Wind state prediction system and wind state prediction method

Abstract

To provide a wind state prediction system and a wind state prediction method for predicting a wind state at a low cost, in realtime, and with high accuracy.SOLUTION: A wind state prediction system is equipped with a wind state analysis section, a linear mapping generation section, and a wind state estimation section. The wind state analysis section calculates secular changes of a three-dimensional flow field using LES (Large-Eddy-Simulation) calculation, on the basis of inflow wind information that is information of an inflow wind flowing into an object space and topographic information that is information of topography in the object space. The linear mapping generation section generates a linear mapping approximation model representing time evolution of the three-dimensional flow field in a selected space that is a part or the whole of a space lattice included in an attention space, wherein the attention space is a part of the object space. The wind state estimation section estimates secular changes of the three-dimensional flow field of the selected space using the Kalman filter, on the basis of an observation result of a wind state at the partial lattice of the selected space and the linear mapping approximation model.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a wind condition prediction system and a wind condition prediction method that predict wind conditions based on CFD analysis and actual measurements. [Background technology]

[0002] Japan's uniquely complex terrain generates orographic turbulence, typified by mountain waves, which poses a meteorological risk to the safety of air and rail transport. Orographic turbulence also poses a risk to the structural safety of fixed ground installations that operate year-round in mountainous regions, such as wind turbines, power transmission towers, and large antennas, in addition to mobile objects. Therefore, there is a need to develop and implement a wind forecasting system that enables highly accurate wind assessments in dangerous areas where turbulence frequently occurs, detailed hazard maps, high-resolution and frequent updates of turbulence warnings, and turbulence forecasts.

[0003] The local model currently in operation at the Japan Meteorological Agency is known as the regional numerical weather forecast model with the highest resolution. However, the horizontal mesh is 2km x 2km, so it can only resolve wind conditions in an area of ​​roughly 20km square. In line with the mesh size, forecast updates are also infrequent, at only every 30 minutes. This makes it difficult to resolve the impact of terrain-induced turbulence caused by the undulations of the Earth's surface on moving objects such as aircraft and trains, as well as ground-based structures in mountainous areas on a scale of several tens of meters, such as wind turbines, power transmission towers, and large antennas. As a result, it cannot be used to provide wind information useful for ensuring the safe landing of aircraft, etc.

[0004] Remote sensing instruments for wind observation include Doppler LIDAR and Doppler SODA. There are many types of instruments available, including those that output three wind speed components in the vertical direction only at intervals of several tens of meters, and those that only output the line-of-sight component but can measure areas by scanning. Each has limitations in spatial resolution and update frequency, and there are no all-weather instruments. The costs of system installation, year-round operation, and maintenance are in the hundreds of millions, making implementation at the local government level difficult. Japan's unique, undulating, and complex topography, with mountainous areas accounting for 70% of the country's land area, easily creates blind spots that cannot be covered.

[0005] It is possible to analyze the impact of actual terrain (Geospatial Information Authority of Japan digital maps) and buildings on wind using CFD (Computational Fluid Dynamics) calculations. This is effective for assessing wind conditions in mountainous or urban areas and creating turbulence hazard maps, and is particularly useful for evaluating the location of wind turbines. However, the number of grid points for outdoor terrain typically reaches the hundreds of millions, and even high-performance computers require several days to 10 days of calculation time. Setting boundary conditions is important for CFD calculations, but inflow conditions in particular are uncertain and constantly changing. Unless inflow winds can be measured and varied and large-scale calculations can be completed in a few minutes, real-time wind forecasts are impossible.

[0006] Meanwhile, technologies for predicting the probability of turbulence occurring have also been developed. For example, Patent Document 1 discloses a weather forecasting device that derives the probability of turbulence occurring based on input weather conditions. However, this weather forecasting device predicts the probability of turbulence occurring for each grid, and does not predict changes over time in the three wind speed components or changes over time in turbulence calculated from the three wind speed components (e.g., Q value). [Prior art documents] [Non-patent literature]

[0007] [Patent Document 1] Japanese Patent Publication No. 2020-176872 Summary of the Invention [Problem to be solved by the invention]

[0008] As mentioned above, while there is a need to visualize terrain-induced turbulence in various industrial fields, there is no technology to perform and predict terrain-induced turbulence in real time. In particular, if a technology for visualizing terrain-induced turbulence that does not require large observation equipment or high-performance computers were realized, it would be possible to operate it at low cost and it would be easy to introduce it at regional airports, etc.

