (also referred to as flap bending) and / or an edge bending parameter (also referred to as edge bending). Furthermore, the at least one rotor loading parameter can be a pitching moment parameter (also referred to as tilt moment), a yaw moment (also referred to as yaw moment) and / or a rotor torque (also referred to as rotor torque). The at least one tower loading parameter can be a tower root bending parameter (in two directions). The at least one tower head moment parameter can be a pitching moment parameter (also referred to as tilt moment), a yaw moment (also referred to as yaw moment) and / or a roll moment parameter (also referred to as roll moment).

[0109] From the load parameters mentioned, a system condition or a component condition of a specific component, in particular a structural component, of a (e.g. first or reference) wind turbine can be determined.

[0110] The load parameter values ​​of a load parameter can be (continuously) measured and, in particular, provided by at least one load sensor of a sensor arrangement. In particular, the at least one load sensor can measure with a high resolution (e.g., 20 Hz). Buffering of the measured values ​​for further processing is possible. For this purpose, the at least one load sensor can be arranged at a suitable position in or on the wind turbine.

[0111] According to a particularly preferred embodiment of the method according to the invention, the method may further comprise:

[0112] Measuring the at least one load parameter of the reference wind turbine (and / or a first wind turbine), wherein the measurement of the at least one load parameter is carried out in particular in accordance with the standard 1EC 61400-13 [2].

[0113] In particular, at least the requirements specified in section 3.3.2 of the standard 1EC 61400-13

[0114] [2] mentioned (key) stress parameters are measured, in particular according to the mentioned standard.

[0115] According to a further embodiment of the method according to the invention, a load parameter value can be provided as a load parameter value of a load parameter, which is selected from the group comprising: maximum load parameter value measured during (the time length) of the load data period, minimum operating parameter value measured during (the time length) of the load data period,

[0116] Stress parameter mean value determined from the stress parameter values ​​measured during (the time length) of the stress data period, standard deviation determined from the stress parameter values ​​measured during (the time length) of the stress data period.

[0117] A load data set of a (e.g., first or reference) wind turbine can be based on at least two of the specified load parameter values, preferably on all of these load parameter values. This allows for the creation of higher-quality training data sets.

[0118] A preferred example of stress parameters and stress parameter values ​​on which a stress data set can be based is given in particular in Table 3.

[0119] Table 3 According to a further embodiment of the method according to the invention, the method may further comprise:

[0120] Measuring at least one load parameter of the reference

[0121] wind turbine during a measurement period, and

[0122] Recording the operating data of the reference wind turbine during the measurement period, wherein the measurement period is in particular at least 3 months, preferably at least 6 months (and e.g., at most 36 months). Each measured (digital) load parameter value can be assigned a timestamp of the measurement time. This can be taken into account (at least indirectly) to create a load data set and, in particular, for the synchronized assignment described above.

[0123] During a measurement period, at least one operating parameter can be continuously recorded and, in particular, measured. For example, the measured data can be temporarily stored.

[0124] As already described, corresponding datasets can be created and then made available, in particular to generate a large number of wind power training datasets for training a machine learning model. To obtain a sufficient number of training datasets (especially with an assumed time length of 10 minutes for each operating dataset (and load dataset), the measurement period can be at least 3 months, preferably at least 6 months.

[0125] For example, approximately 100,000 wind power training datasets can be generated. Preferably, approximately 20,000 of these can be used as validation datasets.

[0126] According to a further embodiment of the method according to the invention, the method may further comprise:

[0127] Forming a (time-dependent) load data set by converting the at least one measured load parameter value of the reference wind turbine into at least one fatigue load indicator, wherein the conversion is based in particular on an RFC (Rainflow Counting) method (also called Rainflow Counting method) and an LDD (Load Duration Distribution) method is used (also called Residence Time Counting method). As already described, a load data set can be based on at least one measured load parameter value of at least one load parameter of the reference wind turbine. In particular, a load data set can be based on at least one converted load parameter value, thus in particular containing the at least one fatigue load indicator.

[0128] Particularly preferably, the at least one measured load parameter value of the reference wind turbine can be converted into at least one fatigue load indicator according to the standard ASTM E1049 - 85 (2017).

[0129] To achieve a high-quality machine learning model, overfitting should be avoided during the training process. This means ensuring that the machine learning model only recognizes patterns in the training data that actually exist, rather than patterns that occurred randomly (once only).

