EP3818384A1

EP3818384A1 - Wind energy system and method for identifying low-frequency oscillations in an electrical supply network

Info

Publication number: EP3818384A1
Application number: EP19737082.8A
Authority: EP
Inventors: Kai BUSKER; Angelo MENDONCA; Aramis SCHWANKA TREVISAN
Original assignee: Wobben Properties GmbH
Current assignee: Wobben Properties GmbH
Priority date: 2018-07-06
Filing date: 2019-07-05
Publication date: 2021-05-12
Also published as: CA3105412C; US11658490B2; US20210159705A1; CA3105412A1; CN112384813A; WO2020008036A1; DE102018116446A1

Abstract

The invention relates to a method for identifying low-frequency oscillations, in particular subsynchronous resonances, in an electrical supply network (510), wherein the electrical supply network (510) has a network voltage with a nominal network frequency, comprising the following steps: detecting at least one electrical signal of the electrical supply network (510) as at least one test signal and filtering and/or transforming the at least one detected test signal into at least one analysis signal, deriving the at least one analysis signal with respect to time or calculating the difference between temporally spaced values of the analysis signal, in order to obtain a gradient signal in each case, identifying the presence of a low-frequency oscillation if the gradient signal or at least one of the gradient signals satisfies a predefined analysis criterion, in particular that at least one predefined analysis limit is exceeded.

Description

Wind energy system and method for detecting low-frequency vibrations in an electrical supply network

The present invention relates to a method for detecting low-frequency vibrations, in particular subsynchronous resonances, in an electrical supply network. The invention also relates to a wind energy system, namely a wind energy installation or a wind farm with a plurality of wind energy installations, for carrying out a method for detecting low-frequency vibrations, in particular subsynchronous resonances, in an electrical supply network.

Wind energy systems, namely wind energy plants or wind farms, are known and they generate electrical power from wind and feed it into an electrical supply network. The proportion of electrical power that is fed into the electrical supply network by wind energy systems, in relation to the total power fed into the electrical supply network, is increasing in most electrical supply networks today. Such wind energy systems are therefore becoming increasingly important for supporting the electrical supply network.

A problem area in which support by wind energy systems would be helpful is the occurrence of low-frequency vibrations. Insofar as these low-frequency vibrations are vibrations between two elements or areas in the electrical supply network, one can also speak of subsynchronous resonances. In any case, low-frequency vibrations, including subsynchronous resonances, can be caused in particular by the vibration behavior of one or more synchronous generators directly coupled to the electrical supply network.

While conventional large-scale power plants, which have such synchronous generators directly coupled to the electrical supply network, possibly counteracting such an effect by setting or regulating their directly coupled synchronous generators, particularly taking into account the physical properties of the directly coupled synchronous generator used, modern wind energy systems feed basically into the electrical supply network by means of frequency inverters or frequency converters. Physical effects of the directly coupled synchronous generator in response to changes in the network that can directly influence current and voltage, as are known from synchronous generators directly coupled to the electrical supply network, do not occur when feeding in by means of inverters or frequency converters. Instead, the amount of current, frequency, and phase that is fed in can essentially be specified by feeding in using an inverter.

Instead, it is particularly important for wind energy systems that feed into the electrical supply network using inverters that any low-frequency vibrations to be steamed are recorded. At least it should be recorded whether low-frequency vibrations occur at all.

Particularly in the case of low-frequency vibrations, however, the low frequency of the vibrations can also result in the fact that such low-frequency vibrations can only be detected or recognized comparatively late, or make a long evaluation time necessary.

In principle, methods are known for detecting vibrations of lower frequency superimposed in a 50 Hz signal or 60 Hz signal. In particular, however, a low amplitude of such superimposed, low-frequency vibrations leads to poor recognition in practice. Known frequency analysis methods can reach their limits, especially when there is little time available for detection.

In the priority application for the present PCT application, the German Patent and Trademark Office researched the following prior art: DE 10 2014 200 740 A1 and WO 2013/102791 A1. The invention is therefore based on the object of addressing at least one of the above problems. In particular, a solution is to be proposed that enables low-frequency vibrations to be quickly recognized in the electrical supply network for wind energy systems. At least an alternative solution to previously known solutions is to be proposed. According to the invention, a method for detecting low-frequency vibrations in an electrical supply network is proposed. In particular, subsynchronous resonances are to be recognized as low-frequency vibrations, but other low-frequency vibrations are also possible. An electrical supply network is assumed that has a network voltage with a nominal network frequency. The low-frequency vibrations to be detected have a lower frequency than the nominal network frequency. The low-frequency vibrations therefore preferably have a lower frequency than the fundamental frequency of the electrical supply network. In particular, the low-frequency vibrations can have values of 1 Hz and less. However, they can also reach up to five times the nominal network frequency. Low-frequency vibrations are vibrations with a frequency of at most five times the nominal network frequency, preferably with a frequency that corresponds at most to the nominal network frequency. In particular, the low-frequency oscillation has no frequency that corresponds to a multiple of the nominal network frequency. It should be noted that the examination and consideration of low-frequency vibrations is used in particular to investigate or to ensure system stability of the electrical supply network. This differs from an assessment of the network quality or signal quality of the voltage signal in the electrical supply network, in which harmonics are particularly important.

It is now proposed that at least one electrical signal of the electrical supply network be recorded as at least one test signal. An electrical voltage and in particular or alternatively an electrical current are particularly suitable as the electrical signal. In particular, a three-phase voltage at a network connection point can be recorded as the electrical voltage, or an electrical voltage that is equivalent, in particular proportional, to the electrical voltage at the network connection point. This is particularly worth considering if a wind energy system, namely a wind energy installation or a wind farm, is used to detect the low-frequency vibrations, which is also proposed according to one aspect of the invention. This wind energy system can feed into the electrical supply network at a grid connection point and record the voltage there. Usually, a transformer is also used to transform the voltage from the wind energy system up to the level in the electrical supply network. The electrical voltage on the low-voltage side of the transformer can also be used here. An output voltage on an inverter can also be used regularly for the grid voltage, namely especially the electrical voltage at the network connection point of the electrical supply network, representative voltage.

In particular, the electrical current fed into the electrical supply network is used as the electrical current. This current is also preferably taken up in three phases. This electrical current fed into the electrical supply network can in particular be an electrical current generated by the wind energy system.