[0009] In view of the above circumstances, an object of the present invention is to provide a wind condition prediction system and a wind condition prediction method that are capable of predicting wind conditions in real time at low cost and with high accuracy. [Means for solving the problem]

[0010] A wind condition prediction system according to an embodiment of the present invention includes a wind condition analysis unit, a linear mapping generation unit, and a wind condition estimation unit. The wind condition analysis unit calculates the time-dependent changes in the three-dimensional flow field using LES (Large-Eddy-Simulation) calculations based on inflow wind information, which is information on the inflow wind flowing into the target space, and topography information, which is information on the topography within the target space. When a part of the target space is set as a space of interest, the linear mapping generation unit generates a linear mapping approximation model that represents the time evolution of a three-dimensional flow field in a selected space that is a part or all of the space of interest. The wind condition estimation unit estimates a time-dependent change in a three-dimensional flow field in the selected space using a Kalman filter based on observation results of wind conditions in a partial grid in the selected space and the linear mapping approximation model.

[0011] a state variable reduction unit that generates, from the linear mapping approximation model, a linear mapping approximation model in which state variables are reduced, the linear mapping approximation model representing time evolution of a three-dimensional flow field in an important space that is a part of the selected space; The wind condition estimation unit may estimate a three-dimensional flow field in the selected space based on the observation results and the linear mapping approximation model in which the state variables are reduced.

[0012] The linear mapping generator may generate the linear mapping approximation model that describes the time evolution of a data set of a three-dimensional flow field in the selected space.

[0013] The linear mapping generation unit may generate the linear mapping approximation model using a system matrix that is averagely fitted between multiple data sets of flow fields calculated by the wind condition analysis unit for multiple inflow conditions for the same wind direction.

[0014] The linear mapping generation unit may generate the linear mapping approximation model using a first system matrix that is averagely fitted between multiple data sets of flow fields calculated by the wind condition analysis unit for a first wind direction, a second system matrix that is averagely fitted between multiple data sets of flow fields calculated by the wind condition analysis unit for a second wind direction, and parameters that connect the first system matrix and the second system matrix.

[0015] The wind condition estimation unit may detect terrain-induced turbulence from the estimated flow field.

[0016] A wind condition prediction system according to an embodiment of the present invention includes a wind condition analysis unit and a linear mapping generation unit. The wind condition analysis unit calculates the time-dependent changes in the three-dimensional flow field using LES (Large-Eddy-Simulation) calculations based on inflow wind information, which is information on the inflow wind flowing into the target space, and topography information, which is information on the topography within the target space. When a part of the target space is set as a space of interest, the linear mapping generation unit generates a linear mapping approximation model that represents the time evolution of a flow field in a selected space that is part or all of the space of interest.

[0017] A wind condition prediction system according to an embodiment of the present invention includes a linear mapping generation unit. The linear mapping generation unit generates a linear mapping approximation model that represents the time evolution of the flow field in a selected space, which is part or all of the target space, when a part of the target space is designated as a focus space, for the time-dependent change of the three-dimensional flow field calculated by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information on inflow wind flowing into the target space, and topography information, which is information on the topography within the target space.

[0018] A wind condition prediction system according to an embodiment of the present invention includes a wind condition estimation unit. The wind condition estimation unit estimates the three-dimensional flow field of a selected space, which is part or all of the target space, using a Kalman filter based on a linear mapping approximation model that represents the time evolution of the flow field in the selected space, which is part or all of the target space, and the observation results of the wind conditions in a partial grid of the selected space, regarding the time-dependent change of the three-dimensional flow field calculated by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information on the inflow wind flowing into the target space, and topography information, which is information on the topography within the target space.

