Method and wind park for feeding electrical power into a power supply grid

Abstract

The invention relates to a method for feeding electrical power into a power supply grid by means of at least one wind farm having a plurality of wind energy installations, wherein the wind farm feeds electrical power into the power supply grid at a grid connection point and an expected power (P) is determined for a predetermined feed-in period. F ), wherein the expected power specifies a value of power or a time profile of power, which is expected to be provided by at least one wind farm as power in the wind in a predetermined feed-in period, in particular as the sum of all rotor powers of all wind energy installations of the wind farm and / or as feed-in power of the wind farm, and wherein an expected accuracy is determined for the expected power, the expected accuracy being a measure representing the accuracy with which the power reaches the expected power in the feed-in period, and / or wherein in order to determine the expected power (P F ), using a meteorological forecast to determine at least one wind forecast value (P MO ), additionally determining or verifying the wind forecast variable from the weather forecast by means of correction rules in relation to local weather data and / or operating data of the wind farm, wherein the expected power (P F ) based on the wind expected variable (P MO ) to determine.

Description

Technical Field

[0001] The invention relates to a method for feeding electrical power into an electrical supply grid by means of at least one wind park having a plurality of wind energy installations. The invention also relates to a corresponding wind park. Background Art

[0002] Methods for feeding electrical power into a power supply grid by means of wind farms combining a plurality of wind energy installations are generally known. In addition to the pure power feed for supplying consumers, such wind farms are also increasingly faced with tasks in the area of ​​grid system services and tasks that include support tasks up to grid reconstruction or even up to the preparation or execution of so-called black starts. These tasks are difficult for wind farms to meet in some cases, especially because their primary energy, namely the wind, can not only fluctuate strongly but can also be difficult to predict.

[0003] It is particularly problematic here that conventional weather forecasts, although relatively accurate and readily available, are only available for geographical areas that are relatively large compared to the relatively small geographical area of ​​the wind farm. In large wind farms, there may be the problem, in addition or as an alternative, that the wind farm is not uniform over the geographical area of ​​the wind farm, in particular with regard to its wind intensity. This means that at one location in the wind farm there may be different wind speeds than at other locations in the wind farm, and the situation may change constantly.

[0004] On the other hand, local weather forecasts can be produced with local resolution for a wind park or even regionally for a wind park, which however have a low accuracy for this purpose.

[0005] Therefore, the basic problem remains that the predictability of the wind speed and thus of the power of the wind farm is very imprecise. This makes it difficult to plan the possible power feed-in. This can be problematic in particular for grid reconstruction or even black starts in cases where a power profile is required that is to be fed in but which cannot be realized or can only be realized inadequately due to wind disturbances.

[0006] The German Patent and Trademark Office searched the following prior arts in the priority application of the present application: DE10 2017 129299A1, US 9 690 884 B2, US2014 / 0 195 159A1, EP 3 432 091A1 and CA 2 770 637A1. Summary of the invention

[0007] The present invention is therefore based on the object of solving at least one of the above-mentioned problems. In particular, the plannability of the available power for a wind farm is to be improved. At least an alternative solution to the previously known solutions is to be proposed.

[0008] According to the invention, a method according to an embodiment of the invention is proposed. Therefore, a method for feeding electrical power into a power supply grid by means of at least one wind farm is taken as a starting point. Such a wind farm has a plurality of wind energy installations, at least one of which feeds into the power supply grid via a common grid connection point. Therefore, the wind farm feeds electrical power into the power supply grid at the grid connection point.

[0009] Furthermore, it is proposed to determine the expected power in a predetermined feed-in period. The expected power specifies the value of the power or the time profile of the power which is to be expected to be available to at least one wind farm as wind-derived power in a predetermined feed-in period. This is firstly to be understood generally as meaning that the wind power is available and can also be received and converted by the wind farm at least in theory. This can firstly ignore technical limitations.

[0010] In particular, the expected power can be understood as the sum of all rotor powers of all wind energy installations of a wind farm. At each wind energy installation there is wind which acts on the rotor of the wind energy installation and to this extent is available to the rotor as power, which is called rotor power. The expected power of a wind farm is then the sum of all of these rotor powers of all wind energy installations of the wind farm or of all observed wind energy installations of the wind farm.

[0011] However, the expected power can also be understood as the power that can be fed into the wind farm. This power that can be fed into the wind farm can differ from the sum of all rotor powers in that the power that can be fed in, which can also be called the feed-in power, can take into account power losses and power limitations and is therefore usually smaller than the sum of all rotor powers.

[0012] It is now additionally proposed that an expected accuracy is determined for the expected power, which is a measure of the accuracy with which the power reaches the expected power during the feeding time period. Therefore, the expected accuracy is a measure of the quality of the power estimation in this regard. It has been particularly recognized here that it is not only important to determine the expected power as accurately as possible, but it is also important for good plannability to be able to estimate to what extent the power actually deviates from the expected power later if the estimated feeding time period is reached. There are different suggestions for giving the expected accuracy, which are further explained in detail below. The possibility of quantifying the expected accuracy gives a P90 value, which describes the wind farm power that occurs with a probability of at least 90%. Therefore, the power quantified thereby or a higher power occurs with a probability of 90%. The P90 value can therefore form the expected accuracy according to a variant.

[0013] In addition or alternatively, it is proposed that, in order to determine the expected power, at least one wind expected variable (P MO ), and additionally determining or verifying a wind forecast variable from a meteorological forecast by means of a correction rule in relation to local meteorological data and / or operating data of the wind farm, wherein based on the wind forecast variable (P MO ) to determine the expected power (P F ).

[0014] Weather forecasts, i.e. weather forecasts created in particular for a geographical area that is larger than the space occupied by a wind farm, can have deviations or errors at the wind farm. It is known that such deviations or errors can at least be reduced when local meteorological data and / or operating data of the wind farm are used for improvement. Local meteorological data and operating data of the wind farm can also be referred to as local wind farm data. It can include meteorological data in the wind farm or its surroundings, such as wind speed, wind direction, temperature, precipitation and air pressure, but also specific operating data, such as generated or output power. Local meteorological data or some of them can also be detected in the wind farm, such as, for example, wind speed and wind direction. The wind direction can also be derived from the azimuth position of the wind energy facility.

[0015] The local wind farm data can then be used to correct the weather forecast in order to obtain a better basis for determining the wind prediction variable. The wind prediction variable is thus determined as an estimate and can be obtained with a high degree of accuracy by means of a correction rule.

[0016] The wind prediction variable itself can be the wind speed or also a curve of the wind speed, which therefore describes the wind speed to be expected as closely as possible. However, it can also be another representative variable, such as, for example, a normalized value, which can be normalized to the rated wind speed, for example. However, it is also conceivable that it is already possible to base on the following power, such as the rotor power, which can be generated with the wind speed to be expected.

[0017] It is also conceivable that the wind prediction variable is verified, in particular, based on the operating data of the wind farm. In particular, a variant is proposed in which the wind prediction variable itself is power and is compared with the actual power. One possibility is that the wind prediction variable is a power change curve, which has been created in a predetermined estimation time period. The starting point of the estimation time period can correspond to the current moment and the remaining estimation time periods are therefore located in the future. Then the wind prediction variable at the starting point of the estimation time period, that is, its power value, can be compared with the current power value of the wind farm, similar power. If the two power values ​​differ by a deviation factor, for example, the estimated power value is 20% lower than the current power value of the wind farm, then the estimation can be corrected by 20%, that is, the deviation factor. This means that the wind prediction variable is corrected by the deviation factor throughout the entire estimation time period. The deviation factor is then a correction rule.

[0018] This principle can also be applied analogously to the wind expected value in the case where the wind expected value is an estimated wind speed and is compared with the measured wind speed, referred to as another example. This can also be applied to other cases, such as, for example, wind direction, where the wind direction can be considered in particular as a boundary condition for the wind speed. However, in particular for considering electric field effects, the wind direction can be an important variable. In this regard, the wind expected variable can include multiple variables, namely, according to one embodiment, wind speed and wind direction.

[0019] Finally, the expected power is determined based on the wind expected variable. The wind expected variable is still essentially independent of specific data of the wind farm, such as technical availability or shadow effects in the farm, and is essentially an estimate of the wind, whereas the expected power specifically relates to the power that the farm can subsequently feed or generate. In one embodiment, the expected power may correspond to the wind expected variable.