According to a further step, it is proposed that the at least one recorded test signal is filtered and / or transformed into at least one test signal. Such a filtering or transforming comes into consideration here in particular that, in addition to measuring noise, the fundamental frequency of the test signal is also filtered, namely ideally eliminated. The nominal network frequency is to be regarded as the basic frequency here, that is usually a frequency of 50 Hz or 60 Hz. However, instead or additionally, it is also possible to transform the test signal in such a way that the time course of the effective value results. This ideally results in a test signal, the amplitude of which essentially has the value of the effective value. This amplitude can fluctuate if an additional signal is superimposed on the fundamental wave. In particular, a low-frequency signal can be superimposed here, which can lead to a slight fluctuation. The detected variables are preferably transformed into space pointer sizes and the space pointer sizes thus transformed are then used further, in particular further filtered. In particular, such variables can also be constant at a constant operating point. If, however, low-frequency vibrations occur, they can be reflected in changes in these detected variables, especially in changes in the respective space vector size.

In a further step it is proposed that the at least one test signal is derived in time in order to obtain a gradient signal in each case. If the test signal reproduces the effective value of the test signal, for example, the derivation of such a test signal would ideally be 0, namely if the test signal were ideally sinusoidal and without fluctuations. If, however, at least one low-frequency oscillation is superimposed, this can emerge from the time derivative of the test signal. The gradient signal obtained by the time derivation then not only has the value 0, but essentially shows the derivation of the superimposed signal.

For this purpose, it is now proposed that the presence of a low-frequency oscillation be recognized when the gradient signal or at least one of the gradient signals if several test signals have been processed, a specified test criterion is met. In particular, it is proposed that the test criterion is a test limit that can be specified and that the test criterion is met if this specified test limit is exceeded. The evaluation is ultimately carried out via the evaluation of the gradient signal. The time derivative is therefore checked from the at least one filtered or transformed test signal and if this at least exceeds a test limit for the test signal of one of the signals, the existence of a low-frequency oscillation is assumed. Whether a low-frequency oscillation is present is therefore only recognized via the gradient formation. In particular, the gradient formation is not applied to an already recognized oscillation, for example in order to recognize whether the oscillation is increasing or decreasing, but rather the oscillation is recognized at all by the gradient formation. In particular, the comparison of the gradients recorded in this way with a test limit forms a test for absolute values. An oscillation is therefore already assumed when this test limit is exceeded. It is not necessary to consider the further development of the vibration. This also enables a quick detection to be realized, because a single exceeding of the test limit can be sufficient to detect the vibration. The criterion is also easy to implement. This method is particularly well suited for online use. Values can be recorded continuously, the at least one electrical test signal can thus be continuously recorded and it can also be continuously filtered or transformed and this in turn can be derived continuously over time. This derivation, that is to say this derivation signal, can be continuously subjected to the test criterion, in particular, this gradient signal can be continuously compared with a predefined test limit. Fulfilling the test criterion, in particular exceeding the test limit, can therefore trigger a measure immediately. In particular, a support measure can be triggered immediately. For example. power supply can be reduced immediately. It is also possible for a controller parameterization to be changed immediately, namely from a controller with a low time constant and / or weakly damped behavior to a controller with a larger time constant and / or more damped behavior. The gradient signal can also be obtained by performing a difference formation of temporally spaced values of the test signal. Insofar as there is a digital test signal, the difference between each sample value corresponds to the time derivation anyway. Here, however, it is also possible that a difference is formed for a predetermined difference period, which is preferably greater than a sampling interval. Such a difference can be successively repeated accordingly. Such differences are preferably carried out in predetermined time windows. Noise can also be suppressed in this way. Such a formation of differences in time windows can act like filtering and can be set by the size of the time window.

In particular, when checking whether a predetermined test limit is exceeded, such a difference between two temporally spaced values can be compared directly with the test limit. In principle, a fast detection of a low-frequency oscillation is also possible because a decision can be made immediately after the difference is formed, namely whether the existence of a low-frequency oscillation is assumed.

According to one embodiment, it is proposed that a gradient maximum value be specified as the test criterion and that the presence of the low-frequency oscillation is recognized if the gradient signal exceeds the gradient maximum value at least once. This allows a clear test criterion to be specified. In particular, it is proposed that the presence of the low-frequency oscillation is recognized when the gradient signal is above the gradient maximum value for at least a predetermined minimum period. Here, the duration of one or two sampling steps can already be selected as the predetermined minimum period. Then, the presence of the low-frequency oscillation is only recognized when the gradient signal is above the maximum gradient value for at least two measured values, or at least for three measured values. In this way it can be avoided that a single exceeding value, which essentially resulted from a measurement noise, triggers the detection of a low-frequency oscillation. By checking for a gradient maximum value, a corresponding oscillation can be recognized immediately as soon as the gradient, which represents the oscillation amplitude, exceeds this value. This distinguishes such a criterion from a method that detects a vibration and the further development considered the vibration, namely, whether the vibration increases, decreases or stagnates.

Such a difference becomes particularly clear in the case of a vibration that occurs suddenly, e.g. by a leap in performance of a fed-in performance. As a result, an oscillation can suddenly appear, but it decreases again. If the method for recognizing the oscillation considers the further development of the oscillation, it recognizes that the oscillation has decayed and therefore does not detect any oscillation, and therefore does not trigger a detection signal. However, if the reaching or exceeding of an absolute value, such as the gradient maximum value, is considered, the oscillation would be recognized in the case mentioned. Such criteria therefore differ significantly.

The maximum gradient value can also depend on a test period over which the gradient is formed and which is described below. The gradient maximum value is preferably also selected with regard to the test period so that it corresponds to a maximum oscillation amplitude of the electrical signal to be tested. For the mains voltage as the signal to be tested, the gradient maximum value is preferably selected from a range which corresponds to a range from 0.1% to 2% of the corresponding maximum oscillation amplitude. According to one embodiment, it is proposed that a difference between the maximum and minimum value of the test signal be used as the gradient signal in a test period under consideration. Accordingly, a test period is specified in which the gradient signal is viewed. For this purpose, the difference between two values that are separated by a predetermined time interval is not considered, but the minimum and maximum value of the test signal that occurs in this test period is considered and its difference is used as a gradient signal, in particular compared to a predetermined test limit , For example. comes into consideration that the test period is chosen so that there are 10 or more values or in particular up to 50 values of the gradient signal. This enables a clearly defined test criterion to be set up and, with a high sampling rate, the period until which the test criterion is checked is comparatively short. The test period under consideration is preferably used as a sliding window and is thus shifted one scanning step further in each new scanning step, and the test signal then located in the window is evaluated accordingly.