[0130] Therefore, according to a further preferred embodiment of the method according to the invention, the invention proposes that at least one regularization technique be applied during the training period. This can at least reduce the risk of overfitting.

[0131] The at least one regularization technique can in particular be selected from the group comprising:

[0132] Linear Bayesian regression (also called Bayesian linear regression),

[0133] Bayesian neural network with concrete dropout (also called Bayesian neural network by concrete dropout) or Bayesian by Variational Inference,

[0134] Adaptive Bayesian spline regression (also called Bayesian adaptive spline regression). Such regularization techniques have proven particularly advantageous.

[0135] According to a further embodiment of the method according to the invention, a model for estimating the aleatoric and / or epistemic uncertainties can be applied during the training period.

[0136] Furthermore, during the training process, a so-called loss function can be used, for example, to evaluate the machine learning model at least partially based on the output. A loss function can, for example, contain a metric comparing the output with a reference value. A loss function can be defined, for example, when a reference output is known for a given input. For example, the loss function can be a so-called "Gaussian negative log likelihood function."

[0137] In a training step of the training process, the machine learning model can be adapted. Adapting can, in particular, involve changing at least one parameter of the machine learning model.

[0138] A wind power training dataset can preferably be in the form of text data, from which a suitable representation can be generated. The generated representation can be formed, for example, from a vector, a matrix, and / or a tensor.

[0139] As already described, the trained machine learning model can be used to determine the turbine condition of a first wind turbine, in particular to estimate it with a high degree of accuracy.

[0140] A machine learning model can, for example, be any algorithm that receives an input (in this case, in particular at least one operating data set, preferably a plurality of historical and stored operating data sets of a (e.g., first) wind turbine) and returns an output (in this case, in particular at least one system status data set, preferably containing at least one fatigue load indicator of at least one component of the (e.g., first) wind turbine), wherein the output depends both on the input and on at least one parameter of the machine learning model.

[0141] As described, the machine learning model is previously trained to learn a relationship between inputs and outputs. The trained machine learning model can then be used to generate a new output from a new input.

[0142] The input for the trained machine learning model in this case is in particular operational data sets, which can preferably be transformed into a vector, a matrix and / or a tensor for input.

[0143] According to the invention, the trained machine learning model can be used for each wind turbine in a wind farm to determine the remaining service life from the respective operating data set.

[0144] According to a further embodiment of the method according to the invention, the first wind turbine can be a wind turbine type that is identical to the wind turbine type of the reference wind turbine. In other words, if the reference wind turbine is of type VI, then the trained machine learning model can preferably be used to determine, in particular to estimate, the turbine state and in particular RUL of first wind turbines of the same or at least a similar type (i.e., type Vx). It is understood that the machine learning model trained according to the invention can still deliver acceptable results even for similar types (for example, with a different hub height or foundation structure). Alternatively or additionally, the first wind turbine and the reference wind turbine can be part of the same wind farm.In other words, the wind farm comprises a plurality of first wind turbines and at least one further wind turbine, which serves in particular as a reference wind turbine.

[0145] To determine the system status and in particular the RUL of preferably all wind turbines in a wind farm, at least one wind turbine of the wind farm (in variants of the invention, more, for example, two, can be selected) can be selected as a reference wind turbine. After training a machine learning model, as described above, the system status and in particular the RUL of at least one first wind turbine, preferably all first wind turbines, of the wind farm can be determined using the trained machine learning model.

[0146] According to a preferred embodiment of the method according to the invention, the method may further comprise:

[0147] Determining the actual remaining operating time of the first wind turbine based on the at least one turbine status data set of the first wind turbine.

[0148] As already described, the actual remaining operating time or RUL of at least one first wind turbine can be estimated with a high degree of accuracy from the output turbine status data set (in particular using already known estimation methods).

[0149] For some components or structural components of a wind turbine, such as tubular steel towers, the damage (and thus the RUL) can be estimated using structural models of the tower shell and the welds (and the turbine condition dataset). For this purpose, for example, welding details, tower geometry (e.g., thicknesses, diameters), and material properties may be known or (conservatively) estimated. However, for structural components such as rotor blades, machine frames, generator supports, and / or hubs, this would entail considerable effort to determine material quality and geometries, create FEM models, and calculate the fatigue strength. Such an approach is only useful if more detailed information can be obtained from the turbine manufacturer or if the number of turbines of the type under consideration is sufficiently large.