[0020] According to one embodiment, it is proposed that the weather forecast is created in a forecast period, and the wind expected variable and / or the expected power are created as a short-term forecast in the forecast period, wherein the forecast period is at least ten times larger, preferably at least fifty times larger, in particular at least one hundred times larger than the forecast period. This means, in particular, that the weather forecast also extends further into the future than the short-term forecast. In particular, it has been recognized that a short forecast period is sufficient for power grid reconstruction or black start, and thus a short, precise forecast is better for this than a long, imprecise forecast.

[0021] In particular, the purpose here is technical, not economic. The determination of the wind forecast variable, i.e., in particular the short-term forecast, is not used to calculate the electricity tariff or to control the operation of the wind farm according to the electricity tariff, but the short-term forecast enables the wind farm to also undertake schedulable support tasks for the power grid. The forecast period of the short-term forecast is in particular in the range of at most one hour, in particular in the range of at most half an hour, whereas the forecast period of the weather forecast is in the range of more than a few hours, in particular more than half a day, or even more than a few days.

[0022] Preferably, a short-term forecast in the range of 10 minutes to 30 minutes, in particular about 15 minutes, can be proposed. The forecast period can in particular include 3 calendar days, i.e. the current day, the next day and the day after that, and thus according to an example is not in the sense of 72 hours. Preferably, it is added that an update is performed every 6 hours, for example by a weather service that provides weather forecasts. Depending on how long it has not been updated, the weather forecast, which can also be used as an estimate stored in a memory, can therefore be used to estimate the next 48 to 72 hours according to the daytime. This forms the basis for a short-term forecast for this improvement. However, it is also considered in principle to consider the entire day instead of the calendar day, so that there is a weather forecast for 72 hours. However, other forecast time periods can also be considered in principle, but the other forecast time periods are usually at least one full day for weather forecasts.

[0023] The estimation period is therefore significantly shorter than the forecast period.The wind expected variable and the expected power usually have the same temporal abscissa and thus the same estimation period, so that both are then preferably provided as short-term estimations.

[0024] In addition or alternatively, it is proposed that the wind prediction variable is predetermined repeatedly or continuously in a comparison time period. The comparison time period can correspond to the feed time period in principle, but is preferably a shorter time period of its own, which is also correspondingly repeatedly redefined or can also be a sliding comparison time period in the case of continuous predetermination.

[0025] In this regard, it is proposed that a forecast comparison be performed in each comparison time period in which a wind forecast variable has been predetermined. In the forecast comparison, the forecast variable is compared with the current wind variable, which represents the current wind speed of the corresponding comparison time period. The forecast variable can be, for example, a wind forecast variable, but the wind forecast variable can also be, for example, the expected power. It is also conceivable that this is, in particular in the case of a wind forecast variable, the wind speed or the curve of the wind speed. This is then compared with the current wind variable, which represents the current wind speed of the corresponding comparison time period.

[0026] In particular, of course, variables of the same unit are also compared here, i.e., for example, two powers or two wind speeds, or two nominal variables. Correspondingly, the current wind variable can also be the wind speed or, for example, also the power, i.e., the power represents the current wind speed at the instant. For example, it is considered that the power of the electric field actually generated later is used. If necessary, this can also be changed by computer in the following situations: for example, if a wind energy facility in the electric field fails, the central control unit can know this.

[0027] If in this case the wind farm reaches its rated power and is therefore subsequently limited, it is possible, for example, to calculate a greater available power from other essential data, such as the rotor blade position.

[0028] In this regard, a comparison is therefore performed between the estimated value and the actual value, which comparison can be referred to as a prediction comparison. From the prediction comparison, at least one matching rule is determined in order to improve the expected power with the aid of the at least one matching rule. As an improvement, it is particularly conceivable to improve the calculation rule for calculating the expected power from the wind expected variable. In addition or as an alternative, the correction rule can be improved with the aid of the matching rule.

[0029] In the simplest case, to illustrate this, the adaptation rule can be a factor, for example 1.2 or 0.8, by which the calculation rule, the correction rule and / or the other rule are multiplied, since it is thus found in the forecast comparison that the expected power is 20% lower or 20% higher than the actual power. The adaptation rule can also be advantageous in particular in the comparison time period and can be applied in an ideal case to any feed time period.

[0030] However, it is also conceivable to apply the matching rule directly to the expected wind variable or expected power. However, it is also conceivable to match it in addition to the factorial matching, which will also be explained later. In particular, it is also conceivable to correct the time offset with the help of the matching rule. This can also be carried out directly or via calculation rules, correction rules and / or other rules.

[0031] In particular, the idea is based here on the fact that weather forecasts are very accurate, but not completely accurate with regard to location information. Similar to clouds, however, the wind speed also continues to advance and it has been recognized that local inaccuracies are usually inaccuracies in time, i.e. the predicted wind speed or also the predicted wind direction will only occur slightly later than predicted at the wind farm, or slightly earlier. It is thus possible, for example, that the wind forecast predicts a specific wind profile in a specific time period and that the wind profile will therefore also actually occur similarly at the wind farm, but slightly later or slightly earlier. This can also be recognized in the forecast comparison and the matching rules can be determined accordingly.

[0032] In any case, the expected power is finally determined based on the wind expected variable and the matching rule, wherein the latter directly or indirectly has the influence as already described above.

[0033] Preferably, the expected power is determined using a power estimator. In this case, it is particularly provided that the expected power is determined from the estimated wind speed or the estimated wind speed profile via the power estimator. Such a power estimator can be designed as a state observer, which simulates a wind energy installation or a wind farm and how it receives the wind speed together with the wind direction as input variables and then models how the wind energy installation or wind farm behaves and accordingly also outputs the power as a system variable.

[0034] In addition or alternatively, the expected power can be additionally determined taking into account boundary conditions. In this case, one or more boundary conditions explained below can be taken into account. In particular, the technical availability of the wind energy installations of the wind farm can be taken into account. In the simplest case, this means taking into account which wind energy installations of the wind farm are in operation or whether one or more wind energy installations are faulty because they must be temporarily maintained, for example.

[0035] However, it is also conceivable that one or more wind energy installations in a wind farm are operated at reduced power. It is also conceivable that a power limit is reached. This power limit can be a rated power, but it can also be a manually preset limit, such as a limit preset for noise protection reasons, just as an example. This limit can then only relate to one or individual wind energy installations in a wind farm. Other limit powers can also be preset by the grid operator of the power grid fed by the wind farm.

[0036] Information about available controllable loads can also be a boundary condition. In particular, it is taken into account that such loads, such as a cold storage operated by a wind farm, may at least temporarily withdraw from their power absorption in the event of a power bottleneck. This may potentially activate power. This is particularly meaningful when it is to be avoided that the minimum power is exceeded.

[0037] As a further boundary condition, information about the storage available for feeding in electrical power is taken into account. Such available storage can provide power in a specific time period and thus overcome excessive power dips. This is also suitable in particular for avoiding excessive drops in the fed-in power.

[0038] In particular, the boundary conditions mentioned above can also be taken into account for determining the expected accuracy. Thus, a small storage capacity or also a small power absorption of the controllable consumers can already increase the guaranteed minimum power and thus the corresponding prediction of the minimum power. The problem of falling below the minimum power is therefore usually that this occurs only briefly and possibly only rarely due to rare ideal and permanent wind conditions. In order to improve the overall power balance of the wind farm, in particular to avoid rare dips, such a small power or a small amount of energy in the storage may sometimes be sufficient, because the rest of the time this support can usually be omitted.

[0039] Another boundary condition is information about the follow-up effect in a wind farm. In particular, it has been recognized that not only the power characteristic curve of the individual wind energy installations together with the wind forecast allow a high-precision power estimate. Rather, it has been recognized that the power of a wind energy installation in a wind farm group can be less than if the wind energy installations were operated as individual wind energy installations. For this purpose, it is usually assumed that the wind in a wind farm is attenuated relative to the open space.

[0040] However, it is also conceivable that a specific follow-up effect influences the power of individual wind energy installations. If a wind energy installation is located behind another wind energy installation, i.e. behind the other wind energy installation with respect to the prevailing wind direction, the wind energy installation located behind will receive significantly less wind and thus generate correspondingly less power. However, this is very strongly dependent on the wind direction and even a small change in the wind direction can cause the rear wind energy installation, which was previously behind the front wind energy installation, to no longer be relevant to the follow-up effect. This can be considered as a boundary condition.

