METHOD AND WIND PARK FOR FEEDING ELECTRIC POWER INTO AN ELECTRIC SUPPLY NETWORK

Abstract

A method for feeding electric power into an electric supply network using a wind park having wind power installations is provided. An expected power is determined for a predetermined feed-in period, where the expected power indicates a power value or temporal profile of power expected to be available to the park as power from wind in the predetermined feed-in period. An expected accuracy is determined and is a measure of how accurately the power reaches the expected power in the feed-in period. To determine the expected power, at least one expected wind variable representative of the expected wind speed is determined using a weather forecast, and the expected wind variable is additionally determined or verified, proceeding from the weather forecast, using a correction rule based on local weather data and/or operating data of the park. The expected power is determined on the basis of the expected wind variable.

Claims

1. A method for feeding electric power into an electric supply network at a network connection point using at least one wind park including a plurality of wind power installations, comprising: determining an expected power for a predetermined feed-in period, the expected power indicating a power value or a temporal profile of wind power expected to be available from the at least one wind park in the predetermined feed-in period, wherein determining the expected power includes: determining, using a weather forecast, at least one expected wind variable representative of an expected wind speed, determining or verifying, using a correction rule, the at least one expected wind variable based on local weather data, operating data of the at least one wind park or both local weather data and the operating data of the at least one wind park, and determining the expected power based on the at least one expected wind variable; and determining an expected accuracy of the expected power, the expected accuracy being a measure of an accuracy associated with wind power reaching the expected power in the predetermined feed-in period.

2. The method as claimed in claim 1, wherein: the weather forecast is generated over a forecast period, the weather forecast is used to generate the at least one expected wind variable and/or the expected power as a short-term prediction over a prediction period, and the forecast period is at least ten times as long as the prediction period.

3. The method as claimed in claim 1, comprising: determining the expected wind variable recurrently or continuously for respective comparison periods; performing a forecast comparison for each comparison period by comparing a forecast variable with a current wind variable representative of a current wind speed in the respective comparison period; determining at least one adjustment rule from the forecast comparison; and adjusting the expected power using the at least one adjustment rule.

4. The method as claimed in claim 1, comprising: determining the expected power using a power estimator, and/or determining the expected power taking into account boundary conditions, wherein the boundary conditions are selected from a list including: technical availability of the wind power installations in the at least one wind park, information relating to available controllable loads, information relating to storage devices available for feeding in electric power, and information relating to follow-on effects in the at least one wind park, and/or determining the expected power by individually considering the plurality of wind power installations in the at least one wind park, wherein: a technical performance of each wind power installation of the plurality of wind power installations is stored in a table, and the table includes a forecast technical performance for each wind power installation of the plurality of wind power installations on the basis of an azimuth orientation of the wind power installation, a wind direction of the wind power installation or a time of day.

5. The method as claimed in claim 1, wherein the expected accuracy includes at least one of: a first power limit indicating a power or a power profile which is not undershot within the predetermined feed-in period with a predetermined realization probability, a second power limit indicating a power or a power profile which is not to be undershot over a period of 10 to 60 seconds, and a third power limit indicating a power or a power profile which is not to be undershot over a period of 5 to 10 seconds.

6. The method as claimed claim 1, comprising: regularly transmitting, by an external weather service, the weather forecast to the at least one wind park; temporarily storing, by the at least one wind park, the respectively transmitted weather forecast; and in response to an interruption to the external weather service, estimating the expected power on the basis of at least one stored weather forecast and adjusting the at least one stored weather forecast based on current local meteorological measured values.

7. The method as claimed in claim 1, wherein: a black start and/or network restoration are planned on the basis of the expected power, the network connection point is connected to a network section of the electric supply network, the expected power is transmitted to a network operator operating the network section in the event that the network section fails, the expected accuracy together with the expected power, and/or an accuracy target value received from the network operator and indicating the expected accuracy with which the expected power is intended to be provided is taken into account.