[0150] A variation of the further option could be to select wind conditions from the measurements that are close to the design conditions of the turbine and to take the reference loads from the measurements themselves and extrapolate them to the specified lifetime (e.g. 20 years).

[0151] According to one embodiment of the computing device, comprising at least one data memory containing computer program code, and at least one processor, the program code and the processor can be configured such that the computing device is caused to generate at least one plant status data set based on at least one provided operational data set and at least partially using a machine learning model that has been trained as described in particular.

[0152] In particular, the machine learning model can be stored in the data store.

[0153] Yet another aspect of the invention is a wind farm, in particular an offshore wind farm, comprising at least one computing device according to one of the preceding claims, and at least one first wind turbine (and preferably at least one reference wind turbine). It is understood that a wind farm may comprise further structures, such as a transformer substation, a measuring station, etc. A previously described module, element, etc. may at least partially comprise hardware elements (e.g. processor, storage means, etc.) and / or at least partially comprise software elements (e.g. executable code).

[0154] The features of the methods and computing devices can be freely combined with one another. In particular, features of the description and / or the dependent claims may be independently inventive, even if they completely or partially circumvent features of the independent claims, either alone or freely combined with one another.

[0155] There are now numerous possibilities for designing and further developing the method and computing device according to the invention. Reference is made, on the one hand, to the claims subordinate to the independent patent claims and, on the other hand, to the description of exemplary embodiments in conjunction with the drawing. The drawing shows:

[0156] Fig. 1 is a schematic view of an embodiment of a wind farm according to the present invention with a plurality of wind turbines,

[0157] Fig. 2 is a diagram of an embodiment of a method according to the present invention,

[0158] Fig. 3a is a diagram of another embodiment of a method according to the present invention,

[0159] Fig. 3b is a diagram of another embodiment of a method according to the present invention,

[0160] Fig. 4 is a diagram of another embodiment of a method according to the present invention, Fig. 5 is an exemplary plant state diagram,

[0161] Fig. 6 shows an exemplary diagram for testing collinearity by correlation, and

[0162] Fig. 7 is a schematic view of an embodiment of a computing device according to the present invention.

[0163] Figure 1 shows a schematic view of a preferred embodiment of a wind farm 100 according to the present invention with a plurality of wind turbines 102, 104. Here, reference numeral 102 denotes the (selected) reference wind turbine and reference numeral 104 denotes the first wind turbines of the wind farm 100. The wind farm 100 can be an onshore wind farm and / or an offshore wind farm.

[0164] The wind farm 100 may preferably have a (central) control device 106 (e.g., formed by the aforementioned computing device), in particular comprising at least one control module 108, for example, a SCADA control module 108. The control device 106 with the control module 108 may be configured to control the wind farm, in particular the wind turbines 102, 104, based on the respective predetermined power framework.

[0165] Furthermore, the operating parameter values ​​of at least one operating parameter of at least one wind turbine 102, 104, preferably all wind turbines 102, 104, of the wind farm 100 can be recorded and transmitted, for example, via a (wireless and / or wired) communication network 110 to the (central) control device 106. Preferred operating parameters are, for example, the parameters listed in Table 1. As already described, the at least one control module 108 can preferably control and / or regulate the wind farm 100 in a conventional manner, based at least on the respective predetermined power ranges of the respective wind turbines 102, 104.

[0166] Furthermore, the operating data records can be recorded or stored (in a data storage device (not shown) (e.g., of the control device 106)). This can be provided for documentation purposes anyway. Preferably, the operating data records of the wind turbines 102, 104 can be stored and used to determine a respective system status of the respective wind turbine 102, 104, in particular to determine the respective actual remaining operating time of the respective first wind turbine 104 (as will be described below).

[0167] As can also be seen, at least the reference wind turbine 102 has a sensor arrangement 112 with at least one load sensor (in particular a plurality of load sensors). The at least one load sensor is configured to measure at least one load parameter (e.g.

[0168] Blade root load parameters, rotor load parameters, tower load parameters, tower torsion parameters and / or tower head moment parameters).

[0169] The measured load parameter values ​​of the at least one load parameter can preferably be stored in a data memory (not shown) (e.g., the control device 106) for subsequent processing, as will be described.