[0041] Alternatively or additionally, it is proposed that the expected power is additionally determined under the condition of individual consideration of the wind energy facilities of the wind farm. In particular, it is proposed that the technical performance of each wind energy facility is saved in a table for the wind farm. In addition, it is proposed that the table saves the predicted technical performance for each wind energy facility in relation to its azimuth direction or in relation to the wind direction and optionally in relation to the time of day. Additionally or alternatively, the performance can be stored according to the wind speed. For this, a multidimensional table can be provided. In addition, together with the predicted wind speed, the expected power can be determined or improved. Preferably, different variants can be proposed for the expected accuracy, according to which the expected accuracy then contains or forms the information explained below. The expected accuracy can also contain a plurality or all of the information explained below.

[0042] A preferred variant here is that the expected accuracy specifies a first power limit, which specifies a power or a power profile which is not undershot with a predetermined probability of achievement in a predetermined feed-in period. Thus, for example, a constant power value can be specified, or a wind profile over time, i.e. over a feed-in period, which is not undershot with a probability of 95% in a predetermined feed-in period, i.e. not only once.

[0043] With the aid of this description, very good planning can then be carried out. If the probability of, for example, 95% mentioned is tolerable, the probability here therefore represents a predetermined probability of realization, i.e., if the remaining 5% possibility can be tolerated, this is particularly the case. It should also be taken into account that generally a plurality of wind farms can be operated in the same manner. As long as a short-term drop below the expected power predicted in a wind farm is only expected by 5%, it is not necessarily the case that the remaining wind farms operating in a similar manner will have the power drop at exactly the same instant, which is only expected by, for example, the 5% mentioned. In other words, the predetermined probability of realization does not necessarily have to be 1, i.e., 100%.

[0044] According to one embodiment, the expected accuracy can additionally or alternatively describe a second power limit, which describes the following power or power variation curve, which is not allowed to be lowered in the medium-term average value. Therefore, it is allowed to be lower than the power variation curve for a very short period of time, but it is not allowed to be lower than the power variation curve for a long period of time. This medium-term average value is in particular an average value determined in a time period of 10-60 seconds. Therefore, if the power drops sharply below the second power limit in 5 seconds, but is significantly higher than it before and after, then the medium-term average value is not lower than the second power limit as a whole. It is also meaningful here that, especially for the grid operator, the single power of the wind farm is less important than the total power of multiple wind farms fed into its grid. Therefore, as long as the second power limit is not lowered in the medium-term average value, it can be assumed that the short-term lowering is hardly noticed by the power supply network, because various other wind farms that feed in - if they exist after all - do not have a power drop at the same moment.

[0045] Likewise, according to another embodiment, it is additionally or alternatively proposed that the expected accuracy specifies a third power limit, which specifies a power or a power profile that may not be dropped below in the short-term average value. Such a third power limit is usually lower than the second power limit here, because the short-term average value fluctuates more strongly than the medium-term average value and can therefore also have lower values. As a short-term average value, an average value in a time period of 5-10 seconds is particularly assumed here. The total power is also important for the grid operator here, which is fed by the multiple wind farms when the expected accuracy of the multiple wind farms is determined identically and the mode of operation is identical.

[0046] According to another embodiment, it is proposed that weather forecasts or weather data for creating weather forecasts are regularly transmitted by an external weather service to at least one wind farm. The wind farm does not carry out its own measurements for the weather forecasts, but in particular also uses generally available weather forecasts. Such weather forecasts can also be prepared regularly so that the process computers in the wind farm can process them.

[0047] In this regard, it is further proposed that the wind farm at least temporarily stores the respectively transmitted weather forecast. This is foreseen for a possible interruption of the external weather service. In the event of an interruption to the external weather service, the expected power is then estimated based on at least one stored weather forecast. Intuitively, there is therefore an older weather forecast than would have been available at that time if there had been no interruption to the external weather service.

[0048] For this case, it is then proposed in particular to adjust the stored weather forecast taking into account the local current weather measurement values, in particular to improve the weather forecast in relation to the local wind farm data by means of correction rules. This is based in particular on the idea that, for example, the weather forecast had predicted an increase in wind speed from 4 m / s at 16:00 to 8 m / s at 17:00. However, the local current weather measurement values ​​did not detect an increase in wind speed from 4 m / s until, for example, 16:20. It can then be assumed together with the stored weather forecast that the wind speed increased to 8 m / s by 17:20.

[0049] In the simplified example, even older weather forecasts can already predict an increase in wind speed, and local measurements can help to refine the forecast, in the simplified example only with respect to the start of the increase in wind speed. In this regard, a solution has been found to adapt the forecast very accurately to the wind farm. The described principle can also be used in cases where it is not necessary to store weather forecasts. However, in cases where it is necessary to store weather forecasts, a very good forecast quality can always be achieved for a specific wind farm.

[0050] It has also been recognized here that a short-term forecast of the expected wind value can be created with good accuracy based on the long-term weather forecast. The long-term weather forecast can also be obtained from the weather service for a long time in advance. The short-term forecast is particularly provided for black start or grid reconstruction. If this happens, the data connection with the weather service will be interrupted. However, there are still older weather forecasts, which the wind farm has stored earlier. The older weather forecast can then be improved based on local wind farm data. For the local wind farm data or at least part of it, there is no need for a data connection to the weather service, which may be interrupted. However, by using the previously stored weather forecast, at least the short-term forecast can at least still trace back to the older weather forecast and improve the forecast quality through the local wind farm data.

[0051] According to another embodiment, it is proposed to plan a black start and / or grid reconstruction based on the expected power. A black start is a situation in which no external auxiliary starting feeder, here the wind farm, the power supply network or its grid section is required. The grid section is therefore voltage-free and the wind farm must build up a corresponding voltage at a corresponding frequency and in doing so supply power to the first consumer connected to the grid section.

[0052] This can then be followed by a grid reconstruction, in which other grid sections that are de-voltaged or at least disconnected from one another are reconnected and then have a grid voltage with a common grid frequency and can transmit power. However, a grid reconstruction without a preceding black start or at least without a black start by a wind farm can also be planned.

[0053] For this kind of planning, it is very important to know the available power. This knowledge is necessary at least for the following time period, in which the grid operator can still react or the automatic grid regulator can react during the black start and / or grid reconstruction. The reason is that it is completely acceptable for a wind farm to reduce the feed-in power due to the upcoming lack of wind power in a 15-minute time period, as long as the grid operator knows this and can start other wind farms or power plants accordingly. However, the time period after the black start or grid reconstruction should also be considered and planned here, in such a way that the started or rebuilt grid must also be established in a stable operating point. In this regard, the time period in which the available power must be known can be relatively long, for example a quarter of an hour or even half an hour. For this, a short-term estimate is still sufficient in time.

[0054] In addition or alternatively, it is proposed that the grid connection point is connected to a grid section of the power supply network and that the expected power is transmitted to the grid operator operating the grid section as information about the expected power. In this respect, the grid operator can use the power planning of the wind farm. In particular, this is also helpful in the case of a grid construction situation in which at least the grid section fails. The grid operator can then use the power reported to him as available and coordinate the grid construction based on this.

[0055] Optionally, in addition to the transmission of the expected power, the expected accuracy can also be transmitted. This enables the grid operator to plan better, since he also clearly knows to what extent he can trust this. This also applies to the fact that this effect can be brought about in particular by having a plurality of wind farms operate in the same way, in particular so that all of this information can be automated, and the grid operator can therefore combine this information in a process computer.

[0056] The expected accuracy can also be related to different variants as explained above. In addition, the expected accuracy can also be transmitted at least partially in the case of a weather forecast. The expected accuracy can therefore be at least partially information obtained by the wind farm from an external weather service, in particular in a computer-processable form.

[0057] In addition or alternatively, it is proposed to take into account a precision expectation value which indicates with what expected precision the expected power should be provided, wherein the precision expectation value is preferably received externally, in particular from a grid operator. It is therefore proposed to orient the estimation to a predefined precision instead of the estimation created by the precision quantification or in addition thereto.

[0058] It has been recognized that, depending on the situation, in particular the state of the power supply grid, a higher accuracy can be required, which can result in a correspondingly lower guaranteed power. It is also possible to require a lower accuracy, which then has a possibly higher power. By transmitting such an accuracy expectation, the grid operator can influence the power that can be provided in a situation-dependent manner in the estimated time period, in particular, for example, in 15 minutes, by requiring a changed accuracy with respect to the accuracy expectation. At the beginning of grid reconstruction, the required accuracy can be the greatest, but if the security in the system is sufficient, in particular when other wind farms whose inaccuracies can be compensated are sufficiently available, more loads can be supplied by reducing the accuracy expectation.