8. The method as claimed in claim 1, wherein: the weather forecast includes at least one temporal profile of an expected wind speed, and/or the weather forecast outputs a value of an expected wind speed at predetermined repetition intervals, the expected power and/or a minimum value of the expected power are determined as the at least one temporal power profile from the weather forecast, and the power profile is shifted, stretched and/or compressed in amplitude and/or temporally using a correction rule.

9. The method as claimed in claim 1, comprising: recording meteorological measured values of the wind park in the wind park and/or in a local vicinity of the wind park to determine the correction rule and/or an adjustment rule, wherein the meteorological measured values includes one of: wind speed; wind direction; temperature; air density; and solar radiation; and/or determining the correction rule and/or the adjustment rule based on measured values and/or operating values of the plurality of wind power installations in the wind park are used to determine; and/or estimating an available power from the measured values and/or operating values.

10. The method as claimed in claim 1, comprising: determining the expected power of the wind park by a wind park computer.

11. The method as claimed in claim 3, wherein comparing the forecast variable with the current wind variable includes comparing a predetermined expected wind variable, or the expected power with the current wind variable.

12. The method as claimed in claim 1, comprising: determining the expected wind variable from the weather forecast using a weather model, or determining an idealized feed-in power from the expected wind variable using a wind power model, and determining the expected power from the idealized feed-in power using an availability model, and determining the expected power from the expected wind variable using a park model, wherein the expected power is determined using a power estimator including at least one model from: the weather model; the wind power model; the availability model; and the park model.

13. The method as claimed in claim 12, wherein: the weather model is configured using weather model adaptation as one of at least one adjustment rules, the wind power model is configured using wind power model adaptation as one of the at least one adjustment rules, the availability model is configured using availability model adaptation as one of the at least one adjustment rules, and/or the park model is configured using park model adaptation as one of the at least one adjustment rules.

14. The method as claimed in claim 12, wherein: the power estimator, for determining the idealized feed-in power or the expected power, is a neural network, the neural network is trained off-line or the neural network is trained using meteorological measured values of the wind park, and the trained neural network is used to determine the idealized feed-in power or the expected power.

15. A wind park for feeding electric power into an electric supply network at a network connection point, comprising: a wind park computer configured to: determine an expected power for a predetermined feed-in period, the expected power indicating a power value or a temporal profile of wind power expected to be available from the wind park in the predetermined feed-in period, wherein determining the expected power includes: determining, using a weather forecast, at least one expected wind variable representative of an expected wind speed, determining or verifying, using a correction rule, the at least one expected wind variable based on local weather data, operating data of the at least one wind park or both local weather data and the operating data of the at least one wind park, and determining the expected power based on the at least one expected wind variable; and determine an expected accuracy of the expected power, the expected accuracy being a measure of an accuracy associated with wind power reaching the expected power in the predetermined feed-in period.

16. The wind park as claimed in claim 15, wherein the park computer is configured: determine the expected wind variable recurrently or continuously for respective comparison periods, perform a forecast comparison for each comparison period by comparing a forecast variable with a current wind variable representative of a current wind speed in the respective comparison period, determine at least one adjustment rule from the forecast comparison, and adjust the expected power using the at least one adjustment rule.

17. (canceled)

18. The method as claimed in claim 1, wherein the expected power is a sum of a plurality of rotor powers of the plurality of wind power installations, respectively.

19. The method as claimed in claim 2, wherein the forecast period is at least one hundred times as long as the prediction period.

20. The method as claimed in claim 10, wherein the wind park computer has an uninterruptible power supply and, in the event of a network failure of the electric supply network, the wind park computer continues operating using the uninterruptible power supply, wherein the operating includes determining the expected power and transmitting the expected power to a network operator.

Description

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0099] The invention is now explained in more detail below, by way of example, with reference to the accompanying figures.