[0170] Preferably, the wind turbines 102, 104 of the wind farm 100 can be of the same wind turbine type.

[0171] Figure 2 shows a diagram of an embodiment of a method according to the present invention. The method can be executed, in particular, by a computing device, which can be formed, in particular, by the control device according to Figure 1. In particular, the method can be used in a wind farm, for example, according to Figure 1. The computing device comprises at least one data memory containing computer program code and at least one processor, wherein the data memory and the processor are configured such that the computing device is caused to execute the method according to Figure 2.

[0172] In a step 201, an actual remaining useful life of a first wind turbine of the wind farm is determined based on a model of at least one reference wind turbine (of the wind farm) and on at least one first operating data set of the first wind turbine, preferably a plurality of operating data sets (as described).

[0173] In a step 202, the determined actual remaining useful life of the first wind turbine is compared with a predetermined remaining useful life of the first wind turbine (as described).

[0174] In a step 203, a power adjustment of a predefined power framework of the first wind turbine is carried out for a defined future period such that an adjustment of the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine is effected upon detection of a predefinable deviation between the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine (as described).Performing a power adjustment of a predefined power range of the first wind turbine for a defined future period may include increasing at least one permissible power parameter limit (by a predefinable value, for example, X percent) upon determining that the actual remaining service life of the first wind turbine is greater than the specified remaining service life of the first wind turbine by a predefinable deviation. Preferably, the rated power can be increased, for example, between 1 and 5%.

[0175] Subsequently, in a step 204, the control device can control the first wind turbine according to the adjusted power framework, at least during the defined future period.

[0176] At least upon expiration of the defined future period, the method can continue with step 201. In particular, the following can be carried out: re-determining (step 201) the actual remaining service life of the first wind turbine based on the model of the reference wind turbine and at least one second operating data set of the first wind turbine acquired during the defined future period; (re-)comparing (step 202) the re-determined actual remaining service life of the first wind turbine with the predefined remaining service life of the first wind turbine; and (re-)performing (step 203) a renewed performance adjustment of a predefined performance framework of the first wind turbine for a further defined future period, such that an alignment of the actual remaining service life of the first wind turbine and the predefined remaining service life of the first wind turbine is effected;if a predefined deviation is again determined between the newly determined actual remaining service life of the first wind turbine and the specified remaining service life of the first wind turbine and / or if a predefined correspondence is determined between the newly determined actual remaining service life of the first wind turbine and the specified remaining service life of the first wind turbine.

[0177] The method may further comprise performing the method for a plurality of wind turbines in the wind farm. In this case, a respective performance adjustment of a predefined respective performance framework of the respective wind turbine for a defined future period may be performed (step 203) in such a way that an alignment of the actual respective remaining service life of the respective wind turbine and the respective remaining service life of the respective wind turbine and an alignment of the respective remaining service lives to one another is effected upon detection of at least one predefined deviation between the respective actual remaining service life of the respective wind turbine and the respective predetermined remaining service life of the respective wind turbine.

[0178] Furthermore, in step 203, the predetermined remaining useful life of the first wind turbine can be extended if it is determined that the actual remaining useful life of the first wind turbine is greater than the predetermined remaining useful life of the first wind turbine by a predefined deviation.

[0179] Figure 3a shows a diagram of an embodiment of a method according to the present invention. The method is, in particular, a computer-implemented method for training a machine learning model. The machine learning model can be used to indirectly determine the remaining service life of a wind turbine, in particular to determine the structural and / or mechanical condition of the wind turbine, in particular in step 201 of Figure 1.

[0180] In a step 301, a plurality of operating data sets of a reference wind turbine can be provided. An operating data set is, in particular, a SCADA operating data set. For an onshore wind turbine, the operating data set can be formed according to the exemplary Table 1. It is understood that other, additional, or fewer operating parameters and / or values ​​can be included. For an offshore wind turbine, the operating data set can be formed according to the exemplary Tables 1 and 2. This includes, in particular, that an operating data set can be formed from two separate sub-operating data sets. It is also understood here that other, additional, or fewer operating parameters and / or values ​​can be included.

[0181] In particular, each operating data record can be assigned an operating data period that refers to the measurement period for the measurement of the at least one operating parameter. For example, an operating data period can include a start time (e.g., 1 / 1 / 21, 1:00 a.m.), a time length (e.g., 10 minutes), and an end time (e.g., 1 / 1 / 21, 1:10 a.m.).