[0059] According to one embodiment, it is proposed that the weather forecast includes at least one time variation curve of the wind speed to be expected. In addition or alternatively, the weather forecast can output the value of the wind speed to be expected in a predetermined repetition interval. In other words, a continuous or approximately continuous time variation curve of the wind speed to be expected can be used. If the weather forecast outputs the value of the wind speed to be expected in a predetermined repetition interval, this can be located in the range of 1-5 minutes in particular. Therefore, a value is determined and output every 1 minute or every 5 minutes or in a range therebetween, so that an almost continuous variation curve is obtained. The value should be different from the update rate, by means of which the existing weather forecast or the meteorological data therefor is updated, which can be located in the range of 6 hours.

[0060] The expected power is then determined from the weather forecast and in addition or alternatively the minimum value of the expected power is determined as a time power profile. This can be done in particular with the aid of a power estimator. The expected power is thus determined from a continuous or approximately continuous time profile of the wind speed as a profile, i.e. as a profile over time.

[0061] It is then proposed for the power variation curve to be shifted, stretched and / or compressed by means of a matching rule. This relates in particular to the amplitude of the power variation curve, so that it is approximately shifted up or down or compressed. Preferably or additionally, it is also possible to relate to its temporal expansion. The power variation curve can thus be shifted forward or back in time or stretched or compressed.

[0062] Based on the following knowledge, the weather forecast is assumed to be correct and reliable in principle. In particular, it is assumed that the power variation curve calculated from the weather forecast is basically applicable in its basic outline. However, it has been recognized that such a power variation curve may appear at a specific wind farm, for example, slightly earlier or later than the power variation curve obtained from the weather forecast. The reason is particularly that the weather forecast is more global, that is, created for a larger area. Depending on the position of the wind farm exactly in the larger area, for example, exactly in the middle of it, the predicted weather variation curve, in particular the power variation curve derived therefrom, appears slightly earlier or later in the wind farm. This can be compensated by a corresponding movement. It is also possible that the variation curve is slightly delayed and thus stretched when reaching the wind farm, which can be compensated by compression. Similarly, compression is also considered, which can be compensated by stretching.

[0063] Similar content can also be applied to the amplitude, because the amplitude may increase or decrease in a longer path when reaching the wind farm. The same content applies to shorter paths, that is, in particular about the midpoint of a larger area, for which the weather forecast has been created. This amplitude change can also be adjusted. By this, the weather forecast and the power variation curve derived therefrom are improved, without in principle having to inquire about the weather forecast or the power variation curve. The weather forecast can thus be matched to the specific location of the wind farm in a simple manner and method. This or other matching can also be referred to as refitting. Possibly, the wind farm or its closer surroundings also have a specific morphology, which, for example, can slow down the wind, so that in principle a reduced power variation curve is assumed in the wind farm. This can be compensated by controlling the amplitude accordingly. According to one embodiment, it is proposed that in order to determine the correction rule and / or matching rule, the meteorological measurement values ​​of the wind farm recorded in the wind farm and / or near the location of the wind farm. By this, the weather forecast can be improved accordingly, in particular, the weather forecast can be matched to the current location value, or adjusted according to the value.

[0064] It is proposed that as meteorological measurement data, the wind speed, additionally or alternatively the wind direction, additionally or alternatively the temperature, additionally or alternatively the air density and additionally or alternatively the sunlight exposure be recorded. With the aid of these values, in particular a comparison with a meteorological forecast can also be made, and from this, for example, it can be deduced whether the climate change curve predicted at the wind farm is behind the meteorological forecast in terms of time or ahead of it. Correspondingly, a correction rule can be determined. This can in principle be carried out each time the meteorological measurement values ​​are compared, but in particular depends on how important the corresponding variable is.

[0065] However, the matching rule can also be determined as a function of when at least one of the measured values, in particular the wind speed, is compared with the estimate. These values, such as the wind direction, can also be considered as boundary conditions.

[0066] In addition or alternatively, it is proposed that in order to determine the correction rules and / or the matching rules, the measured values ​​and / or operating values ​​of the wind energy installation of the wind farm can be used. In particular, it is proposed to estimate the available power from the measured values ​​and / or operating values. Here, it is particularly based on the following idea that the wind energy installation itself already allows at least a conclusion about the prevailing wind conditions due to its operating performance. In particular, it belongs to this that the wind speed can be derived from the power, the rotation speed and the blade position and the azimuth position and of course the characteristic variables of the wind energy installation. From the azimuth position, which is usually also clearly known in the wind energy installation, the wind direction can be derived accordingly. All of this can also be used for comparison, in particular for predictive comparison, in order to determine the matching rules.

[0067] The correction rule can also be determined with the aid of at least one of the values ​​of the wind park, for example by the estimated available power of the wind park allowing the weather data to be inferred in order to correct the weather forecast therefrom.

[0068] According to another embodiment, it is proposed that the determination of the expected power of the wind farm is carried out on a wind farm computing unit of the wind farm. The calculation is thus performed locally and has the advantage that possible installation values ​​of the wind energy installation or the wind farm can be evaluated together in a simple manner and method without complex transmission being required for this.

[0069] In particular, it is proposed that the wind farm calculation unit is provided with an uninterruptible power supply and that, in the event of a grid failure of the power supply network, the wind farm calculation unit continues to determine the expected power while using the uninterruptible power supply and transmits it to the grid operator or the grid operator. In particular, the idea is based on the proposed method that the wind farm can, even in the event of a grid failure, estimate the current meteorological situation as accurately as possible and thus have information about how much power is available for restarting the wind farm and thus for supporting or rebuilding the power supply network or a part thereof. This information can be transmitted to the grid operator and the grid operator can use it for planning.

[0070] The uninterruptible power supply can also be used to control and / or perform the recording of local weather data and / or operating data of the wind farm or to provide the supply current required for this. For computing units, in particular wind farm computing units, which have data memories on which older weather forecasts are stored, the uninterruptible power supply can provide the supply current.

[0071] According to one embodiment, it is proposed that a predetermined wind forecast variable or expected power is compared with the current wind variable as a forecast variable during the forecast comparison. The expected power corresponds in particular to the output power that the wind farm ultimately feeds into the power supply grid and can be compared with the current wind variable, which then corresponds to the current output power of the wind farm.

[0072] However, it is also possible to use intermediate variables for comparison. The predetermined wind prediction variable corresponds to such an intermediate value, which can, for example, indicate a power that is still greater than the power actually fed in, because, for example, effects in the farm, such as the follow-up effect, are also taken into account. The wind prediction variable can also be a wind speed that has been predicted and compared with representative measured values ​​in the wind farm or in the vicinity of the wind farm.

[0073] According to one embodiment, it is proposed that the wind expected variable is determined from the weather forecast, in particular from the weather data of the weather forecast, by means of a weather model. The weather model may contain a correction rule, or the correction rule is connected downstream of the weather model. Optionally, the idealized feed-in power is determined from the wind expected variable by means of a wind power model. If the idealized feed-in power is determined, the expected power is determined from the idealized feed-in power by means of an available model. For the described case, the method therefore comprises at least three steps, namely determining the wind expected variable from the weather forecast, determining the idealized feed-in power therefrom, and determining the expected power therefrom.

[0074] Alternatively, it is conceivable that the expected power is determined from the expected wind variable with the aid of a wind farm model. The expected power is in this respect determined from the expected wind variable without an intermediate step via an idealized feed-in power.

[0075] In particular, it is proposed that the expected power is determined using a power estimator or the power estimator, wherein the power estimator comprises at least one or one of the following models: a meteorological model, a wind power model, an availability model or a wind farm model.

[0076] Four individual models are described in detail above and are based in particular on the suggestion that the models can be combined in a power estimator. It is also possible to consider that only some of the models are combined in a power estimator. In particular, the weather forecast or the variables determined here can then form the input for the power estimator, wherein other input variables can be added. The result, i.e. the output variable of the power estimator, is then the expected power, based on the corresponding input data. In this way and method, the above steps can also be combined in a power estimator. For example, the models used in each case can be connected in sequence.

[0077] According to one embodiment, it is proposed that the meteorological model is adjusted by means of a meteorological model adjustment. The meteorological model adjustment can be regarded as a matching rule or one of at least one matching rule. Likewise, additionally or alternatively, the wind power model can be adjusted by means of a wind power model adjustment. Here, the wind power model adjustment is the matching rule or one of the matching rules.

[0078] Preferably, the correction rule is adapted by means of one of the at least one adaptation rule. According to a variant, if the correction rule is part of the weather model, this can also be included in the adaptation of the weather model.

[0079] Likewise, in addition or alternatively, the availability model can be adjusted by means of an availability model adjustment. In this case, the availability model adjustment can then be regarded as one of the at least one matching rule.