[0100] FIG. 1 shows a perspective illustration of a wind power installation.

[0101] FIG. 2 shows a schematic illustration of a wind park.

[0102] FIG. 3 shows a structure for schematically describing the method according to at least one embodiment.

[0103] FIG. 4 illustratively shows a graph with an uncertainty funnel.

[0104] FIG. 5 illustratively shows a distribution curve.

DETAILED DESCRIPTION

[0105] FIG. 1 shows a wind power installation 100 having a tower 102 and a nacelle 104. A rotor 106 having three rotor blades 108 and a spinner 110 is arranged on the nacelle 104. During operation, the rotor 106 is caused to rotate by the wind and thereby drives a generator in the nacelle 104.

[0106] FIG. 2 shows a wind park 112 having, by way of example, three wind power installations 100 which may be identical or different. The three wind power installations 100 are therefore representative of fundamentally any desired number of wind power installations in a wind park 112. The wind power installations 100 provide their power, specifically the generated current in particular, via an electric park network 114. In this case, the respectively generated currents or powers of the individual wind power installations 100 are added, and a transformer 116 is usually provided and steps up the voltage in the park in order to then feed it into the supply network 120 at the feed-in point 118, which is also generally referred to as the PCC. FIG. 2 is only a simplified illustration of a wind park 112, which does not show a controller, for example, even though a controller is naturally present. The park network 114 may also be configured differently, for example, by virtue of a transformer also being present at the output of each wind power installation 100, for example, to name just one other exemplary embodiment.

[0107] The wind park 112 also has a central park computer 130 which can be synonymously referred to, not only for the embodiment in FIG. 2, as a wind park computing unit. This central park computer 130 is provided, in particular, for the purpose of communicating with each wind power installation 100, in particular transmitting control commands and/or information to the latter or receiving information from the wind power installations 100. Such information may include operating values, such as currently generated powers, and measured values, such as a captured wind speed or a measured temperature.

[0108] The central park computer 130 may also be coupled to a weather service 132 in order to receive weather forecasts therefrom. However, the fact that the wind park provides the weather service with information, which is indicated by the double-headed arrow, also comes into consideration. The connection 134 between the central park computer 130 and the weather service 132 is partially depicted using dashed lines in order to indicate that the weather service may be arranged regionally far away from the wind park 112.

[0109] A network operator 136, with which the wind park 112 can likewise communicate by means of the central park computer 130, is also indicated. A network operator connection 138 which enables mutual communication, indicated by the double-headed arrow, is also provided here. The network operator connection 138 is also partially illustrated using dashed lines in order to indicate the possible regional distance.

[0110] FIG. 3 shows a structural diagram 350 which illustrates a method for achieving a wind prediction or a power prediction which is locally as accurate as possible. The method may be performed by the central park computer 130. In this respect, it is initially assumed that a weather service 532 provides weather data, for example various distributed values of air pressure, temperature, wind and precipitation for a region, to name just a few examples, and additionally or alternatively can provide weather data from one or more weather satellites.

[0111] These are then input to a forecast block 352. The forecast block may store, for example, numerical weather models or at least one of the latter, with the aid of which and with the aid of the input weather data a weather forecast can be created. A correction rule is additionally implemented in the forecast block 352 and can improve the weather forecast on the basis of local weather data and/or operating data of the wind park, namely can specifically adjust it to the wind park. In this respect, the correction rule can correct and at least improve the weather forecast based on the wind park and can therefore refit the weather forecast for the wind park. The correction rule may also be implemented as part of the weather model. The wind speed, in particular, is important for the present purposes and the forecast block 352 therefore outputs, in particular, an expected wind variable P.sub.MO.