It should be noted that a test limit, especially for a vibration amplitude, can be used. The test limit is therefore preferably not included compared to an absolute value, but it is compared with a maximum oscillation amplitude. Such an oscillation amplitude basically describes the distance between a positive and a negative envelope of an oscillating signal. If a signal fluctuates, for example, by a value of 10, to name just one example, and in the extreme case the signal fluctuates from the value 9 to the value 1 1 and back, the oscillation amplitude would have the value 2 in this example. If, in order to stay with this illustrative example, the associated test limit, for example the value 3, this test limit would not have been reached. If the value of the test limit were 1, 5, for example, it would have been reached. According to one embodiment, it is proposed that a mains voltage of the electrical supply network, in particular three-phase, is recorded as a first test signal and a feed current, in particular three-phase, fed into the electrical supply network is recorded as a second test signal, and in particular the first and second test signal in at least one test signal be transformed. Information about low-frequency vibrations can be detected via the grid voltage, in particular at the relevant grid connection point, and the feed-in current, which a wind energy system in particular feeds into the electrical supply network. It is particularly important to take into account that a low-frequency oscillation can regularly affect the oscillation of electrical power in the electrical supply network. The input power and the input reactive power can thus be recorded by detecting the feed current and at the same time detecting the mains voltage, that is to say the voltage with which the feed current is fed in. From this in turn, the oscillation of power in the electrical supply network can be derived or it can be detected thereby. This detected mains voltage and this detected feed-in current are preferably transformed into at least one test signal. It is also possible to transform them into a common test signal or to transform them together into several test signals. A transformation into an active power signal as a test signal and / or a reactive power signal as a test signal is particularly suitable. Such a first and second test signal, that is to say the detected mains voltage and the detected fed-in feed current, are preferably transformed into a voltage signal, an active power signal and a reactive power signal, which then form a voltage test signal, active power test signal or reactive power test signal. The voltage signal then represents the mains voltage, the active power signal the active power fed in and the reactive power signal the reactive power fed in.

These three test signals are then derived in time in order to obtain a gradient signal, namely a voltage gradient signal, an active power gradient signal and a reactive power gradient signal. The time derivation can also be realized in each case by forming a difference, or instead of the time derivation, a difference is formed for values separated by time.

The voltage gradient signal, the active power gradient signal and the reactive power gradient signal are then checked for the presence of a low-frequency oscillation.

The test is carried out in such a way that the presence of a low-frequency oscillation is assumed if a low-frequency oscillation was detected at least in the voltage gradient signal and the active power gradient signal, or if a low-frequency oscillation was detected in the voltage gradient signal and the reactive power gradient signal. It is also contemplated that a low frequency oscillation is assumed if a low frequency oscillation was detected in all three gradients, i.e. if a low frequency oscillation was detected in the voltage gradient signal, in the active power gradient signal and in the reactive power gradient signal. The positive test of the presence of a low-frequency oscillation in at least two of the gradient signals mentioned avoids in particular that a measurement error or excessive noise already leads to incorrect detection of low-frequency oscillations. In particular, the simultaneous test in the voltage gradient signal on the one hand and the active power gradient signal or the reactive power gradient signal on the other hand has been recognized as advantageous because the voltage gradient signal can quickly detect fluctuations in the voltage signal, which may not be related to low-frequency vibrations. By further considering the reactive or active power, the current component is also taken into account and not only voltage fluctuations are recognized, which can also have another cause and do not necessarily have to indicate a low-frequency oscillation immediately. According to one embodiment, it is proposed that a network frequency of the electrical supply network is detected as a further test signal, the further test signal is transformed into a frequency signal as a frequency test signal, and the frequency test signal is derived in time, or a difference is formed in order to obtain a frequency gradient signal. The frequency gradient signal and in particular at least one further gradient signal are then checked for the presence of a low-frequency oscillation, in particular in such a way that the presence of a low-frequency oscillation is assumed if a low-frequency oscillation was detected in the frequency gradient signal and in at least one of the gradient signals, namely in one Voltage gradient signal, an active power gradient signal and / or a reactive power gradient signal.

It is therefore proposed here that the network frequency is recorded and evaluated as such as a signal. Such a signal would ideally be constant, especially at 50Hz or 60Hz. In fact, however, it will fluctuate and this fluctuation, to put it graphically, forms the frequency signal. Such a frequency signal can then basically be processed like the other signals described.

Here, too, it is possible not only to evaluate the frequency signal, but also to take into account at least one further signal, in particular the active power gradient signal. The detection of the low-frequency vibration can be additionally secured. By transforming the recorded voltages and currents, among others In an active power signal and a reactive power signal as test signals, these variables can also be taken into account as effective values. In particular, a transformation can take place in each case in its effective values, and then only the fluctuation of the effective values is considered by deriving them.

According to one embodiment, it is proposed that a three-phase mains voltage is recorded as a test signal for the detection of the at least one electrical signal of the electrical supply network, and a constant, in particular a space vector size of the voltage is formed therefrom via a transformation, in particular that a positive sequence system voltage according to the method of symmetrical components is determined, which forms a test signal and / or a three-phase feed-in current is detected as a test signal and a co-system current is determined therefrom according to the method of symmetrical components, which forms a test signal. The transformation thus determines a constant size, that is to say a non-oscillating variable, which can therefore also be referred to as a DC variable, which basically describes a sinusoidal signal by means of a fixed variable. As a result, the signal is essentially freed from its oscillating basic signal, and signal changes, namely deviations from the sinusoidal basic signal, emerge strongly and can thus be easily recognized and possibly processed. Despite three phases examined, only one size needs to be considered. These effects and the resulting advantages also result from a corresponding transformation of a three-phase input current, for which such a transformation is therefore also proposed. In particular, a transformation according to the method of symmetrical components is proposed for the mains voltage and / or the feed current as a transformation into a test signal, in which case only the co-system component, ie basically the symmetrical component, is considered. It was particularly recognized that this transformation using the method of symmetrical components basically results in an effective value which is therefore an equivalent value and which can be used as a description of the basic component. Actual fluctuations, in particular power fluctuations, which are not limited to asymmetries, are then superimposed on this symmetrical component and can then be easily recognized by the suggested time derivation of these test signals. Through the transformation into a co-system voltage and a co-system current, the three-phase voltage or the three-phase current is also considered to be only one component in a simple manner known from other applications. Especially for the further processing and later application of a test criterion, the question does not arise how to apply a single criterion to three phases. It was also recognized that such a transformation, in particular a transformation in co-system and counter-system components, has a filtering effect and thereby particularly high-frequency components are filtered out and that there is sufficient bandwidth for the detection of low-frequency signals. However, other transformations are also possible, such as a d / q transformation, which can also lead to an equivalent value if the reference frequency is selected accordingly.