[0080] Additionally or alternatively, it is provided that the wind farm model is adjusted by means of a wind farm model adjustment. Accordingly, the wind farm model adjustment can then be regarded as an adaptation rule.

[0081] Therefore, according to one embodiment, it is proposed to adjust each of the four models. The adjustment can be carried out in particular in that the result of the corresponding model is compared with the corresponding measured value or at least with the current value determined from the current value and this comparison is used to adjust the corresponding model.

[0082] Preferably, the power estimator is designed as a neural network, in particular when it is used to determine an idealized feed-in power or expected power. It is proposed that the neural network is trained in an offline method. In particular, the neural network is trained with the aid of meteorological measurement data of the wind farm. In addition, it is proposed that the trained neural network is used to determine the idealized feed-in power or expected power.

[0083] In particular, the use and training of such neural networks is a possibility for achieving the adjustment. The use of neural networks has the advantage in particular that the structural relationships on which the power estimator should be based as far as possible do not need to be known so precisely when using such neural networks, as long as only the structure of the neural network, including the area, is selected to be sufficiently large.

[0084] In particular, this is also based on the idea that particularly critical situations, in which the power should be predicted as accurately as possible, occur only in rare cases of grid reconstruction or even during black starts. Under normal circumstances, the neural network can be trained for this purpose. As long as the results are not used to control the wind farm, offline methods are also considered here, because the values ​​determined here are not directly included in the control, but are stored. As a result, the neural network and thus in particular the power estimator are constantly improved during operation.

[0085] The results can be used in particular when such a grid reconstruction or black start has to be carried out and the connection to the weather service providing the weather forecast is interrupted. It is therefore preferred to then go back to all older stored forecasts, which may require a high degree of adaptation of the wind farm accordingly.

[0086] According to the invention, a wind farm is also proposed. The wind farm is provided for feeding electrical power into a power supply grid, and it is proposed that the wind farm is connected to a grid connection point in order to feed electrical power into the power supply grid. In addition, the wind farm has an estimation device in order to determine the expected power in a predetermined feed-in time period. The estimation device can be implemented as a device, but can also be implemented as an implemented program in a control computer, in particular in a central farm computer.

[0087] It is proposed that the expected power describes the value of the power or the time profile of the power which is expected to be available as power from the wind for at least one wind farm in a predetermined feed-in period, in particular as the sum of all rotor powers of all wind energy installations of the wind farm and / or as feed-in power of the wind farm. Different possible meanings of the expected power have already been explained in conjunction with the method for feeding in and the explanations explained there also apply here.

[0088] Furthermore, an evaluation device is proposed in order to determine an expected accuracy for the expected power, which is a measure representing the accuracy with which the power reaches the expected power in the feed-in period. The evaluation device can also be provided as a device or implemented as a program.

[0089] In addition or alternatively, it is proposed that the evaluation device for determining the expected power comprises a wind expectation estimator, which is configured to determine or verify at least one wind expectation variable representing the wind speed to be expected with the help of a meteorological forecast, and comprises a correction unit, which is configured to determine the wind expectation variable based on the meteorological forecast and in addition with the help of a correction rule in relation to local meteorological data and / or operating data of the wind farm, wherein

[0090] The estimation device is configured to determine the expected power based on the wind expected variable. The wind farm is thereby configured to determine the expected power as explained above in conjunction with the aspects of the method for feeding in electrical power. The wind expected estimator, the correction unit and / or the estimation device can be provided individually or in combination as a device or as a solution implemented in a process computer.

[0091] In addition or alternatively, it is provided that the wind expected estimator is configured to repeatedly or continuously predetermine the wind expected variable in each comparison time period with the aid of a weather forecast. The wind expected estimator thus uses the weather forecast in order to always reconstruct it, in particular to predetermine the wind expected variable continuously or quasi-continuously.

[0092] The wind forecast variable represents the wind speed to be expected and can therefore respectively include a wind speed value. However, normalized values ​​or also power values ​​are also conceivable, which can also represent the wind speed to be expected. The power value represents the wind speed in particular if the power is based on the assumption that the power value respectively indicates the power that can be generated by means of a wind energy installation or a wind farm in dependence on the wind speed.

[0093] Furthermore, a comparison unit is provided in order to carry out a prediction comparison for each comparison period of a predetermined wind forecast variable, wherein the prediction variable is compared with a current wind variable which represents the current wind speed for the respective comparison period. The comparison unit can also be designed as a device or as an implemented program.

[0094] Furthermore, a matching unit is proposed in order to determine at least one matching rule from the prediction comparison in order to improve the expected power by means of at least one matching rule. The matching unit thus determines the matching rule. This was also explained above in conjunction with the method for feeding in electrical power and applies analogously here. The matching unit that determines the matching rule can therefore also be provided as an implementation of a device or a program.

[0095] The estimation device is then configured to determine the expected power based on the wind expected variable and the adaptation rule.

[0096] It is therefore proposed in particular that the wind farm is configured to carry out a method according to at least one of the above-described embodiments. In particular, the method can also be implemented in a central farm computer using the described devices and units. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] The invention will now be explained in more detail hereinafter by way of example with reference to the attached drawings.

[0098] Figure 1 A perspective view of a wind energy installation is shown.

[0099] Figure 2 A schematic diagram of a wind farm is shown.

[0100] Figure 3 The present invention shows a structure for schematically describing a method according to at least one embodiment.

[0101] Figure 4 A diagram with an uncertainty funnel is illustrated graphically.

[0102] Figure 5 The distribution curves are shown graphically. DETAILED DESCRIPTION

[0103] Figure 1A wind power plant 100 is shown having a tower 102 and a nacelle 104. A rotor 106 having a spinner 110 and three rotor blades 108 is arranged on the nacelle 104. During operation, the rotor 106 is set into rotational motion by the wind and thus drives a generator in the nacelle 104.

[0104] Figure 2 A wind farm 112 is shown by way of example with three wind energy installations 100, which can be identical or different. The three wind energy installations 100 therefore represent in principle any number of wind energy installations of a wind farm 112. The wind energy installations 100 provide their power, i.e., in particular the generated current, via a wind farm grid 114. In this case, the respectively generated currents or powers of the individual wind energy installations 100 are added together and a transformer 116 is usually provided, which steps up the voltage in the power plant so that it can then be fed into a supply grid 120 at a feed point 118, which is also generally referred to as a PCC. Figure 2 This is only a simplified illustration of a wind farm 112 which, for example, does not show a control device, although it is naturally present. The wind farm grid 114 can also be designed in another way, for example by also having a transformer at the output of each wind energy installation 100, which is merely another exemplary embodiment.

[0105] Furthermore, the wind farm 112 has a central power plant computer 130 which can be used synonymously, not only for Figure 2 The embodiment of the central power plant computer 130 is referred to as a wind farm computing unit. The central power plant computer 130 is particularly designed to communicate with each wind energy installation 100, in particular to transmit the control commands and / or information, but also to receive information from the wind energy installation 100. The information may include operating values ​​such as the currently generated power and also measured values ​​such as the detected wind speed or the measured temperature.

[0106] The central power plant computer 130 can also be coupled to a weather service 132 in order to receive weather forecasts therefrom. However, it is also conceivable that the wind farm provides information to the weather service, which is indicated by the double arrow. The connection 134 between the central power plant computer 130 and the weather service 132 is shown partially in dashed form in order to indicate that the weather service can be arranged separately at a location remote from the wind farm 112.

[0107] Furthermore, a grid operator 136 is shown, with which the wind farm 112 likewise communicates by means of the central power plant computer 130. Here too, a grid operator connection 138 is provided, which, as indicated by the double arrow, enables mutual communication. The grid operator connection 138 is also shown partially in dashed lines in order to indicate possible location distances.

[0108] Figure 3A block diagram 350 is shown, which illustrates a method for realizing a wind forecast or power forecast that is as accurate as possible at a location. It is first assumed that a weather service 532 provides weather data, such as, for example, various distribution values ​​for air pressure, temperature, wind and precipitation for a region, just to name a few examples, and that weather data may additionally or alternatively be provided by one or more weather satellites.

[0109] This is then input into the forecast block 352. In the forecast block, for example, a digital meteorological model or at least a part thereof can be stored, with the aid of which and with the aid of the input meteorological data a meteorological forecast can be created. In addition, correction rules are implemented in the forecast block 352, which correction rules can improve the meteorological forecast in relation to the operating data of the wind farm and / or local meteorological data, i.e. can be specifically adapted to the wind farm. In this regard, the correction rules can correct, at least improve, the meteorological forecast with respect to the wind farm, which can thus refit the meteorological forecast for the wind farm. The correction rules can also be implemented as part of the meteorological model. In particular, the wind speed is important for the present purpose and therefore the forecast block 352 outputs in particular the wind expectation variable P MO .