[0112] This expected wind variable P.sub.MO may be a wind speed or a profile of a wind speed which is specifically expected in a comparison period in the future. In particular, the expected wind value is provided as a short-term prediction. Since information relating to electric power is particularly relevant to the wind park, provision may also be made here for the expected wind variable P.sub.MO to be a power which is representative of a wind speed; the same then also applies to a wind speed profile. In this case, it is possible to use, in particular, a rotor power as an expected wind variable, which indicates a value above a value applied to a rotor of a wind power installation. Such a value may also be extrapolated to the relevant wind park examined here. The expected wind variable P.sub.MO could therefore be the sum of all expected rotor powers in the comparison period under consideration.

[0113] Nevertheless, this should be understood as meaning a variable which is representative of a wind speed and, in this respect, also subsequently excludes addressed park problems or park effects or wind power installation effects. If the sum of the rotor powers is therefore taken as a basis here, this is primarily a fictitious variable.

[0114] The expected wind variable P.sub.MO is then input to a wind power block 354. The wind power block 354 comprises a wind power model which is used to determine an idealized feed-in power P.sub.i from the expected wind variable P.sub.MO. The wind power block 354 may simulate the wind power model by means of a neural network. The wind power block 354 can receive properties of the neural network for this purpose from a parameter block 356. In this case, the parameter block 356 may also transfer the entire structure of the neural network including parameterization after a corresponding learning process.

[0115] However, the use of a neural network is only one example, and other control-related implementations, which can likewise receive corresponding parameterizations, possibly structures and/or initial values from the parameter block 356, also come into consideration for the wind power model. The fact that the correction rule is used only or additionally in the wind power block also comes into consideration. This is provided, in particular, for the case in which the expected wind variable forms a power prediction and this is verified in the wind power block 354 by means of the correction rule. This correction in the wind power block 354 can also be combined with further changes in the expected wind variable.

[0116] The result of the wind power block 354 is an idealized feed-in power P.sub.i which could be theoretically generated by the examined wind park if the forecast, that is to say, in particular, the expected wind variable P.sub.MO, is correct and all wind power installations in the park are also completely available. The expected power would then correspond to the verified expected wind variable. In this case, the wind direction is also concomitantly taken into account. The idealized feed-in power P.sub.i can be considered to be a verified expected wind variable and also depends, in particular, on park effects in the wind park. These include the general topology of the terrain in which the wind park is located and which surrounds the wind park, but also the mutual influence of the wind power installations. This generally includes, on the one hand, an attenuation of the wind field by the wind park, but, on the other hand, may also relate to specific influences of an upstream wind power installation with respect to a downstream wind power installation, that is to say a wind power installation precisely downwind. All of these relationships are taken into account in the wind power model which is used by the wind power block 354.

[0117] The idealized feed-in power P.sub.i is then input to the availability block 358. The availability block 358 takes into account the technical availability of each individual installation in the wind park. The availability block obtains the data for this purpose from the data block 360. All availability data relating to the wind power installations in the wind park are therefore collected and continuously updated in the data block 360. Said data block contains, in particular, when a wind power installation fails, for example. However, the fact that a wind power installation can be operated only in a reduced manner because this is prescribed by noise protection rules, for example, also comes into consideration. All information of this type is stored for each wind power installation in the park in the data block 360 and is transferred to the availability block 358. In this case, general data relating to the relevant wind power installation, for example its nominal power, may possibly also be transferred if such data are not already permanently stored in the availability block 358.

[0118] In any case, the availability block 358 can determine an expected power P.sub.F of the wind park from the idealized feed-in power P.sub.i. This expected power P.sub.F is a forecast power, in particular in the form of a short-term prediction, which then ideally corresponds to the power actually captured overall in the wind park, which may also be the power fed in by the wind park. However, it is also possible here to use a wind equivalent, that is to say a wind speed, which would result in such a power.

[0119] Accordingly, a measured wind park power P.sub.M or said wind equivalent is captured by a symbolically illustrated wind park 312.