Preferably, the same size or space vector size of the voltage or the system voltage and the size or space vector size of the current or the system current is transformed into a voltage signal, an active power signal and reactive power signal as a voltage test signal, active power test signal or reactive power test signal. In this respect, the system voltage can form the voltage test signal directly. The positive sequence current can, together with the positive sequence voltage, be further transformed into an active power signal and a reactive power signal.

This further transformation, in particular also for obtaining the active power test signal and reactive power test signal, enables the fed-in powers to be viewed in a simple and efficient manner, and in particular their fluctuations can then be viewed, in particular by the proposed time derivation or difference formation of the respective test signal ,

The electrical supply network is preferably fed into the electrical supply network by means of a wind energy plant or a wind farm, and low-frequency vibrations are detected by means of the wind energy plant or by means of the wind farm. It was particularly recognized here that a wind energy installation or a wind farm can act as very fast control units in the electrical supply network and it can therefore be advantageous to use them to detect low-frequency vibrations. Details of such wind turbines or wind farms are described below.

According to one embodiment, it is proposed that a vibration caused in the electrical supply network is recognized if a low-frequency vibration was detected in the voltage gradient signal and the active power gradient signal or in the voltage gradient signal and the reactive power gradient signal, and these detected low-frequency vibrations have the same vibration frequency, in particular additionally it is checked whether the mains frequency oscillates at the same oscillation frequency.

It was particularly recognized here that a low-frequency vibration can have different causes and that it can also depend on how such vibrations are to be handled. If a feeder, especially a wind power plant or a wind farm, has caused vibrations, the reason for the vibration is particularly to be found in the dynamics of the wind power plant or the wind farm. It can then also be assumed that the vibrations must be able to be remedied by the wind energy installation or the wind farm. The cause of the vibration then does not necessarily have to lie directly in the area of the feed, but can also relate to a mechanical vibration and / or a vibration in the generator. However, if the cause of the vibration lies in the electrical supply network, the wind energy installation or the wind farm can be used, if at all, to carry out vibration damping, that is to say reduction. In addition, such damping will essentially affect the feed. Therefore, the differentiation of the cause of vibration has been recognized as important.

For this purpose, it was further recognized that the cause of the vibrations in the electrical supply network is to be sought when at least two of the signals mentioned are recorded simultaneously. Vibrations of the grid voltage and the grid frequency in particular indicate a vibration in the electrical supply network, whereas the active power signal, i.e. the fed-in active power, and the reactive power signal, i.e. the fed-in reactive power, tend to indicate vibrations in the feed unit, particularly in the wind power plant or the Indicate wind farm.

If the oscillations of the mains voltage and / or the mains frequency have the same oscillation frequency as the oscillations of the active and / or reactive power, then there is an oscillation that affects the feed-in unit, especially the wind turbine or the wind farm, but its cause in the electrical supply network Has. For example. it may be that large electrical consumers that are connected to the electrical supply network interfere with network operation, for example due to power fluctuations and thereby trigger the vibrations. According to a further embodiment, the method is characterized in that a difference between the maximum and minimum value of the corresponding test signal is used in each test period as a voltage gradient signal, active power gradient signal and reactive power gradient signal and an identical oscillation frequency is recognized by using the same test period and / or - The respective time intervals between the maximum and minimum value of the corresponding test signal for the voltage gradient signal, the active power gradient signal and the reactive power gradient signal are the same or similar.

It was particularly recognized here that the detection of the low-frequency oscillation by means of gradient formation can be carried out in a simple and efficient manner by binding the difference between two signal values, namely the maximum and the minimum. This is based on a test period that can be set to the distance between the two values, that is, the maximum and minimum values, and thus the oscillation frequency can also be formed directly. For a check on same frequency is but that is not necessary either, then the time intervals are the same, and the assigned frequencies are also the same.

According to a further embodiment, the method is characterized in that a gradient quotient is formed as the quotient of two gradient signals and is recognized depending on the gradient quotient for an oscillation caused in the electrical supply network. Here, in particular, as also particularly in the above embodiments, the wind energy installation or the wind farm is used as the feed unit and it is examined whether a vibration caused in the electrical supply network or a vibration caused in or by the feed unit is to be assumed.

By considering the gradient quotients, a simple test option is created which can also be implemented in a control unit or detection unit with little effort. In particular, it is proposed to implement such a criterion in a central parking control. Two gradient signals can be directly related to one another without the need for complex individual evaluation of the electrical signals or the gradient signals formed therefrom.

In particular, it is proposed that a voltage-active power quotient dU / dP is formed as the gradient quotient as a quotient between the voltage gradient signal and the active power gradient signal and / or that a voltage-reactive power quotient dU / dQ is formed as the quotient between the voltage gradient signal and the reactive power gradient signal. For this purpose, it is proposed that a vibration caused in the electrical supply network is recognized when the voltage-active power quotient dU / dP and / or the voltage-reactive power quotient dU / dQ are negative. This is based on the consideration that an oscillation of the active power P as well as the reactive power Q can lead to a reaction, namely an oscillation of the mains voltage, because the mains voltage also depends on these variables. So if the active power and / or the reactive power vibrates, this can trigger a simultaneous oscillation of the mains voltage. This would lead to gradients of the same sign of the oscillation of the mains voltage on the one hand and the oscillation of the active power or reactive power on the other hand. The gradient quotient would therefore be positive. However, if the oscillation of the mains voltage U is not the result of the oscillation of the active power or reactive power, it can be expected that the mains voltage on the one hand and the active power or reactive power on the other hand oscillate in opposite directions. Your gradients would then have different signs. The gradient quotient would be negative. In the event that only one of the two quotients of the voltage-active power quotient dU / dP and the voltage-reactive power quotient dU / dQ is negative, the quotient at which the denominator is larger can be decisive for the evaluation. In most cases, however, the active power P and the reactive power Q will not oscillate against each other, so that they are fed in together as apparent power S. A further embodiment proposes a method which is characterized in that vibrations are classified and a found vibration classification is output. The following are particularly considered as vibration classifications:

A low-frequency oscillation was recognized for one or the mains voltage signal and one or the active power signal. A low-frequency oscillation was recognized for the mains voltage signal and one or the reactive power signal.