[0110] The wind prediction variable P MO It can be the wind speed or the change curve of the wind speed, that is, the change curve is expected in a comparative time period in the future. In particular, the wind prediction value is set as a short-term prediction. Since information about electrical power is particularly important for wind farms, it can also be proposed here that the wind prediction variable P MO is the power, which represents the wind speed, and the same applies to the wind speed curve. In particular, the rotor power can be used as a wind prediction variable, a value that indicates the value present at the rotor of the wind power plant. This value can also be used to infer the wind farm that is important and is being studied here. Wind prediction variable P MO This can therefore be the sum of all rotor powers to be expected in the comparative time period under consideration.

[0111] However, this is still understood to be a variable which represents the wind speed and to this extent the power plant issues or power plant effects or wind power plant effects which are the subject of the text are still not taken into account. Therefore, if the sum is based on the rotor power, this is initially a dummy variable.

[0112] Wind prediction variable P MO The wind power block 354 is then input. The wind power block 354 includes a wind power module, which is used to calculate the wind power from the expected variable P MO Determine the idealized feed power P iThe wind power block 354 can simulate a wind power model by means of a neural network. The wind power block 354 can obtain the characteristics of the neural network from the parameter block 356. The parameter block 356 can also transmit the entire structure of the neural network, including the parameterization, after a corresponding learning process.

[0113] However, the use of a neural network is only an example and other control implementations for the wind power model are also considered, which can also obtain corresponding parameterizations, possible structures and / or initial values ​​from the parameter block 356. It is also considered that the correction rule is applied first or additionally in the wind power block. This is particularly provided for the following situation, that is, the wind expected variable forms a power estimate and the power estimate is verified in the wind power block 354 by means of the correction rule. The correction in the wind power block 354 can also be combined with other changes in the wind expected variable.

[0114] The result of wind power block 354 is the idealized feed power P i , when the forecast, i.e., in particular the wind expected variable P MO The wind farm under consideration will theoretically produce the idealized feed-in power if all wind power facilities in the farm are fully available and correct. The expected power then corresponds to the verified wind expected variable. The wind direction is also taken into account here. The idealized feed-in power P i It can be regarded as a verified wind expectation variable and it is also related to the power plant effect in the wind farm. This includes the general appearance of the region in which the wind farm is located and the region surrounds the wind farm, but the wind energy facilities influence each other. On the one hand, this usually includes the weakening of the wind farm by the wind farm, but on the other hand, it also involves the specific influence of the upstream wind energy facilities relative to the downstream wind energy facilities, that is, just in the leeward side. All these relationships are taken into account in the wind power model, and the wind power box 354 applies the wind power model.

[0115] Idealized feed power P i This is then entered into the availability box 358. The availability box 358 takes into account the technical availability of each individual facility in the wind farm. The availability box obtains data on this from the data box 360. In the data box 360, all the availability data of the wind energy facilities of the wind farm are therefore collected and continuously updated. In particular, this includes, for example, when the wind energy facility fails. However, it is also considered that the wind energy facility is only allowed to operate in a reduced manner, because this is, for example, preset noise protection rules. All this information is stored in the data box 360 for each wind energy facility in the power plant and is passed to the availability box 358. Here, if necessary, the conventional data of the relevant wind energy facility, such as its rated power, can also be passed, if this data has not been permanently stored in the availability box 358.

[0116] In any case, the availability block 358 can be calculated from the idealized feed power P i Determine the expected power P of the wind farm F The expected power P F is the predicted power, in particular as a short-term forecast, which then ideally corresponds to the total power actually detected in the wind farm, which can also be the power fed in by the wind farm. However, the wind equivalent can also be used here, ie the wind speed that would cause this power.

[0117] Correspondingly, the measured wind farm power P is detected from the symbolically shown wind farm 312 M , or the wind equivalent.

[0118] Not only the expected power P F Moreover, the measured wind farm power P M It is then input into the comparison box 362 and compared. In this case, of course, the corresponding values ​​or curves of the same comparison time period are used. Thus, if the prediction is performed in a time period in the future, for example, half an hour, the expected power P obtained therefrom F Corresponding to the measured wind farm power P M In comparison, the wind farm power is measured correspondingly half an hour later than mentioned. In this respect, the comparison block 362 can also be correspondingly complex and in particular also have a memory for a plurality of expected powers.

[0119] In particular, the comparison block 362 then also performs an evaluation, wherein a separate block may also be used for this purpose, namely a matching unit, which performs the evaluation and creates the matching rules. Figure 3 In an embodiment, this is integrated together in the comparison block 362 .

[0120] As a result, in Figure 3 In the embodiment of , not only the adaptation rule for wind power block 354 and thus the wind power model implemented there is obtained, but also the adaptation rule for prediction block 352 and thus the meteorological model or models used there. The adaptation rule is called AV1 or AV2.

[0121] However, it is also conceivable to use only one of the two matching rules AV1 or AV2 or to use the other matching rule as well. Matching rule AV2 can also be used for matching the correction rule that is implemented in prediction block 352 .

[0122] In particular, it has been recognized that in grid reconstruction situations and the like, no current, ie newly calculated, estimates are available for renewable power plants, in particular wind farms, because in particular the computing center is offline or no longer has a data connection.

[0123] Forecasts are required in particular in the reconstruction of power grids, where already calculated forecasts can have significant deviations from the actually available power after a power outage of several hours. Forecasts should also be available at the wind farm level and should even be calculated at the wind farm level. Short-term forecasts are used for active grid operation control of power distribution networks or in the case of accounting grid management in a manner provided via a grid operator interface.

[0124] One concept is to provide the forecast by refitting old forecasts at wind farm level. Thus, old forecasts or weather forecasts are used and adapted to the specific wind farm.

[0125] In particular, the super-high wind farms at medium voltage substations and at high and low voltage levels.

[0126] This makes it possible to achieve improved operational control, in particular in critical grid situations.

[0127] It is therefore proposed in particular to adjust the forecasts, in particular the entire forecast or old weather forecasts, in order to create a wind park specific feed-in forecast in order to thereby create a short-term forecast for active operational control.

[0128] According to at least one embodiment:

[0129] - Proposes a device for maintaining and providing short-term forecasts at wind farm level, said device comprising the following functions:

[0130] - receiving forecasts at regular intervals, for example from a forecast service provider, such as a weather service, which forecasts here in particular represent weather forecasts,

[0131] - storing the estimate in a wind farm memory, in particular in a central power plant computer,

[0132] - determining and providing a short-term forecast, for example via a grid operator interface via which a connection of the wind farm to a grid operator can be established,

[0133] - refitting, i.e. adjusting, the saved estimate by available local measurements, which may be current and previous measurements,

[0134] And provide improved estimates, in particular as wind expected variables and / or expected power.

[0135] For local measurements you can use:

[0136] - anemometers, in particular nacelle anemometers, on wind energy installations at wind farms,

[0137] - nearby wind parks, in particular their anemometers or corresponding information about the fed-in power,

[0138] - a wind mast, which can be interfaced via Meteo DB or directly, and / or - a local weather station.

[0139] The proposed method according to at least one embodiment includes the following or a part thereof:

[0140] - Creation of wind farm specific feed-in forecasts based on external numerical weather forecasts and SCADA data, in particular through space-time regression methods, and

[0141] - a fast update cycle, i.e. a fast update of the detected and / or estimated values ​​with the help of an internal weather forecast model, and

[0142] -Use real-time SCADA data for data assimilation.

[0143] The controllable local load and the available memory, ie at least one memory loading state and / or the available power, are taken into account in the estimation or in order to minimize the uncertainty.

[0144] One goal is to improve the forecast accuracy of wind farm power by combining the file forecasts into a wind farm focused estimate.

[0145] For this purpose, the forecast model is divided into individual models which can be validated individually, namely in particular the meteorological model, the wind power model and the availability model.

[0146] It has been recognized that the interconnection of individual models optimized for a wind farm leads to more precise wind farm-specific estimates than a model which depicts the entire uncertainty chain, in particular wind forecasts, azimuth stability, follow-up models, technical availability and power losses, fuzzily or fused in one model and is optimized to form a combined file, i.e., for rewards and less for specific and local values.