[0120] Both the expected power P.sub.F and the measured wind park power P.sub.M are then input to a comparison block 362 and compared. In this case, the respective values or profiles for the identical comparison period are naturally used. If a forecast is therefore made for a period which is approximately half an hour in the future, the expected power P.sub.F determined therefrom is accordingly compared with the measured park power P.sub.M which is accordingly measured said half an hour later. In this respect, the comparison block 362 may also be accordingly complex and may also have, in particular, a storage device for a plurality of expected powers.

[0121] In particular, the comparison block 362 then also carries out an evaluation, wherein a separate block could also be used for this purpose, specifically an adjustment unit which carries out this evaluation and creates an adjustment rule. In the embodiment in FIG. 3, this is concomitantly integrated in the comparison block 362.

[0122] As the result, both an adjustment rule for the wind power block 354 and therefore the wind power model implemented there and an adjustment rule for the forecast block 352 and therefore the weather model(s) used there emerge in the embodiment in FIG. 3. The adjustment rules are referred to as AV1 and AV2.

[0123] However, the practice of using only one of the two adjustment rules AV1 or AV2 mentioned or of using other adjustment rules also comes into consideration. The adjustment rule AV2 can also be used to adjust the correction rule which is implemented in the forecast block 352.

[0124] In particular, it has been recognized that a current, that is to say newly calculated, prediction is no longer available under certain circumstances for regenerative power plants, in particular wind parks, in a network restoration situation because computing centers, in particular, are off-line or no longer have a data connection.

[0125] A prediction is needed, in particular, during network restoration where the predictions which have already been calculated may already have a considerable deviation from the actual available power after a few hours of power failure. The prediction should also be available at the wind park level and should even be calculated, under certain circumstances, at the wind park level. A short-term prediction, provided via the network operator interface, may also be used for active network operational management of distribution networks or in accounting grid management.

[0126] One idea is to provide this prediction by refitting old predictions at the wind park level. Old predictions or weather forecasts are therefore used and are adjusted to the specific wind park.

[0127] In particular, wind parks on medium-voltage substations and on high-voltage and extra-high-voltage levels are proposed.

[0128] This makes it possible to achieve, in particular, improved operational management in critical network situations.

[0129] It is therefore proposed, in particular, to adjust predictions, in particular old predictions or old weather forecasts, in order to create wind-park-focused feed-in predictions in order to thus in turn create short-term predictions for active operational management.

[0130] At least one embodiment proposes an apparatus for holding and providing a short-term prediction at the wind park level, comprising the following functionality: [0131] receiving a prediction, which here represents a weather forecast in particular, at regular intervals, for example from prediction service providers such as weather services; [0132] storing the prediction in a wind park storage device, in particular in a central park computer; [0133] determining and providing a short-term prediction, for example providing a prediction via a network operator interface which can be used to establish a connection from the wind park to a network operator; [0134] refitting, that is to say adjusting, a stored prediction by means of available local measured values which may be current and past measured values; and [0135] providing an improved prediction, in particular in the form of the expected wind variable and/or expected power.

[0136] The following can be used for the local measured values: [0137] an anemometer on wind power installations in the wind park, in particular a nacelle anemometer; [0138] nearby wind parks, in particular their anemometers, or corresponding information relating to a feed-in power; [0139] wind metmasts which can be connected via Meteo DB or directly; and/or local weather stations.

[0140] The proposed method at least according to one embodiment comprises the following or parts thereof: [0141] creating wind-park-focused feed-in predictions on the basis of external numerical weather forecast and SCADA data, in particular by means of spatiotemporal regression methods; [0142] rapid-update cycle, that is to say the practice of carrying out rapid updates of the captured and/or predicted values using an internal weather forecast model; and in the process [0143] using live SCADA data for data assimilation.

[0144] Controllable local loads and usable storage devices, specifically at least one storage device state of charge and/or available power, are taken into account in the prediction and to minimize uncertainty.

[0145] One aim is to increase the forecast accuracy of the wind park power by changing a portfolio forecast to a wind-park-focused prediction.