A low-frequency vibration was detected for the voltage signal, the active power signal and the reactive power signal.

A low-frequency oscillation was recognized for the mains frequency and also for the voltage signal, the active power signal and / or the reactive power signal.

It is thus proposed that the vibration classifications simply indicate the signals for which a low-frequency vibration was recognized. The recipient of this vibration classification can then derive further qualified conclusions from this. For example. An oscillation of the mains voltage together with an oscillation of the fed-in active power can indicate a power oscillation in the electrical supply network, or an oscillation triggered by a change in the active power balance in the electrical supply network. On the other hand, an oscillation of the mains voltage together with an oscillation of the fed reactive power may indicate a problem of voltage stabilization in the electrical supply network, to name another example.

According to the invention, a wind energy system is also proposed. The wind energy system can be a single wind energy installation or a wind farm with several wind energy installations. It is prepared for the detection of low-frequency vibrations, in particular subsynchronous resonances, in an electrical supply network. The electrical supply network is based on the assumption that it has a network voltage with a nominal network frequency and that the low-frequency vibrations to be detected have a lower frequency than the nominal network frequency. The proposed wind energy system comprises a detection device for detecting at least one electrical signal of the electrical supply network as at least one test signal. In particular, a measuring device at the output of an inverter and / or at the grid connection point at which the wind energy system feeds into the electrical supply network is proposed here. The detection device is preferably set up to detect a voltage, in particular the mains voltage of the electrical supply network. In addition or alternatively, it is preferably proposed that the detection device be provided for detecting an electrical current that is fed in. Furthermore, the wind energy system comprises a filter unit for filtering and / or transforming the at least one detected test signal into at least one test signal. Here, in particular, a digital filter or a digital transformation unit comes into consideration, in particular to filter or transform the at least one test signal, if this is present as a digital signal. In particular, a filter comes into consideration which filters out frequencies which are at and above the nominal network frequency in order to achieve in particular that frequencies of the expected low-frequency vibrations are retained. In particular, however, it is also possible to carry out a transformation which also functions in whole or in part as a filter in order in particular to transform an effective value of the detected variable. Such a transformation into the effective value can be considered, in particular, when a line voltage or a current fed in is recorded as a test signal. Basically, the fundamental is filtered out or transformed out of the respective test signal.

Furthermore, a derivation unit is provided for deriving the at least one test signal in time in order to obtain a gradient signal in each case. Every test signal is thus through changed this derivation unit into a gradient signal. The derivation can also be carried out by forming a difference or, instead of deriving, a difference can be formed by temporally spaced values of the test signal. For this purpose, a time interval can be specified, for example, by which two values of the test signal are to be spaced, if the difference is to be carried out between them. Ideally, only a derivation of any low-frequency vibrations present is then present in the gradient signal, when the basic signal, that is to say in particular a 50 Hz signal or a 60 Hz signal, has ideally been removed by the filter unit. The respective signals, in particular the superimposed signals, can be better recognized by the derivation.

In addition, a detection unit is provided for detecting the presence of a low-frequency vibration. This works in such a way that when the gradient signal or at least one of the gradient signals fulfills a predetermined test criterion, the presence of a low-frequency oscillation is recognized. In particular, a test limit is specified as the specified test criterion and the test criterion is met if the at least one specified test limit is exceeded. In particular, if a predetermined test limit was exceeded by the gradient signal.

The wind energy system, in particular the detection unit, thus operates in such a way that low-frequency vibrations are detected by considering and evaluating the change of at least one filtered test signal.

The detection device, the filter unit, the derivation unit and / or the detection unit can preferably be combined in a process computer and in particular in a control device. It is also possible to consider these elements as program code. In particular, the wind energy system works in such a way that it implements a method according to at least one embodiment described above.

In particular, the wind energy system has a control device and the control device is prepared to carry out a method according to an embodiment described above. In particular, the method can be implemented in the control device for this purpose. It is preferably provided that a method for recognizing low-frequency vibrations and / or a wind energy system for recognizing low-frequency vibrations is also prepared to react in the event of one or more detected low-frequency vibrations, in particular to dampen the electrical supply network. For this purpose, it is proposed that when a low-frequency oscillation has been detected, the feeding of electrical power, in particular electrical active power, into the electrical supply network is reduced, in particular by 30% -70%, preferably 50%. It was recognized that by reducing the active power fed in, the electrical supply network can be calmed with regard to low-frequency vibrations. It should be emphasized here in particular that no detailed information about the detected low-frequency vibration is required. It is sufficient to carry out the proposed steaming measure, ie to trigger it as soon as a low-frequency oscillation has been detected.

According to a further variant, in the case of one or more detected low-frequency vibrations, it is provided that a damping measure is activated in which, for example, a modulated power is fed in.

It is preferably proposed, both for the method and therefore also for the wind energy installation, that the decay of the one or more detected low-frequency vibrations is assumed when the at least one gradient signal has fallen below a predetermined termination limit, which is in each case less than the test limit. In particular, it is proposed that the termination limit is in each case at least less than 80% of the test limit, in particular in each case less than 50% of the test limit.

In this respect, it is proposed that the detection of the presence of a low-frequency oscillation is assumed if a predetermined test limit is exceeded, but the decay of the low-frequency oscillation is not yet assumed if this predefined test limit is undershot again. Rather, the gradient signal must be significantly smaller than the specified test limit and the specified termination limit is proposed for this, which is selected to be significantly smaller than the test limit. In particular, it should have a maximum of 80%, preferably a maximum of 50% of the test limit.