[0019] A wind condition prediction system according to an embodiment of the present invention includes a wind condition analysis unit, a linear mapping generation unit, an observation device, and a wind condition estimation unit. The wind condition analysis unit calculates the time-dependent changes in the three-dimensional flow field using LES (Large-Eddy-Simulation) calculations based on inflow wind information, which is information on the inflow wind flowing into the target space, and topography information, which is information on the topography within the target space. When a part of the target space is set as a space of interest, the linear mapping generation unit generates a linear mapping approximation model that represents the time evolution of a flow field in a selected space that is part or all of the space of interest. The observation equipment observes wind conditions in a partial grid of the selected space. The wind condition estimation unit estimates a time-dependent change in the three-dimensional flow field in the selected space using a Kalman filter based on the observation results and the linear mapping approximation model.

[0020] The observation equipment may observe the wind conditions by spot observation.

[0021] A wind condition prediction method according to one aspect of the present invention includes calculating a time-dependent change in a three-dimensional flow field by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information about inflow winds flowing into a target space, and topography information, which is information about the topography within the target space; When a part of the target space is designated as a space of interest, a linear mapping approximation model is generated that represents the time evolution of a flow field in a selected space that is a part or all of the space of interest; Based on the observation results at a spatial grid in part of the selected space and the linear mapping approximation model, a Kalman filter is used to estimate the time-dependent change in the three-dimensional flow field in the selected space. [Effects of the Invention]

[0022] According to the present invention, it is possible to provide a wind condition prediction system and a wind condition prediction method that are capable of predicting wind conditions in real time with high accuracy at low cost. [Brief explanation of the drawings]

[0023] [Figure 1] 1 is a block diagram of a wind condition prediction system according to an embodiment of the present invention. [Figure 2] 4 is a flowchart showing the operation of the wind condition prediction system. [Figure 3] FIG. 2 is a schematic diagram of a target space related to the wind condition prediction system. [Figure 4] FIG. 2 is a schematic diagram of a spatial grid provided in the target space. [Figure 5] 10 is a graph showing a height direction profile of wind conditions set in a wind condition analysis unit included in the wind condition prediction system. [Figure 6] FIG. 2 is a schematic diagram of a space of interest in the target space. [Figure 7] FIG. 2 is a schematic diagram of a data set generated by a linear mapping generation unit included in the wind condition prediction system. [Figure 8] 1 is a schematic diagram showing the generation of a data matrix by the linear mapping generation unit and an example of a system matrix calculation (Equation 3). [Figure 9] 1 shows a schematic diagram of multiple data sets generated by the linear mapping generator and the relational equation (Equation 4) between the system matrices. [Figure 10] FIG. 1 is a schematic diagram showing a common grid (circular portion) for different wind direction conditions. DETAILED DESCRIPTION OF THE INVENTION

[0024] A wind condition prediction system according to an embodiment of the present invention will be described.

[0025] [Configuration and operation of wind forecasting system] FIG. 1 is a block diagram showing the configuration of a wind condition prediction system 100 according to this embodiment, and FIG. 2 is a flowchart showing the operation of the wind condition prediction system 100. As shown in FIG. 1, the wind condition prediction system 100 comprises a server 110, a terminal 120, and observation equipment 130. The server 110 is preferably an information processing device with high processing power, such as a workstation. The terminal 120 can be a general information processing device, such as a PC (personal computer). The observation equipment 130 observes actual wind conditions and supplies the observation results to the terminal 120.

[0026] 1, the server 110 includes a wind condition analysis unit 111, a spatial grid selection unit 112, a linear mapping generation unit 113, and a state variable reduction unit 114. The terminal 120 also includes a wind condition estimation unit 121. These are functional configurations realized by the cooperation of hardware and software of an information processing device that includes a CPU (Central Processing Unit), RAM (Random Access Memory), etc. The software is a program that can be executed by the information processing device, and may be a program recorded on a recording medium that can be read by the information processing device.