[0146] For this purpose, the forecast model is subdivided into three individual models which can be separately validated, specifically, in particular, the weather model, the wind power model and the availability model.

[0147] It has been recognized that interconnecting the individual models which have been optimized for the wind park results in a more accurate wind-park-focused prediction than models which represent the entire uncertainty chain, in particular the wind forecast, location stability, follow-on model, technical availability and electrical losses, in a blurred or fused manner in a model and are optimized for a large portfolio, that is to say are aimed at remuneration and less at specific and local values.

[0148] It is proposed that a correction loop or a plurality of correction loops is/are incorporated in the forecast chain, specifically, in particular, extending from the weather data from the forecast block 532 to expected powers which have been determined, which correction loop(s) statistically correct(s) the model forecast, in particular using the improvement or correction means of smoothing, linear regression, a sliding average, dynamic weighting, bias correction, using measured data. Depending on the available data source, SCADA power data (for example also from adjacent wind parks), nacelle anemometer data, status codes or meteorological measurement stations are used. Accordingly, the correction with respect to wind data is applied to the weather models and, in the case of power data, is applied to the wind park model which may be a neural network, in particular.

[0149] According to one aspect, the accuracy of a prediction, in particular, is concomitantly taken into account. This is based on the following concepts.

[0150] In a network restoration situation, a current, that is to say newly calculated, prediction is no longer available under certain circumstances for regenerative power plants such as wind parks because computing centers are off-line, for example, or no longer have a data connection. After a few hours of power failure, the predictions which have already been calculated already have a considerable deviation from the actual available power and considerable uncertainty. However, in critical network situations, in addition to a prediction, the network operator also requires a quantification of the uncertainty in order to concomitantly take it into account in the control of the wind parks. The prediction and the uncertainty of the prediction must also be available at the wind park level and may even be calculated at the wind park level under certain circumstances.

[0151] However, a short-term prediction including uncertainty, provided via the network operator interface, can also be used for active network operational management of distribution networks or in accounting grid management.

[0152] The idea is, in particular, to quantify the prediction inaccuracy.

[0153] This applies, in particular, to wind parks on medium-voltage substations and high and extra-high voltages.

[0154] Improved operational management in critical network situations is also intended to be achieved, in particular.

[0155] The object can also be considered to be that of ensuring that the maximum possible power is always available depending on the network situation and prediction deviation which can be tolerated.

[0156] An apparatus or a method for holding and providing a prediction including prediction uncertainty at the wind park level is proposed, in particular. In this respect, the following steps or approaches are proposed: [0157] receiving a prediction, in particular with a probability distribution, at regular intervals, for example from prediction service providers; [0158] storing the prediction in a wind park storage device; and [0159] calculating a possible prediction deviation or a probability distribution on the basis of the weather situation, available installations, age of the prediction and/or number of available measurement sensors.

[0160] Alternatively: calculating and communicating a guaranteed minimum power on the basis of a minimum certainty to be predefined (e.g., 95%).

[0161] The following can be explained as an example: [0162] The network operator could specify, for example, via the network operator interface, the availability with which the network operator requires the minimum available power and the (short-term) prediction, or the network operator can directly request the uncertainty and can therefore itself consider this in the control.

[0163] The following are proposed as additional possibilities: [0164] Reducing the uncertainty by taking a plurality of wind parks into account;

[0165] considering available controllable loads and storage devices when calculating the uncertainty. This may mean the following, for example: in the case of a half-full state of charge of a storage device, half of the storage capacity is available for compensating for the prediction uncertainty.

[0166] Preliminary studies have shown that the uncertainties of each individual model can be reduced using a multi-model approach by means of additional information and manufacturer know-how relating to wind power installations and the wind park, which external forecast companies/traders do not have. The expected uncertainty of a wind park transfer model which has been created by means of a neural network may be −2.5% (on the basis of the quality and available historical power data).