[0182] In particular, a large number of operational data sets can be provided for the entire recording or measurement period.

[0183] In step 302, a plurality of load data sets of the reference wind turbine can be provided, wherein a load data set is based on at least one load parameter measured on the reference wind turbine.

[0184] In particular, each load data record can be assigned a load data period that refers to the measurement period for the at least one load parameter. This load data period can be selected according to the operating data period.

[0185] For example, a load data period can comprise a start time (e.g., 1 / 1 / 21, 1:00 a.m.), a time length (e.g., 10 minutes), and an end time (e.g., 1 / 1 / 21, 1:10 a.m.). In other variants of the invention (particularly if the time length is always fixed), an operating data period and / or a load data period can also be sufficiently determined by a single timestamp.

[0186] In a further step 303, a plurality of wind power training data sets can be generated for training a machine learning model. The generation occurs by synchronously assigning a respective operating data set to a respective load data set.

[0187] The (temporally) synchronized assignment of an operating data record to a load data record is carried out, in particular, depending on the respective assigned load data durations or operating data durations. In particular, the start time and / or end time can be evaluated for this purpose. For example, if, as in the example above, the start time (e.g., 1 / 1 / 21, 1:00 a.m.) and / or end time (e.g., 1 / 1 / 21, 1:10 a.m.) of an operating data record and a load data record are the same, then these operating data records can be assigned to each other.

[0188] In particular, the matching process involves creating a wind power training dataset containing the data from the operational dataset and the associated load dataset. In other words, an operational dataset can be correlated with a load dataset to create a wind power training dataset.

[0189] Example wind power training datasets WTDi, WTD2, ..., WTD n (n is a natural number for different training data sets to represent different operational data periods) are shown in Table 4. The parameter values ​​mentioned are illustrative examples.

[0190] Table 4

[0191] Here, Mean represents the mean value, Min the minimum value, Max the maximum value, Dev the standard deviation, and Ind a fatigue load indicator. It is understood that (as indicated by ...) a variety of other data may be included.

[0192] In a step 304, the machine learning model can be trained using the generated wind power training data sets. Preferably, a portion of the generated wind power training data sets can be used as validation data sets. In a particularly conventional manner, a machine learning model can preferably be trained in the form of a neural network. For input, a wind power training data set can be transformed into a vector, a matrix, and / or a tensor.

[0193] Figure 3b shows a diagram of another embodiment of a method according to the present invention. To avoid repetition, only the differences from the embodiment shown in Figure 3a are described below, and otherwise reference is made to the previous explanations.

[0194] In order to provide the at least one load data set, at least one load parameter of the reference wind turbine (i.e., of at least one component of the reference wind turbine) can be measured in step 306 during a measurement period (preferably at least 6 months). In particular, at least almost continuous measurement and thus monitoring can be performed. As already described, the measurement of the at least one load parameter can be performed, in particular, in accordance with the IEC 61400-13 standard [2]. Preferably, the load parameter values ​​of a plurality of load parameters of a plurality of components of the reference wind turbine can be measured.

[0195] In step 307, which can be performed at least partially in parallel with step 305, the at least one measured load parameter value of the reference wind turbine can be converted into at least one fatigue load indicator. Preferably, all load parameter values ​​can be converted. The conversion can be based, in particular, on a rainflow counting method. Particularly preferably, the at least one measured load parameter value of the reference wind turbine can be converted into at least one fatigue load indicator in accordance with the ASTM E1049-85(2017) standard.

[0196] In step 307, the load data sets can also be created. For this purpose, the fatigue load indicators (in particular according to the timestamp provided with the load parameter values ​​(or the corresponding sample values)) can be divided into a plurality of load data sets, each with a previously described load data period. The load data period can preferably correspond to the operating data period. In particular, an operating data period is assigned to each provided (SCADA) operating data set.

[0197] The operating parameter values ​​of the at least one operating parameter can be recorded during the measurement period and, in particular (in a known manner), provided in the form of operating data sets in step 305 (see, for example, Table 1 and / or 2). In step 308, a plurality of wind power training data sets (see Table 4) for training a machine learning model can be generated by synchronously assigning a respective operating data set to a respective load data set, as previously described.