[0147] It is proposed that in the forecast chain, i.e. in particular from the meteorological data from the forecast block 532 to the specific expected power, a correction loop or multiple correction loops are embedded, which correct the model forecast statistically, in particular via improvement or correction means: smoothing, linear regression, sliding average, dynamic weighting, deviation correction with the help of measured data. Depending on the available data source, SCADA power data (e.g. also from adjacent wind farms), nacelle anemometer data, status codes or meteorological measuring stations are used. Correspondingly, the correction with respect to the wind data is applied to the meteorological model and in the case of power data to the wind farm model, which can in particular be a neural network.

[0148] According to one aspect, in particular the accuracy of the estimate is also taken into account. This is based on the following idea.

[0149] In the case of a grid reconstruction, there are no longer current, i.e. recalculated, estimates available for renewable power plants, such as wind farms, because, for example, the computing center is offline or no longer has a data connection. After a power outage of several hours, the already calculated estimates already have significant deviations or significant uncertainties from the actually available power. However, in critical grid situations, grid operators also need a quantitative analysis of the uncertainties in addition to the estimates in order to take them into account when regulating the wind farm. The estimates and the uncertainty of the estimates must also exist at the wind farm level and can also be calculated at the wind farm level.

[0150] However, the short-term forecasts including uncertainties provided via the grid operator interface can also be used for active grid operation control of the power distribution grid or in accounting grid management.

[0151] The concept is, in particular, to carry out a quantitative analysis of the estimation inaccuracy.

[0152] This concerns in particular medium voltage substations, high voltage and extra-high voltage wind farms.

[0153] In particular, an improved operational control should also be achieved in critical grid situations.

[0154] It can also be seen as a goal to always provide the maximum possible power, depending on the grid situation and the permissible estimated deviations.

[0155] In particular, a device or method is proposed for maintaining and providing an estimate including an estimate uncertainty at the wind farm level. The following steps or methods are proposed for this purpose:

[0156] - receiving estimates at regular intervals, for example via an estimate service provider, in particular with a probability distribution,

[0157] - save the estimate in the wind farm memory,

[0158] - Calculation of possible deviations from the estimate or probability distribution depending on the meteorological situation, the available facilities, the estimated age and / or the data of the available measuring sensors.

[0159] Alternatively: the reliable minimum power is calculated and communicated according to a predefined minimum reliability (eg 95%).

[0160] As an example the following can be explained:

[0161] The grid operator can specify via the grid operator interface, for example, what availability the grid operator requires a minimum available power and a (short-term) forecast, or he can directly query the uncertainty and thus take it into account in the control.

[0162] As additional possibilities we propose:

[0163] Uncertainties are reduced by taking into account a plurality of wind farms. The available controllable loads and storage are taken into account in the calculation of the uncertainty. This can mean, for example, that in a half-full loading state of the storage, half of the storage capacity is available to compensate for the estimated uncertainty.

[0164] Preliminary studies have shown that with a multi-model approach, the uncertainty of each individual model can be reduced by additional information and manufacturer expertise about wind energy facilities and wind farms, which do not have external forecasting companies / traders. The expected uncertainty of a wind farm transfer model created with the help of a neural network can be -2.5% (depending on the quality and availability of historical power data).

[0165] In-house expertise on wind energy facility technology reduces the uncertainty in the estimated technical turbine availability to 1%. The greatest uncertainty lies in the meteorological forecast. This can be reduced by intelligent weighting of different numerical meteorological models and by statistical corrections based on various measured data in the wind farm environment. It is therefore advantageous to combine these aspects.

[0166] The uncertainty funnel of the forecast can be reduced on the one hand by short-term correction using the measured data and additionally or alternatively by coupled storage solutions in the wind park. The combination of multiple wind parks operated at the grid operator interface can also reduce the uncertainty.

[0167] All three measures support the possibility of providing significantly higher P90 values ​​to grid operators for forecasting, which describe the wind farm power that occurs with a probability of 90%. Figure 5 The diagram in FIG. 1 shows a probability distribution for different matches assuming an approximately normal distribution. The ordinate shows the probability density and the abscissa shows the normalized scale for the values ​​of the investigated results. The scale can be assumed to be a normalized representation of the wind farm power, in particular the rotor power, for example with respect to the maximum rotor power or the maximum sum of all rotor powers, i.e., for example, without taking into account restrictions to the rated power. The value denoted by 50 here is the predicted power. The value can also represent an absolute value, such as 50 MW, for example. In this figure, in particular, a basic comparison of a plurality of prediction curves is involved.

[0168] In this case, the matching, in particular by means of correction rules, is carried out or simulated for a 1-hour forecast (1h fitting), for a 3-hour forecast (3h fitting) and for a 6-hour forecast (6h fitting) and for this, a 1-hour distribution 501, a 3-hour distribution 503 and a 6-hour distribution 506 are shown correspondingly. The P1, P10, P50, P90 and P99 positions are also marked for the 3-hour forecast. In particular, it can be seen that the accuracy is improved when matching the short-term forecast and thus in the short-term forecast. For the 1-hour forecast, the P90 value is also indicated as "P90 (1h)". The P90 value indicates the wind farm power, which occurs with a probability of at least 90%. This therefore corresponds to the area of ​​the surface below the corresponding curve to the right of the P90 value with respect to the entire surface below the curve.

[0169] The surface to the left of the P90 value has an area of ​​10% accordingly, and this surface is plotted for the 3-hour forecast as residual surface 510. The residual surface is improved with the aid of shorter forecasts, which can be seen at the high value of 50 for the 1-hour forecast. However, this also causes the curve to become thinner and the 10% residual surface to end further to the right, so that the P90 value also moves further to the right and is thus closer to the predicted power.

[0170] based on Figure 3 It can be illustrated that the uncertainty of the weather forecast of the weather service 532 can be around 6 to 8%. The uncertainty of the model describing the wind farm, which can be illustrated by means of the parameter box 356 or by means of the wind power box 354 in which calculations are subsequently performed, can be around 2.5%. The availability of the technology, which can be illustrated by means of the data box 36 or the availability box, is around 1%.

[0171] Figure 4 Explain the so-called uncertainty funnel. Figure 4 An uncertainty diagram 400 is shown in which a predicted value is shown as a prediction curve over time together with its uncertainty.

[0172] The time axis is divided into three regions for this purpose. A first time region 40 is not yet in the future, so that the estimated value corresponds to the measured value, so that there is no uncertainty.

[0173] The second time region 402 is in the future and thus has an uncertainty, which increases with time. The actual value may therefore be in the uncertainty region shown and deviate from the prediction curve 404 accordingly.

[0174] The third time region 403 is still further in the future and now a clear difference can be distinguished between the wind farm uncertainty curve 405 and the meteorological uncertainty curve 406. The wind farm uncertainty curve 405 describes the uncertainty due to wind farm inaccuracies and measurement uncertainties, while the meteorological uncertainty curve 406 describes the uncertainty due to meteorological forecasts. The wind farm uncertainty curve 405 forms a significantly smaller funnel than the meteorological uncertainty curve 406.

Claims

1. A method for feeding electrical power into a power supply grid by means of at least one wind farm, the wind farm having a plurality of wind energy installations, wherein - the wind farm feeds electrical power into the power supply grid at a grid connection point, and - Determine the expected power (P F ),in The expected power specifies a value of a power or a time profile of a power which is expected to be provided as power from the wind to at least one wind farm in the predetermined feed-in time period, and wherein - determining an expected accuracy for the expected power, the expected accuracy being a measure representing with what accuracy the power reaches the expected power during the feeding period, and / or wherein - In order to determine the expected power (P F ), - determining with the aid of a meteorological forecast at least one wind expectation variable representing the wind speed to be expected ( P MO ),and - additionally determining or verifying the wind prediction variable from the weather forecast by means of correction rules in relation to local weather data and / or operating data of the wind farm, wherein - Based on the wind prediction variable (P MO ) to determine the expected power (P F ), A black start and / or grid reconstruction is planned based on the expected power.

2. The method according to claim 1, It is characterized in that - create weather forecasts for the forecast time period, and - Creation of wind expected variables and / or expected power as short-term forecasts in the forecast period, where - the predicted time period is at least ten times longer than the estimated time period.

3. The method according to claim 1 or 2, It is characterized in that - repeatedly or continuously predetermining the wind prediction variable (P MO ),and - The wind expected variable (P MO ) performs a forecast comparison for each comparison period, where the predictor variable (P MO , P F , P i ) and the current wind variable (P M ), the current wind variable represents the current wind speed of the corresponding comparison time period, and - determining at least one matching rule from the prediction comparison in order to improve the expected power (P F ).