[0027] The wind condition analysis unit 111 performs LES (Large-Eddy-Simulation) calculations for the target space. FIG. 3 is a schematic diagram showing the target space S. The target space S is a space designated by the user for which the wind condition prediction system 100 predicts wind conditions, and is, for example, a space where terrain-induced turbulence is a problem. In FIG. 3, the target space S is shown as an example of a space including an aircraft's approach path to airport K. In addition, the target space S can be a space where terrain-induced turbulence is a problem, such as a space around a fixed ground installation such as a wind turbine for wind power generation or a large antenna. The size of the target space S is not particularly limited, but is, for example, several kilometers to several tens of kilometers square in the horizontal direction and several kilometers in the vertical direction.

[0028] FIG. 4 is a schematic diagram showing a spatial grid G ​​provided in the target space S. As shown in the figure, the wind condition analysis unit 111 divides the target space S into a large number of spatial grids G. The size of the spatial grid G ​​is not particularly limited, but for example, one side is several meters to several tens of meters in the horizontal and vertical directions. The size of the spatial grid G ​​may be set in advance or may be specified by the user.

[0029] In addition, the wind condition analysis unit 111 acquires information on the inflow wind flowing into the target space S (hereinafter referred to as "inflow wind information"). The inflow wind information includes the direction, wind speed, and altitude profile of the inflow wind. The wind condition analysis unit 111 can acquire inflow wind information set by the user. The user can grasp the direction and flow speed of the wind flowing into the target space S in a particular season by using a wind rose or the like around the target space S. For example, if it is desired to predict terrain-induced turbulence occurring around airport K in winter, and it is assumed that westerly winds blow around airport K in winter, the user inputs information on the westerly wind as inflow wind information.

[0030] Figure 5 shows an example of an elevation profile that constitutes inflow wind information. As shown in the figure, the elevation profile is set as a power-type elevation profile with a power exponent as a parameter. However, the actual elevation profile fluctuates over time and is uncertain. The difference between the actual elevation profile and the setting is corrected in the processing described below.

[0031] Furthermore, the wind condition analysis unit 111 acquires information on the topography within the target space S (hereinafter referred to as "topography information"). The topography information is, for example, the elevation of each point within the target space S. The wind condition analysis unit 111 can acquire the topography information from a digital map of the Geospatial Information Authority of Japan or the like.

[0032] The wind condition analysis unit 111 performs LES (Large-Eddy Simulation) calculations based on inflow wind information and topographical information (Figure 2, St1). In the LES calculations, the wind flow field (wind condition field) in the target space S is decomposed into large flow fields that can be captured by the spatial grid G ​​and flow fields smaller than the spatial grid G. The large flow fields are calculated directly, and the small flow fields are modeled and subjected to unsteady calculations. The wind condition analysis unit 111 analyzes the wind conditions in the target space S using LES calculations and calculates the time-varying changes in the three wind speed components (three-dimensional flow field) in each spatial grid G. Hereinafter, the time-varying changes in the three wind speed components in all spatial grids G included in the target space S and the time-varying changes in turbulence calculated from the wind speed field are referred to as "spatiotemporal high-density time series data." The wind condition analysis unit 111 supplies the generated spatiotemporal high-density time series data to the linear mapping generation unit 113.

[0033] The spatial lattice selector 112 selects a part of the spatial lattice G as a "selected lattice" and downsamples the spatial lattice (FIG. 2, St2). a 1 is a schematic diagram showing the selection of a target space S. As shown in the figure, the user can specify a part of the target space S as a target space T. The target space T is a space in the target space S where the occurrence of terrain-induced turbulence is predicted, and its shape is not particularly limited. The user can specify the target space T by enclosing the space or area where the occurrence of terrain-induced turbulence is predicted with a graphic, or by inputting the coordinates of the perimeter, or by other operations.

[0034] The spatial grid selection unit 112 selects all of the spatial grids G included in the space of interest T as the selected grids G a The spatial lattice selection unit 112 extracts spatial lattices G included in the attention space T at predetermined intervals and selects the selected lattice G a It is also possible to set the selected grid G a must include the observation point of the observation equipment 130. The spatial grid selection unit 112 selects the selected grid G ​​according to the processing capacity of the server 110. a The spatial grid selection unit 112 can adjust the number of selected grids G a The linear mapping generator 113 is supplied with information indicating the above.