[0198] In step 309, in particular, preprocessing or pretreatment of the generated wind power training data sets can take place. For example, invalid values ​​and / or indicators can be removed. Feature selection can also be performed in the manner described above in step 309. Examples include sequential feature selection (e.g. Ferri, FJ & Pudil, Pavel & Hatef, M. (2001). Comparative Study of Techniques for Large-Scale Feature Selection. Pattern Recognition in Practice, IV: Multiple Paradigms, Comparative Studies and Hybrid Systems. 16. 10.1016 / B978-0-444-81892-8.50040-7), recursive feature selection (e.g. Guyon, I, J Weston, S Barnhill, and V Vapnik. 2002. “Gene Selection for Cancer Classification Using Support Vector Machines.” Machine Learning 46 (1): 389-422.), linear regression with LI regulation (Lasso) and / or Bayesian LASSO (e.g.Trevor Park & ​​George Casella (2008) The Bayesian Lasso, Journal of the American Statistical Association, 103:482, 681-686, DOI: 10.1198 / 016214508000000337).

[0199] In step 310, the generated wind power training data sets are divided into training data sets (e.g., 80%) and validation data sets (e.g., 20%). Furthermore, the wind power training data sets are transformed into data representations that are suitable as input for the machine learning model to be trained, i.e., that can depend on the machine learning model to be trained (in a known manner). Preferably, a wind power training data set is transformed into a vector, a matrix, and / or a tensor.

[0200] In step 311, the machine learning model is trained or learned during a training period. Preferably, a neural network can be trained. Preferably, so-called supervised training or learning takes place using the training data sets and the validation data sets.

[0201] During the training period, a model can be applied to estimate the aleatoric and / or epistemic uncertainties, whereby the estimation can be achieved by applying at least one machine learning method. During the training period or during the training process, at least one regularization technique can be applied. This can at least reduce the risk of overfitting. The at least one regularization technique (in particular the at least one machine learning method) can in particular be selected from the group comprising:

[0202] Linear Bayesian regression (also called Bayesian linear regression),

[0203] Bayesian neural network with concrete dropout (also called Bayesian neural network by concrete dropout) or Bayesian neural network with variational inference (Bayesian neural network by variational inference),

[0204] Adaptive Bayesian spline regression (also called Bayesian adaptive spline regression).

[0205] The methods mentioned already include an estimation of aleatory and / or epistemic uncertainties. Bayesian neural networks, in particular, allow the representation of heteroscedastic uncertainties, i.e., uncertainties whose magnitude depends on the expected value of the estimated target variable (plant condition).

[0206] Alternatively (or additionally), these can be determined during the training period or during the training process using other methods (especially bootstrapping). After training, a trained machine learning model can be made available for further use in step 312.

[0207] Figure 4 shows a diagram of an embodiment of a method for using a machine learning model trained according to the present invention, for example, according to the embodiment of Figures 3a and / or 3b. The method can be performed or used in particular in step 201 of Figure 2.

[0208] In a first step 401, at least one operating data set (preferably a plurality of operating data sets) of a first wind turbine can be input into the machine learning model. Preferably, the operating data sets recorded over a specific period of time (in particular at least 2 months, preferably at least 3 months (and e.g., at most 24 months)) can be input. The first wind turbine is preferably of the same, or at least a similar, type as the reference wind turbine. Preferably, in addition, the first wind turbine can be part of the same wind farm as the reference wind turbine (see, e.g., Fig. 1). It is understood that the first wind turbine can also originate from a different wind farm.

[0209] In step 402, the machine learning model can output at least one system condition data set. The system condition data set can, for example, contain at least one determined fatigue load indicator of the first wind turbine. It is understood that a different load parameter indicator can also be output.

[0210] In a step 403, the actual remaining operating time (RUL) of the first wind turbine can be determined based on the at least one output system status data set of the first wind turbine. This includes, in particular, determining, in particular estimating, the actual RUL of at least one component of the first wind turbine. By training a machine learning model with load data and operating data of a reference wind turbine in order to recognize patterns between these data and to "store" them accordingly in the trained machine learning model, at least the system status of a first wind turbine can be determined, in particular estimated with a high degree of reliability, using the trained machine learning model solely on the operating data of this first wind turbine (i.e., without actually measuring a load parameter).As already described, the system condition of the wind turbine includes the condition of one (or more) structural component(s) of the wind turbine.