When checking a plurality of gradient signals, that is to say when several test signals have been recorded, an individual test limit can be provided for each gradient signal his. Correspondingly, in order to recognize the decay of recognized low-frequency vibrations, it is then to be assumed that there are several termination limits which are correspondingly individual. A test limit and a termination limit are preferably assigned to each gradient signal, the respective termination variable being smaller than the test limit of the same gradient signal.

It is preferably proposed that damping measures that were triggered upon detection of a low-frequency oscillation be ended when the low-frequency oscillations have decayed. The damping measures can therefore be initiated when a gradient signal exceeds the predetermined test limit and they can be terminated when the same gradient signal falls below its termination limit.

The invention is explained in more detail below by way of example using embodiments with reference to the accompanying figures.

Figure 1 shows a wind turbine in a perspective view. Figure 2 shows a wind farm in a schematic representation.

FIG. 3 schematically shows a sequence structure of a method according to one embodiment.

Figure 4 shows a schematic diagram of several test signals

FIG. 5 shows a wind energy system with a control device. FIG. 1 shows a wind energy installation 100 with a tower 102 and a nacelle 104. A rotor 106 with three rotor blades 108 and a spinner 110 is arranged on the nacelle 104. During operation, the rotor 106 is set into a rotary movement by the wind and thereby drives a generator in the nacelle 104.

Figure 2 shows a wind farm 1 12 with three wind turbines 100 as an example, which can be the same or different. The three wind energy plants 100 are therefore representative of basically any number of wind energy plants of a wind farm 1 12. The wind energy plants 100 provide their output, namely in particular the electricity generated, via an electrical parking network 1 14. The respectively generated are Currents or powers of the individual wind energy plants 100 are added up and mostly a transformer 116 is provided, which transforms up the voltage in the park in order to then feed into the supply network 120 at the feed-in point 118, which is also generally referred to as PCC. FIG. 2 is only a simplified illustration of a wind farm 1 12, which shows no control, for example, although of course there is a control. The parking network 114 can also be designed differently, for example, in that, for example, there is also a transformer at the output of each wind energy installation 100, to name just one other exemplary embodiment.

FIG. 3 shows in process structure 300 method steps for the method for recognizing low-frequency vibrations. Accordingly, a voltage detection block 302 and a current detection block 304 are initially provided. The voltage detection block 302 receives the three phase voltages Ui, U2 and U3 and forwards a common voltage signal U to a filter block 306. The three phase voltages Ui, U2 and U3 can be recorded especially as a line voltage at a line connection point.

The current detection block 304 receives the three phase currents h, I2 and I3 and forwards a common current signal I to the filter block 306. The three phase currents h, I2 and I3 can in particular have been recorded as feed currents which a wind energy system has generated and feeds into the electrical supply network at the same grid connection point, to which the three phase voltages Ui, U2 and U3 were also recorded.

In the filter block 306, a filtering of the common voltage signal U obtained and the common current signal I obtained is then carried out first. This filtering is tailored to the frequency spectrum that is of interest. In particular, the filter is designed so that low-frequency vibrations can be maintained as far as possible and are not filtered out.

In addition, in the filter block 306, the signals thus filtered are converted into a rms voltage value Um, an effective power rms value P _m and a reactive power rms value Q _m . All of these three values are output as signals, that is to say as a voltage signal, active power signal and reactive power signal, each signal representing the effective value of the relevant variable as a function of time. These signals output by filter block 306 can form test signals. These three RMS signals are input to deriving block 308. In the derivation block 308, gradients are determined for the RMS values by derivation or difference formation, and these gradients are each compared with a test limit. In this embodiment, it is assumed that there is a low-frequency oscillation if the voltage rms signal Um has been detected to exceed its test limit and, in addition, to at least one of the two other rms value signals, namely the active power rms value P _m and the reactive power rms value Q _m , it has been detected that its test limit has been exceeded , Only then is a low-frequency oscillation assumed. Of course, it can also be considered that all three signals that are entered in the deriving block 308 each exceed their test limit.

If this test criterion is thus met, the deriving block 308 outputs a corresponding signal, which is referred to here as a trigger signal. The signal is called a trigger signal because it can still be used to trigger reactions. Such triggering reactions can be to take damping measures and, additionally or alternatively, can be a safety shutdown of the wind energy system using this method. It is also possible for the trigger signal to always be output, but depending on the situation detected, that is to say depending on whether a low-frequency oscillation has been detected, has a different value or has a different signal amplitude.

FIG. 4 schematically shows the course of three test signals, namely the voltage test signal Um, the active power test signal P _m and the reactive power test signal Q _m , for measurements taken over a period of about 30 seconds. These three test signals correspond to the three RMS signals Um, P _m and Q _m according to FIG. 3, which the filter block 306 outputs there.

FIG. 4 also shows in the diagram a trigger signal that corresponds to the trigger signal T _hg according to FIG. 3 that the deriving block 308 outputs there.

The three test signals Um, Pm and Q _m are shown standardized, namely standardized to nominal values. The numbers are shown as "milli", so that the scale ranges from -1000 to +1000 instead of -1 to +1. According to the proposed method, derivations are also formed from these three test signals for further evaluation, especially in the derivation block 308, before a further evaluation takes place. These derivatives are not shown here for the sake of simplicity. It can be seen in FIG. 4 that all three test signals initially have few vibrations. The voltage test signal Um and the reactive power test signal Q _m initially initially have a constant value. So constant reactive power is fed in. The voltage test signal Um drops slightly, the drop being less than 1%.

The active power test signal shows a slightly increasing value. This increase can also be attributed to increasing wind speeds. However, the increase of around 3% in 15 seconds is comparatively small, and in any case does not allow conclusions to be drawn about low-frequency vibrations. Shortly before time h it can be seen that all three test signals have increasing oscillations. In the graph of the schematic representation according to FIG. 4, the increase in the oscillations appears to be obvious and easily recognizable. However, this relationship cannot easily be identified for automatic evaluation by means of a process computer. It is therefore proposed to derive these three test signals, namely the voltage test signal Um, the active power test signal P _m and the reactive power test signal Q _m . With such a derivation, which is however not shown in FIG. 4, the vibrations then emerge more intensely. The derivatives then become so large at time h that they exceed their respective test limits and were therefore recognized for the presence of a low-frequency oscillation.