[0035] The linear mapping generator 113 generates a linear mapping approximation model that represents the time evolution of the flow field in the selected space. The selected space is a part or all of the attention space T, and the selected grid G a The linear mapping generator 113 generates a linear map for each selected grid G a The flow field in the selected grid G ​​is extracted to form a data set. FIG. 7 is a schematic diagram of this data set. a For the flow field in the , determine the system matrix A that satisfies the system equation shown in Equation 1 below. The system matrix A is a matrix that represents the time evolution of this data set. Note that x k indicates the state variables at time k (turbulence calculated from the three wind speed components or wind speed gradient), and x k+1 denotes the state variables at time k+1.

[0036] In addition, in the data set shown in FIG. 7, the selected grid G ​​where the wind conditions are actually observed by the observation equipment 130 as described later is a The observation grid G b The observation equation for this data set is expressed as the following equation 2. k indicates the observation value at time k, and the observation matrix C is the observation lattice G b Indicates the location of.

[0037] x k+1 =Ax k + Noise ... (Equation 1) y k =Cx k + Noise ... (Equation 2) (C,A): Observable

[0038] Figure 8 shows the selection grid G a As shown in the figure, the linear mapping generator 113 generates a data matrix of a flow field in each selected grid G aThe time series of is placed in each row to generate a data matrix. From this data matrix, Equation 3 shown in Figure 8 is obtained. However, Equation 3 cannot express data sets for other cases where the inflow conditions are changed, which causes the adverse effect of overfitting, where the data is overfitted to only a single inflow condition, and the degrees of freedom of the system matrix A are not utilized.

[0039] Therefore, the linear mapping generation unit 113 determines a system matrix A that can fit multiple data sets generated for the same wind direction on average (St3 in Fig. 2). The multiple data sets generated for the same wind direction are composed of flow fields calculated by the wind analysis unit 111 for a specific target space S by changing conditions such as wind speed and altitude profile.

[0040] FIG. 9 is a schematic diagram showing the calculation of the system matrix A from the data matrices of multiple data sets. As shown in the figure, Equation 4 is obtained using the data matrices of multiple data sets (data sets (1) to (N) in the figure), and the system matrix A is found from this equation. By using multiple data sets generated for the same wind direction, it is possible to averagely represent multiple cases using a single nominal model, prevent overlearning, and ensure the versatility of the system matrix A. In this way, the linear mapping generation unit 113 generates a linear mapping approximation model using the system matrix A. The linear mapping generation unit 113 supplies the generated linear mapping approximation model to the state variable reduction unit 114.

[0041] The state variable reduction unit 114 reduces the number of state variables to reduce their dimensions (FIG. 2, St4). By performing the above steps (St1 to St3), a linear mapping approximation model that expresses the time evolution of the flow field can be generated. However, even at this stage, the dimensions of the state variables (number of spatial grids) are expected to be on the order of thousands to tens of thousands, so further reduction in dimensions is required.

[0042] Specifically, the state variable reduction unit 114 reduces the selected lattice G a rearrange the important lattice group Φ (1) and the non-important lattice group Φ (2)The rows and columns of the corresponding system matrix A are also swapped, so the new swapped matrix is ​​Ω=[Ω ij ] is written as follows.

[0043]

number

[0044] From the above equation 5, the system matrix A of only the important lattice groups is r can be calculated as in the following equation 6.

[0045]

number

[0046] In this way, the state variable reduction unit 114 reduces the system matrix A r In other words, the state variable reduction unit 114 generates a linear mapping approximation model in which the state variables are reduced, which represents the time evolution of the three-dimensional flow field in the important space, which is a set of important lattice groups. The state variable reduction unit 114 supplies the generated linear mapping approximation model to the wind condition estimation unit 121. Note that the reduction of state variables by the state variable reduction unit 114 depends on the calculation processing capacity of the terminal 120 and the selected lattice G a In this case, the server 110 can supply the linear mapping approximation model generated by the linear mapping generation unit 113 to the wind condition estimation unit 121 instead of the linear mapping approximation model generated by the state variable reduction unit 114.