4. The method according to claim 1 or 2, It is characterized in that The expected power (P F ) - using a power estimator to determine, and / or - In addition, the expected power (P F ), wherein the boundary conditions are selected from the following list: - technical availability of said wind energy facility of said wind park, - information about available controllable loads, - information about the storage available for feeding electrical power, and - information on the follower effect in wind farms, and / or - additionally determining the expected power (P F ).

5. The method according to claim 1 or 2, It is characterized in that The expected accuracy includes at least one of the following information: a first power limit, which specifies a power or a power profile which does not fall below the power or the power profile with a predetermined probability of achievement within a predetermined feed-in period, a second power limit, which specifies a power or a power characteristic curve which may not be fallen below in the medium-term average value, A third power limit specifies a power or a power profile which may not be fallen below in the short-term average value.

6. The method according to claim 1 or 2, It is characterized in that - said weather forecast is regularly transmitted to at least one of said wind farms by an external weather service, - the wind farm at least temporarily stores the respectively transmitted weather forecast, - In case of an outage with an external weather service, estimating said expected power based on at least one stored weather forecast.

7. The method according to claim 1 or 2, It is characterized in that the grid connection point is connected to a grid section of the power supply grid, - the expected power (P F ) is transmitted to a grid operator operating the grid section.

8. The method according to claim 1 or 2, It is characterized in that the weather forecast comprises at least one time profile of the wind speed to be expected and / or the weather forecast outputs values ​​of the wind speed to be expected in predetermined repetition intervals, - determining the expected power and / or the minimum value of the expected power from the weather forecast as a time power curve, and - The power curve is shifted, stretched and / or compressed by means of a correction rule.

9. The method according to claim 3, It is characterized in that - recording meteorological measurements of the wind park in the wind park and / or in the vicinity of the location of the wind park in order to determine the correction rule and / or the adaptation rule, and - The meteorological measurements include one, more or all variables from the following list: - wind speed, -wind direction, -temperature, - air density, and - Sun exposure, and / or To determine the correction rule and / or the adaptation rule, measured values ​​and / or operating values ​​of the wind energy installations of the wind park are used.

10. The method according to claim 1 or 2, It is characterized in that - performing a calculation on a wind farm computing unit of the wind farm to determine the expected power (P F ).

11. The method according to claim 1 or 2, It is characterized in that As a predictor variable in predictive comparison, - Predetermined wind expected variable (P MO ),or - the expected power (P F ), -With the current wind variable (P M ) for comparison.

12. The method according to claim 1 or 2, It is characterized in that - the wind expected variable (P MO ) determined from weather forecasts with the aid of a weather model, or - Determine the idealized feed-in power (P) from the wind prediction variables with the aid of a wind power model i ),and - If the ideal feed power (P i ), then with the aid of the availability model, we can get the idealized feed-in power (P i ) is used to determine the expected power (P F ), while in other cases - Determine the expected power (P) from the expected wind variables with the aid of a wind farm model F ).

13. The method according to claim 12, It is characterized in that The weather model is adapted by means of a weather model adaptation as one of the at least one adaptation rule.

14. The method according to claim 1 or 2, It is characterized in that - The power estimator is constructed as a neural network, where - The neural network is trained in an offline method.

15. The method according to claim 1, It is characterized in that The power is provided to at least one wind park as the sum of all rotor powers of all wind energy installations of the wind park and / or as feed-in power of the wind park during the predetermined feed-in time period.

16. The method according to claim 2, It is characterized in that The predicted time period is at least one hundred times longer than the estimated time period.

17. The method according to claim 4, It is characterized in that - For a wind farm, the technical characteristics of each wind energy facility are stored in a table, where The table stores the predicted technical performance for each wind energy installation according to its azimuth direction or according to the wind direction or according to the time of day.

18. The method according to claim 5, It is characterized in that The medium-term average is formed over a period of 10 seconds to 60 seconds.

19. The method according to claim 5, It is characterized in that The short-term average is formed over a period of 5 to 10 seconds.

20. The method according to claim 6, It is characterized in that In the event of an outage with an external weather service, the expected power is estimated based on at least one stored weather forecast, such that the stored weather forecast is adjusted taking into account local current weather measurements.

21. The method according to claim 7, It is characterized in that - In the case of a grid construction scenario, the expected power (P F ) is transmitted to a grid operator operating the grid segment, at least the grid segment failing in the grid construction scenario.

22. The method according to claim 7, It is characterized in that - In addition to transmitting the expected power (P F ), transmit the expected accuracy, and / or A desired accuracy value is taken into account, which indicates with which desired accuracy the desired power should be provided.

23. The method according to claim 22, It is characterized in that The expected accuracy is received externally.

24. The method according to claim 22, It is characterized in that The expected accuracy is received from a grid operator.

25. The method according to claim 8, It is characterized in that The repetition interval is in the range of 5 minutes to 30 minutes.

26. The method according to claim 8, It is characterized in that - determining the expected power and / or the minimum value of the expected power from the weather forecast as a time power profile by means of a power estimator.

27. The method according to claim 8, It is characterized in that - The amplitude and / or the time extension of the power curve is shifted, stretched and / or compressed by means of a correction rule.

28. The method according to claim 9, It is characterized in that The available power is estimated from the measured values ​​and / or operational values.

29. The method according to claim 10, It is characterized in that The wind farm calculation unit is provided with an uninterruptible power supply and, in the event of a grid failure of the power supply grid, continues to determine the expected power using the uninterruptible power supply and transmits it to the grid operator or the grid operator.

30. The method according to claim 12, It is characterized in that - using a power estimator or said power estimator to determine said expected power (P F ), and the power estimator is at least one model or includes a model from the following list: - Weather models, - Wind power model, - Availability model, and - Wind farm model.

31. The method according to claim 14, It is characterized in that The power estimator is designed to determine an idealized feed-in power or an expected power.

32. The method according to claim 14, It is characterized in that - the neural network is trained with the aid of meteorological measurement data of the wind farm, and - A trained neural network is used to determine the idealized feed-in power or the expected power.

33. The method according to claim 12, It is characterized in that The wind power model is adapted by means of a wind power model adaptation as one of the at least one adaptation rule.

34. The method according to claim 12, It is characterized in that The availability model is adapted by means of an availability model adaptation as one of the at least one adaptation rule.

35. The method according to claim 12, It is characterized in that The wind farm model is adapted by means of a wind farm model adaptation as one of the at least one adaptation rule.

36. A wind farm for feeding electrical power into a power supply grid, wherein the wind farm - connected to a grid connection point in order to feed electrical power into said power supply grid, and - having an estimation device in order to determine the expected power (P F ),in the expected power specifies a value of a power or a time profile of a power which is expected to be provided as power from the wind to at least one of the wind farms in the predetermined feed-in time period, and wherein An evaluation device is provided in order to determine an expected accuracy for the expected power, the expected accuracy being a measure representing with which accuracy the power reaches the expected power during the feeding period, and / or wherein - the estimation device is used to determine the expected power (P F ) - comprising a wind expectation estimator configured to determine or verify, with the aid of a meteorological forecast, at least one wind expectation variable (P MO ),and - comprises a correction unit which is configured to determine a wind prediction variable from the meteorological forecast, additionally with the aid of a correction rule as a function of operating data of the wind farm and / or local meteorological data, wherein - the estimating device is configured to, based on the wind expected variable (P MO ) to determine the expected power (P F ), A black start and / or grid reconstruction is planned based on the expected power.

37. The wind farm according to claim 36, It is characterized in that The wind forecast estimator is configured to repeatedly or continuously predetermine the wind forecast variable (P MO ), and where The wind farm has a comparison unit for determining the wind expected variable (P MO ) performs a forecast comparison for each comparison period, where the predictor variable (P MO , P F , P i ) and the current wind variable (P M ), the current wind variable represents the current wind speed of the corresponding comparison time period, and a matching unit is provided in order to determine at least one matching rule from the prediction comparison in order to improve the expected power (P F ).

38. A wind farm according to claim 36 or 37, It is characterized in that The wind farm is configured to carry out at least one method according to any one of claims 1 to 35.

39. The wind farm according to claim 36, It is characterized in that The power is provided to at least one wind park as the sum of all rotor powers of all wind energy installations of the wind park and / or as feed-in power of the wind park during the predetermined feed-in time period.

40. The wind farm according to claim 38, It is characterized in that A central power plant computer is provided for executing the methods, so that at least one of the methods is carried out on the power plant computer.

Images