This is based on an evaluation that recognizes the presence of a low-frequency oscillation when the voltage test signal and the reactive power test signal each exceed their test limit and / or when the voltage test signal and the active power test signal each exceed their test limit. In the example in FIG. 4, both criteria are met at time t1. For the sake of simplicity, the trigger signal T g shows at the time ti that the trigger signal T rig jumps from 0 to the value 1. If only one of the criteria is met, the trigger signal Thg assumes a smaller value, but which is significantly greater than zero, e.g. 0.8.

The trigger signal Thg assumes the value 0 only if none of the criteria is met. For this reason, the trigger signal Thg partially drops to this smaller value of approximately 0.8, because there the active power test signal or the reactive power test signal in the time ranges dropped below their test limit. The voltage test signal did not drop below its test limit during the entire time shown from time h, because in that case the trigger signal T rig would have dropped to the value 0.

If the trigger signal T rig does not assume the value 0, a damping measure is initiated, or even a wind energy system is shut down, or even the wind energy system is disconnected from the electrical supply network.

FIG. 5 shows an illustration of a wind energy installation 500 with a control device 502, which, like the inverter 504 just shown, is to be regarded as part of the wind energy installation 500 and could, for example, be arranged in the tower 506 of the wind energy installation, with the inverter 504 and the only for the sake of clarity Control device is shown outside the remaining wind turbine 500.

The inverter 504 receives power generated from the wind as a DC voltage signal and executes the inverter based thereon and generates a three-phase feed current li, _2,3 at a three-phase voltage U-1,2,3. This can be fed via a transformer 508 into the electrical supply network 510 at a network connection point 512 indicated there.

To carry out a proposed method for recognizing low-frequency vibrations, current and voltage can first be measured with an indicated measurement sensor 514 and transferred to the detection device 526. The detection device 516 and the measurement sensor 514 can also form a common unit.

The detection device 516 thus detects at least one test signal from the transferred measurements. Here voltage and current can each form a test signal. This test signal or here these test signals are then passed to the filter unit 518, which carries out filtering and in particular carries out this filtering in such a way that essentially only signal components with desired frequencies remain, namely in the range of the expected low-frequency vibrations. These signals filtered in this way are used as test signals and transferred to the derivation unit 520. The symbol of the derivation unit 520 indicates a time-continuous derivation, but of course, especially when discrete signals are present, a discrete derivation by difference formation can also be considered. In any case, the signal or signals derived in this way is transferred to the recognition unit 522, which then checks a predetermined test criterion, in particular for each received derived test signal, checks whether a predetermined test limit has been exceeded in each case. As a result, the recognition unit 522 can transmit a trigger signal to a process computer 524.

The process computer 524 basically controls the inverter, possibly takes on further control tasks, and can also make this control dependent on the trigger signal received by the recognition unit 522. Particularly when one low-frequency oscillation or several low-frequency oscillations have been detected, the process computer 524 can control the inverter 504 accordingly in a modified manner, for example by specifying a reduction in the power to be fed in. For this purpose, which is not shown in FIG. 5, the process computer 524 can also carry out further controls in the wind energy installation, such as, for example, adjusting the rotor blades in order to also draw correspondingly less power from the wind. In addition or alternatively, it is possible that in the event of detected low-frequency vibrations to protect the wind energy installation, the feeding is stopped and, if necessary, a safety switch is opened, which, however, is not shown in FIG.

In particular, a method for detecting low-frequency vibrations has been proposed. This takes into account the fact that energy systems are vibratory systems that have natural modes below and possibly also above the system frequency. The system frequency is to be assumed here as the system frequency, usually 50 Hz or 60 Hz. When excited, such vibrations can impair the system stability.

Wind turbines can also help stabilize the energy system or, if handled incorrectly, even destabilize it. It should be noted that the lifespan of a wind turbine can be around 25 years and during this time the energy system can also change and develop significantly.

The now proposed observation of low-frequency vibrations, which are also referred to as energy system vibrations or Power System Oscillations (PSO), can not only be used as a warning system for the operation of wind turbines or wind farms. are used, but this information as a result of the observation can also be used to use suitable damping signals from the wind farm or, if applicable, from a wind power plant, to dampen the energy system vibrations.

The proposed method can also be implemented as an algorithm in a control device, in particular a process computer. In the case of a wind farm as a wind energy system, a central parking computer or a central parking control unit on which the method can be implemented can also be considered. In particular, the detection device, filter unit, derivation unit and detection unit, as illustrated in FIG. 5, can also be combined in a common control unit or implemented as algorithms or software blocks.

The proposed algorithm or the proposed method is based in particular on the analysis of voltage and power gradients. Incidentally, evaluations can be carried out on site at the wind energy installation or in the wind farm, or also in a remote monitoring center. It is then possible for the necessary data to be transferred via SCADA.

Otherwise, the proposed method can also be used for consumer units, and in principle also for conventional power plants. For example. consumer units may change their behavior if low-frequency vibrations are detected or may disconnect from the electrical supply network. In particular, a solution is also created that enables a method for online detection of energy system vibrations.

It is particularly provided for the method that an online analysis of transient measurement data takes place at a network connection point of a wind farm. This is particularly advantageous because a central park control unit is usually arranged there in wind farms. The analysis can thus be carried out immediately and on site. It is proposed that voltages and currents are evaluated, suitably filtered, and that the time gradient of the filtered voltage and calculated power signals is finally calculated last. Appropriate parameterization of the test limits, which can also be referred to as threshold values, can then be used to detect low-frequency vibrations.

Claims

Expectations

1 Method for detecting low-frequency vibrations, in particular subsynchronous resonances, in an electrical supply network (510), the electrical supply network (510) having a mains voltage with a nominal network frequency, comprising the steps:

Detecting at least one electrical signal of the electrical supply network (510) as at least one test signal and

Filtering and / or transforming the at least one recorded test signal into at least one test signal,

deriving the at least one test signal over time or forming differences in time-spaced values of the test signal in order to obtain a gradient signal in each case,

Detect the presence of a low-frequency oscillation when the gradient signal or at least one of the gradient signals fulfills a predetermined test criterion, in particular at least one predetermined test limit is exceeded.

2. The method according to claim 1,

characterized in that

a maximum gradient value is specified as a test criterion and the presence of the low-frequency oscillation is recognized if the gradient signal exceeds the maximum gradient value at least once, in particular if

the gradient signal is above the gradient maximum value for at least a predetermined minimum period.