[0047] The wind condition estimation unit 121 estimates the observation grid G b The time-dependent change of the three-dimensional flow field in the selected space is estimated based on the observation results of the wind conditions at the observation grid G ​​and the linear mapping approximation model (Fig. 2, St5). b The observation results of wind conditions at observation grid G bThe wind condition estimation unit 121 calculates the wind speed by converting the observation results of the observation equipment 130 into the observed value y k Then, the state variable x k The Kalman filter can be a standard linear Kalman filter.

[0048] Here, the above system matrix A is a nominal characteristic that expresses the average dynamic characteristics of the flow field, and it is inevitable that it will involve errors. The following equations 7 and 8 are the system equation and observation equation that include the error matrix ΔA. By evaluating the magnitude (norm) of the error matrix ΔA, the wind condition estimation unit 121 can design a Kalman gain that makes the Kalman filter robust (does not cause instability).

[0049] x k+1 =(A+ΔA)x k + noise ... (Equation 7) y k =Cx k + noise ... (Equation 8) (C,A): Observable

[0050] In this way, the wind condition estimation unit 121 selects the grid G a The state variable x k That is, the flow field of each spatial grid G ​​included in the space of interest T (see FIG. 6) can be estimated. The wind condition estimation unit 121 can visualize the estimated flow field and present it to the user. The wind condition estimation unit 121 may also detect terrain-induced turbulence in the estimated flow field and present the detected terrain-induced turbulence to the user.

[0051] Observation equipment 130 is observation grid G b and the observation grid G bThe observation equipment 130 measures wind conditions, i.e., wind direction and wind speed. The observation equipment 130 may be a device capable of spot observation, such as a wind vane and wind speed meter that is standardly installed at airports, and there is no need to use expensive equipment capable of observing a wide airspace, such as a Doppler LIDAR or Doppler soda. The observation equipment 130 supplies the observation results to the wind condition estimation unit 121. There is no particular limitation on the number of observation equipment 130, as long as there is one or more. An anemometer already installed at the airport may also be used.

[0052] The wind condition prediction system 100 has the above-described configuration. The wind condition prediction system 100 may be any system capable of realizing the above-described functional configuration, and its hardware configuration is not limited to the server 110 and the terminal 120. For example, the functional configuration of the server 110 may be realized by the cooperation of multiple information processing devices. Furthermore, all of the functional configuration of the wind condition prediction system 100 may be realized by a single information processing device. The terminal 120 may be an information processing device such as a laptop PC, or may be a device installed in an airport control tower or on board an aircraft.

[0053] [Effects of the wind forecast system] As described above, in the wind condition prediction system 100, the server 110 generates a linear mapping approximation model, and the terminal 120 estimates a flow field based on this linear mapping approximation model and the observation results from the observation equipment 130. The LES calculations required to generate the linear mapping approximation model require a large amount of calculation, but the server 110 can perform the LES calculations in advance to generate the linear mapping approximation model. This allows the terminal 120 to estimate a flow field from the linear mapping approximation model and the observation results with a small amount of calculation, making it possible to estimate a flow field in real time with high accuracy even on a general information processing device.

[0054] The observation equipment 130 can be low-cost observation equipment for spot observations, rather than large, expensive equipment costing hundreds of millions of yen, and since it is not necessary to deploy a large number of such equipment, it can be realized at low cost. Existing observation equipment can also be used.

[0055] [Generation of linear mapping approximation models using different wind direction conditions] In the above description, the linear mapping generator 113 determines the system matrix A for the same wind direction condition, but it is also possible to determine the system matrix A for different wind direction conditions. Specifically, the linear mapping generator 113 can determine a system matrix A that is fitted on average between a plurality of data sets generated under different wind direction conditions.

[0056] The multiple data sets generated under different wind direction conditions are composed of time series of vibration characteristics calculated by the wind analysis unit 111 for different target spaces S. Fig. 10 is a schematic diagram showing the calculation of the system matrix A from the data matrices of the multiple data sets generated under different wind direction conditions.