[0211] Figure 5 shows an exemplary plant status diagram, with the help of which in particular the actual remaining running time can be determined, i.e. in particular estimated.

[0212] The y-axis shows the system condition WKcond of the wind turbine or the component condition WKcond of a structural component of the wind turbine and the x-axis shows the time t.

[0213] First, it can be assumed that at the (current) time ti the (current) plant state is WKcond.i. This can be determined, in particular, according to the method shown in Figure 4.

[0214] The determined plant condition limit value WK C ond_ gThe difference can be specified. Then, using models as described, time t2 can be determined. Time t2 is, in particular, the point in time beyond which the wind turbine can probably no longer be operated (for safety reasons). The actual RUL or TRUL can then generally be calculated as follows: TRUL = t2 - t1. Figure 6 shows an example diagram illustrating the testing of collinearity by correlation, as can be performed, for example, in step 309.

[0215] As can be seen particularly in Figure 6, the correlations (expressed, for example, by the Pearson correlation coefficient in the interval [-1 1]) of the operating parameters intended for use are calculated. Operating parameters with a high correlation coefficient (typically >= 0.8 to 0.9 or <= -0.8 to -0.9) are referred to as collinear and should not be used together for training machine learning methods.

[0216] Figure 7 shows a schematic view of an embodiment of a computing device 700 according to the present invention. The computing device 700 comprises at least one processor (e.g., microprocessor, DSP, FPGA, and / or the like) and at least one data memory 720 containing computer program code. In particular, the data memory 720 contains a computer program corresponding to the (computer-implemented) method according to Figure 2 and the trained or yet-to-be-trained machine learning model 750.

[0217] Furthermore, at least one communication interface 730 and / or at least one user interface 740 is provided, configured to input and / or output data, such as operating data sets, training data sets and / or system status data sets.

[0218] The data memory 720 and the processor 710 can be configured such that the computing device 700 is prompted to generate at least one plant status data set based on at least one provided operational data set and at least partially using a machine learning model 750, which has been trained, for example, in accordance with the embodiment of Figure 3a and / or Figure 3b. In particular, the data memory 720 and the processor 710 can be configured such that the computing device 700 is prompted to execute at least steps 201, 202, 203, and preferably 204.

[0219]

[0220] October 28, 2024

[0221] List of reference symbols:

[0222] 100 Wind farm 102 Reference wind turbine

[0223] 104 first wind turbine

[0224] 106 Control device

[0225] 108 Control module

[0226] 110 Communication network 112 Sensor arrangement

[0227] 700 Calculating device

[0228] 710 processor

[0229] 720 data storage

[0230] 730 Communication interface 740 User interface

[0231] 750 model

Claims

Patent claims 1. A method, in particular a computer-implemented method, for a wind farm, comprising: Determining an actual remaining useful life of a first wind turbine of the wind farm, based on a model of at least one reference wind turbine and on at least one first operating data set of the first wind turbine, Comparing the determined actual remaining useful life of the first wind turbine with a predetermined remaining useful life of the first wind turbine, and Carrying out a performance adjustment of a predefined performance framework of the first wind turbine for a defined future period such that an alignment of the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine is effected upon detection of a predefined deviation between the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine.

2. The method according to claim 1, characterized in that performing a power adjustment of a predefined power framework of the first wind turbine for a defined future period comprises: increasing at least one permissible power parameter limit value of the power framework upon determining that the actual remaining useful life of the first wind turbine is greater than the predetermined remaining useful life of the first wind turbine by a predefined deviation.

3. Method according to one of claims 1 and 2, characterized in that at least at the end of the defined future period, the method comprises: re-determining the actual remaining service life of the first wind turbine, based on the model of the reference wind turbine and at least one second operating data set of the first wind turbine, which is recorded during the defined future period, Comparing the newly determined actual remaining useful life of the first wind turbine with the specified remaining useful life of the first wind turbine, and Carrying out a renewed performance adjustment of a predefined performance framework of the first wind turbine for a further defined future period in such a way that an alignment of the actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine is brought about, upon a renewed determination of a predefined deviation between the newly determined actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine and / or upon a renewed determination of a predefined correspondence between the newly determined actual remaining useful life of the first wind turbine and the predetermined remaining useful life of the first wind turbine.