3 The method according to claim 1 or 2,

characterized in that

a difference between the maximum and minimum value of the test signal in a considered test period is used as the gradient signal.

4 Method according to one of the preceding claims,

characterized in that

a line voltage of the electrical supply network, in particular three-phase, is detected as a first test signal and A feed current, in particular three-phase, fed into the electrical supply network (510) is recorded as a second test signal, and in particular

the first and second test signals are transformed into at least one test signal.

5. The method according to claim 4,

characterized in that

the first and second test signal into a voltage signal, a real power signal and a reactive power signal as Spannungsprüfsignal (Um), Wirkleistungsprüfsignal (P _m) and Blindleistungsprüfsignal be transformed, and the Spannungsprüfsignal (Um), the Wirkleistungsprüfsignal (P _m) and the Blindleistungsprüfsignal (Q _m ) are derived in time, or a difference is formed in order to obtain a gradient signal, namely a voltage gradient signal, an active power gradient signal and a reactive power gradient signal, wherein

the voltage gradient signal, the active power gradient signal and the reactive power gradient signal are checked for the presence of a low-frequency oscillation, in particular in such a way that

the presence of a low-frequency oscillation is assumed if a low-frequency oscillation was detected at least in the voltage gradient signal and the active power gradient signal or in the voltage gradient signal and the reactive power gradient signal.

6. The method according to any one of the preceding claims,

characterized in that

a network frequency of the electrical supply network is detected as a further test signal,

the further test signal is transformed into a frequency signal as a frequency test signal,

the frequency test signal is derived in time, or a difference is formed in order to obtain a frequency gradient signal,

the frequency gradient signal and in particular at least one further gradient signal are checked for the presence of a low-frequency oscillation, in particular in such a way that the presence of a low frequency oscillation is assumed to have in the frequency gradient signal and in at least one of the gradient signals of the list

a voltage gradient signal,

- an active power gradient signal and

a reactive power gradient signal

a low frequency vibration was detected.

7. The method according to any one of the preceding claims,

characterized in that

for detecting the at least one electrical signal of the electrical supply network as a test signal

a three-phase mains voltage is recorded and a constant, in particular a space vector size of the voltage is formed from this via a transformation, in particular that a co-system voltage is determined according to the method of symmetrical components, which forms a and / or a three-phase feed-in current is recorded and from this via a transformation an equal size, in particular a space vector size of the voltage is formed, in particular that a positive sequence system current is determined according to the method of the symmetrical components, which forms a test signal, preferably

the same size or space vector size of the voltage or the system voltage and the size or room vector size of the current or the system current are transformed into a voltage signal, an active power signal and a reactive power signal as a voltage test signal (Um), active power test signal (P _m ) or reactive power test signal (Q _m ) ,

8. The method according to any one of the preceding claims,

characterized in that

it is assumed that the one or more detected low-frequency vibrations have decayed if the at least one gradient signal has fallen below a predetermined termination limit, which is in each case smaller than the test limit.

9. The method according to any one of the preceding claims,

characterized in that is fed into the electrical supply network by means of a wind energy installation or a wind farm and

the detection of low-frequency vibrations is carried out by means of the wind energy installation or by means of the wind farm.

10. The method according to any one of the preceding claims,

characterized in that

to a vibration caused in the electrical supply network is detected if

In the voltage gradient signal and the active power gradient signal or in the voltage gradient signal and the reactive power gradient signal, a low-frequency oscillation was recognized in each case, and these detected low-frequency oscillations have the same oscillation frequency, in particular

an additional check is carried out to determine whether the network frequency is oscillating at the same oscillation frequency.

1 1. The method according to claim 10,

characterized in that

as the voltage gradient signal, active power gradient signal and reactive power gradient signal, a difference between the maximum and minimum value of the corresponding test signal is used in a test period under consideration and

an identical oscillation frequency is recognized by the fact that

the same test period is used and / or

respective time intervals between the maximum and minimum value of the corresponding test signal for the voltage gradient signal that

Active power gradient signal or the reactive power gradient signal are the same or similar.

12. The method according to any one of the preceding claims, characterized in that a gradient quotient is formed as the quotient of two gradient signals and

depending on the gradient quotient to an oscillation caused in the electrical supply network is detected, in particular as a gradient quotient a voltage-active power quotient (dU / dP) is formed as a quotient between the voltage gradient signal and the active power gradient signal and / or

a voltage-reactive power quotient (dU / dQ) is formed as the quotient between the voltage gradient signal and the reactive power gradient signal, and

to a vibration caused in the electrical supply network is detected if

the voltage-active power quotient (dU / dP) and / or - the voltage-reactive power quotient (dU / dQ) are negative.

13. The method according to any one of the preceding claims,

characterized in that

Vibrations are classified, and

- A vibration classification found is output, in particular

a vibration classification is selected from the classification list comprising:

a low-frequency oscillation was recognized for one or the mains voltage signal and one or the active power signal, a low-frequency oscillation was recognized for the mains voltage signal and one or the reactive power signal,

a low-frequency oscillation was recognized for the voltage signal, the active power signal and the reactive power signal, and - it was found for the mains frequency and at least one of the signals from the list

the voltage signal,

the active power signal and

the reactive power signal,

a low frequency vibration is detected.

14. A wind energy system, namely a wind energy installation (100, 500) or a wind farm with a plurality of wind energy installations, for detecting low-frequency vibrations, in particular subsynchronous resonances, in an electrical supply network (510), the electrical supply network (510) having a mains voltage with a nominal network frequency and the wind energy system includes: a detection device (516) for detecting at least one electrical signal of the electrical supply network as at least one test signal,

a filter unit for filtering and / or transforming the at least one recorded test signal into at least one test signal,

a derivation unit (520) for deriving the at least one test signal over time or for forming differences in time-spaced values of the test signal in order to obtain a gradient signal in each case,

a detection unit (522) for detecting the presence of a low-frequency oscillation when the gradient signal or at least one of the gradient signals fulfills a predetermined test criterion, in particular at least one predetermined test limit is exceeded.

15. Wind energy system according to claim 14,

characterized in that

- The wind energy system has a control device (502) and

the control device (502) is prepared to carry out a method according to one of claims 1 to 13.