[0057] As shown in the figure, the system matrix that is averagely fitted between a plurality of data sets based on the vibration characteristics calculated for the first target space S1 is called the first system matrix A. (1) In addition, a system matrix that is averagely fitted between a plurality of data sets based on vibration characteristics calculated for a second target space S2 different from the first target space S1 is defined as a second system matrix A (2) Let's say.

[0058] The first target space S1 and the second target space S2 have different wind directions flowing into each target space. When the wind direction flowing into the target space changes by more than a certain amount, the dynamics of the flow field also changes, and it becomes difficult to express its time evolution with a single linear model. For this reason, the endpoint matrix A (1) , A (2) It is necessary to introduce a parameter θ that connects them and expand the expressive capabilities of the model. The system equation for multiple data sets generated from different dynamics is expressed in the following equation 9, and the observation equation is expressed in the following equation 10. The system equation is expressed using the system matrix A(θ) that changes depending on the parameter θ. Such a linear system is called a polytopic linear system.

[0059] x k+1 ={θ1A (1) +θ2A(2)}x k +Noise =A(θ)x k + noise ... (Equation 9) y k =Cx k + noise ... (Equation 10) (C,A(θ)): Observable

[0060] The linear mapping generation unit 113 supplies a linear mapping approximation model based on the system matrix A(θ) to the state variable reduction unit 114, and the state variable reduction unit 114 supplies the linear mapping approximation model, which has been reduced in dimension by reducing the state variables, to the wind condition estimation unit 121. Alternatively, if there is no need to reduce the dimension of the linear mapping approximation model, the linear mapping approximation model generated by the linear mapping generation unit 113 is supplied to the wind condition estimation unit 121.

[0061] The wind condition estimation unit 121 uses the observation grid G b Based on the observation results of wind conditions in the area and the linear mapping approximation model, the selected grid G a The wind condition estimation unit 121 estimates the vibration characteristics of the wind speed by converting the observation result of the observation equipment 130 into the observation value y k Then, the state variable x k and parameter θ. The wind condition estimation unit 121 estimates the state variable x k Since it is necessary to simultaneously estimate both the parameter θ and the parameter θ, a nonlinear Kalman filter such as an extended Kalman filter must be used.

[0062] The above system matrix A(θ) is a nominal characteristic that expresses the average dynamic characteristics of the flow field, and it is inevitable that it will contain errors. The following equations 11 and 12 are the system equation and observation equation that include the error matrix ΔA. By evaluating the magnitude (norm) of the error matrix ΔA, the wind condition estimation unit 121 can design a Kalman gain that makes the Kalman filter robust (does not cause instability).

[0063] x k+1 =(A(θ)+ΔA)x k + noise ... (Equation 11) y k =Cxk + noise ... (Equation 12) (C,A(θ)): Observable

[0064] In this way, the wind condition estimation unit 121 estimates the selected grid G a The state variable x k That is, the vibration characteristics of each spatial grid G ​​included in the space of interest T (see FIG. 6) can be calculated, and the flow field in the space of interest T can be estimated.

[0065] Although the embodiment of the present invention has been described above, the present invention is not limited to this embodiment, and various modifications can be made, as a matter of course. [Explanation of symbols]

[0066] 100...Wind forecast system 110...Server 111...Wind condition analysis department 112...Spatial grid selection unit 113...Linear mapping generator 114...State variable reduction section 120...Terminal 121...Wind Condition Estimation Section 130...Observation equipment

Claims

1. a wind condition analysis unit that calculates time-dependent changes in a three-dimensional flow field by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information on inflow winds flowing into a target space, and topography information, which is information on the topography within the target space; and a linear mapping generation unit that generates a linear mapping approximation model that represents time evolution of a three-dimensional flow field in a selected space that is a part or all of the target space, where the selected space is a space of interest; a wind condition estimation unit that estimates a time-dependent change in a three-dimensional flow field in the selected space using a Kalman filter based on the observation results of wind conditions in a partial grid of the selected space and the linear mapping approximation model; A wind forecasting system equipped with the above.