4. Method according to one of the preceding claims, characterized in that the method further comprises: Carrying out the method for a plurality of wind turbines of the wind farm, and carrying out a respective power adjustment of a predefined respective power framework of the respective wind turbine for a defined future period, such that an adjustment of the actual respective remaining useful life of the respective wind turbine and the respective Remaining useful life of the respective wind turbine and an alignment of the respective remaining useful lives to one another is effected if at least one pre-definable deviation is detected between the respective actual remaining useful life of the respective wind turbine and the respective specified remaining useful life of the respective wind turbine.

5. Method according to one of the preceding claims, characterized in that the method further comprises: Extending the specified remaining useful life of the first wind turbine if it is determined that the actual remaining useful life of the first wind turbine is greater than the specified remaining useful life of the first wind turbine by a predefined deviation.

6. Method according to one of the preceding claims, characterized in that determining the actual remaining service life of a first wind turbine comprises using a machine learning model of the reference wind turbine, wherein training the machine learning model of the reference wind turbine comprises: Providing a variety of operating data sets of the reference wind turbine, Providing a plurality of load data sets of the reference wind turbine, wherein a load data set is based on at least one load parameter measured on the reference wind turbine, and Generating a plurality of wind turbine training datasets for training a machine learning model by synchronously mapping a respective operational dataset to a respective load dataset.

7. Method according to claim 6, characterized in that the machine learning model is trained with the generated wind power training datasets during the training period. and / or a portion of the generated wind power training datasets is used as validation datasets during a training period.

8. Method according to one of the preceding claims 6 or 7, characterized in that a provided operating data set contains at least one operating parameter of the reference wind turbine, wherein the at least one operating parameter is selected from the group comprising: Tower head acceleration, Pitch angle, Pitch speed, rotor speed, Electrical power, in particular electrical active power, nacelle wind speed, and in particular: an operating data set as an operating parameter value of an operating parameter comprises at least one operating parameter value selected from the group comprising: maximum operating parameter value recorded during the operating data period, minimum operating parameter value recorded during the operating data period, Operating parameter mean value determined from the operating parameter values recorded during the operating data period, standard deviation determined from the operating parameter values recorded during the operating data period.

9. Method according to one of the preceding claims 6 to 8, characterized in that a provided load data set is based on at least one measured load parameter of the reference wind turbine, wherein the load parameter is selected from the group comprising: Leaf root stress parameters, Rotor load parameters, Tower loading parameters, Tower torsion parameters, Tower head moment parameter, and in particular: as a load parameter value of a load parameter, a load parameter value is provided, selected from the group comprising: maximum load parameter value measured during the Load data duration, minimum operating parameter value measured during the Load data duration, Stress parameter mean value determined from the Load parameter values measured during the load data period, standard deviation determined from the values measured during the Stress parameter values measured over the duration of the stress data.

10. Method according to one of the preceding claims 6 to 9, characterized in that the method comprises: Measuring at least one load parameter of the reference Wind turbine, wherein the measurement of at least one load parameter is carried out in particular in accordance with the standard 1EC 61400-13 [2].

11. Method according to one of the preceding claims 6 to 10, characterized in that the method comprises: Measuring at least one load parameter of the reference wind turbine during a measurement period, and Recording the operating data parameter values of the reference wind turbine during the measurement period, wherein the measurement period is in particular at least 3 months, preferably at least 6 months.

12. Method according to one of claims 6 to 11, characterized in that determining the actual remaining useful life of the first wind turbine further comprises: Entering at least one first operating data set of the first wind turbine into the machine learning model, and Output, through the machine learning model, at least one turbine status data set of the first wind turbine.

13. The method according to claim 12, characterized in that the first wind turbine is a wind turbine type that is identical to the wind turbine type of the reference wind turbine and / or the first wind turbine and the reference wind turbine are comprised of the same wind farm.

14. A method according to claim 12 or 13, characterized in that the method comprises: Determining the actual remaining operating time of the first wind turbine based on the at least one turbine status data set of the first wind turbine.

15. Computing device comprising at least one data memory containing computer program code and at least one processor, wherein the data memory and the processor are configured such that the computing device is caused to carry out the method according to one of the preceding claims.

16. Wind farm, in particular offshore wind farm, comprising: at least one computing device according to one of the preceding claims, and at least one first wind turbine.

Images