[0167] The internal know-how relating to wind power installation technology reduces the uncertainty with respect to predicted technical turbine availabilities to 1%. The greatest uncertainty is in the weather forecast. This can be reduced both by means of intelligent weighting of different numerical weather models and by means of statistical corrections based on various measurement data in the wind park environment. It is therefore advantageous to combine these aspects.

[0168] The uncertainty funnel of the forecast can be reduced, on the one hand, by means of short-term corrections using measurement data and additionally or alternatively with coupled storage device solutions in the wind park. A combination of a plurality of wind parks running on a network operator interface can likewise reduce the uncertainty.

[0169] All three measures support the possibility of providing the network operators with a considerably higher P90 value, which describes the wind park power which occurs with a probability of 90%, for the forecast. This is illustrated in FIG. 5 which shows probability distributions, assuming an approximate normal distribution, for different adjustments. The ordinate shows the probability densities and the abscissa shows a normalized scale for the value of the examined result. The scale may be assumed here to be a normalized indication of a park power, in particular a rotor power, for example based on a maximum rotor power or a maximum sum of all rotor powers, that is to say without considering a limitation to nominal power, for example. In this case, the value marked 50 is the forecast power. The value may also represent an absolute value such as 50 MW, for example. This Figure is concerned, in particular, with the fundamental comparison of a plurality of prediction curves.

[0170] In this case, an adjustment, in particular by means of the correction rule, was carried out or simulated for a 1-hour prediction (1 h fit), for a 3-hour prediction (3 h fit) and for a 6-hour prediction (6 h fit), and a 1-hour distribution 501, a 3-hour distribution 503 and a 6-hour distribution 506 were accordingly represented. The P1, P10, P50, P90 and P99 positions were also marked for the 3-hour prediction. It can be seen, in particular, that the accuracy is increased in the short-term prediction adjustment and therefore in the short-term prediction. A P90 value is also indicated as “P90(1 h)” for the 1-hour prediction. The P90 value indicates the wind park power which at least occurs with a probability of 90%. This therefore corresponds to the surface area of the area below the respective curve to the right of the P90 value, based on the total area below the curve.

[0171] The area to the left of the P90 value accordingly has a surface area of 10%, and such an area is depicted as a remaining area 510 for the 3-hour prediction. This is improved with a shorter forecast, which can be seen from the high value of the 1-hour prediction at 50. This also results in the curve becoming narrower and the 10% remaining area ending further to the right, with the result that the P90 value is also further to the right and is therefore closer to the forecast power.

[0172] On the basis of FIG. 3, it can be illustrated that the uncertainty of the weather forecast from the weather service 532 may be approximately 6 to 8%. The uncertainty of the model which describes the wind park and can be illustrated using the parameter block 356 or using the wind power block 354, in which the calculation then takes place, may be approximately 2.5%. The technical availability which can be illustrated by the data block 36 or by the availability block is approximately 1%.

[0173] FIG. 4 explains a so-called uncertainty funnel. In this respect, FIG. 4 shows an uncertainty graph 400 in which forecast values are illustrated as a forecast curve over time with their uncertainty.

[0174] The time axis is divided into three regions for this purpose. The first time region 401 is not yet in the future, with the result that the prediction values correspond to the measured values, and so there is no uncertainty.

[0175] The second time region 402 is in the future, and there is therefore uncertainty which increases with time. The actual value can therefore be in the region of the uncertainty range shown and can accordingly deviate from the forecast curve 404.

[0176] The third time region 403 is yet further in the future, and a difference between a park uncertainty curve 405 and a weather uncertainty curve 406 can now be clearly distinguished. The park uncertainty curve 405 describes the uncertainty on account of the park inaccuracy and measurement uncertainties, whereas the weather uncertainty curve 406 describes an uncertainty on account of the weather forecast. The park uncertainty curve 405 forms a substantially smaller funnel than the weather uncertainty curve 406 in this case.