2. The wind condition prediction system according to claim 1, a state variable reduction unit that generates, from the linear mapping approximation model, a linear mapping approximation model in which state variables are reduced, the linear mapping approximation model representing time evolution of a three-dimensional flow field in an important space that is a part of the selected space; The wind condition estimation unit estimates a time-dependent change in the three-dimensional flow field in the selected space based on the observation results and the linear mapping approximation model in which the state variables are reduced. Wind forecasting system.

3. The wind condition prediction system according to claim 1, The linear mapping generator generates the linear mapping approximation model that describes the time evolution of a data set of a flow field in the selected space. Wind forecasting system.

4. The wind condition prediction system according to claim 3, The linear mapping generation unit generates the linear mapping approximation model using a system matrix that is averagely fitted between a plurality of data sets of flow fields calculated by the wind condition analysis unit for a plurality of inflow conditions for the same wind direction. Wind forecasting system.

5. The wind condition prediction system according to claim 4, The linear mapping generation unit generates the linear mapping approximation model using a first system matrix that is averagely fitted between a plurality of data sets of flow fields calculated by the wind condition analysis unit for a first wind direction, a second system matrix that is averagely fitted between a plurality of data sets of flow fields calculated by the wind condition analysis unit for a second wind direction, and a parameter that connects the first system matrix and the second system matrix. Wind forecasting system.

6. A wind condition prediction system according to claim 1, The wind condition estimation unit detects terrain-induced turbulence from the estimated flow field. Wind forecasting system.

7. a wind condition analysis unit that calculates time-dependent changes in a three-dimensional flow field by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information on inflow winds flowing into a target space, and topography information, which is information on the topography within the target space; and a linear mapping generation unit that generates a linear mapping approximation model that represents the time evolution of a flow field in a selected space that is a part or all of the target space, where the selected space is a space of interest; A wind forecasting system equipped with the above.

8. a linear mapping generation unit that generates a linear mapping approximation model that represents the time evolution of a three-dimensional flow field in a selected space, which is a part or all of a space of interest, when a part of the target space is designated as a space of interest, for a time-varying three-dimensional flow field calculated by a large-eddy-simulation (LES) calculation based on inflow wind information that is information on inflow winds flowing into a target space and topography information that is information on topography within the target space; A wind forecasting system equipped with the above.

9. a wind condition estimation unit that, regarding time-varying changes in a three-dimensional flow field calculated by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information on inflow winds flowing into a target space, and topography information, which is information on the topography within the target space, estimates the three-dimensional flow field of the selected space by a Kalman filter, based on a linear mapping approximation model that represents the time evolution of the flow field in a selected space, which is part or all of the target space, and observation results of wind conditions in a partial grid of the selected space, when a part of the target space is taken as a space of interest; A wind forecasting system equipped with the above.

10. a wind condition analysis unit that calculates time-dependent changes in a three-dimensional flow field by LES (Large-Eddy-Simulation) calculation based on inflow wind information, which is information on inflow winds flowing into a target space, and topography information, which is information on the topography within the target space; and a linear mapping generator that generates a linear mapping approximation model that represents time evolution of a flow field in a selected space that is a part or all of the target space, where the selected space is a space of interest; an observation device that observes wind conditions in a partial grid of the selected space; a wind condition estimation unit that estimates a three-dimensional flow field in the selected space using a Kalman filter based on the observation results and the linear mapping approximation model; A wind forecasting system equipped with the above.

11. A wind condition prediction system according to claim 10, The observation equipment observes the wind conditions by spot observation. Wind forecasting system.

12. Calculating time-dependent changes in a three-dimensional flow field using LES (Large-Eddy-Simulation) calculations based on inflow wind information, which is information on inflow winds flowing into a target space, and topography information, which is information on the topography within the target space; When a part of the target space is designated as a space of interest, a linear mapping approximation model is generated that represents the time evolution of a three-dimensional flow field in a selected space that is a part or all of the space of interest; A time-dependent change in the three-dimensional flow field in the selected space is estimated using a Kalman filter based on the observation results in a partial grid of the selected space and the linear mapping approximation model. Wind forecasting methods.

Images