Method and system for generating initial field of hybrid hot start in regional numerical forecast model
Through the hybrid hot-start initial field generation method, combined with Fourier decomposition and low-pass filter technology, the global mode and regional numerical forecast mode are separated and merged, solving the problem of error accumulation in long-term integral solution of regional numerical forecast mode. Fast response and efficient forecasting are achieved.
Patent Information
- Application Number
- CN202210198456.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-02
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The model error, initial value error and boundary error accumulated in the existing regional numerical forecasting mode during long-term integral solution causes the weather system to be distorted, which cannot effectively reflect large-scale weather movements. The cold start method requires a large amount of computing resources and is not suitable for many Changes in weather.
Using the hybrid hot-start initial field generation method, the global mode and regional numerical forecast mode are separated by Fourier decomposition or low-pass filter to extract the large-scale information of the global mode and the small-scale information of the regional numerical forecast mode. Combination is carried out to generate a hot-start initial field, calculate the kinetic energy spectrum and its error distribution in real time, and control the residual differential kinetic energy to maintain stability.
While not affecting the stability of regional numerical forecasting mode, global mode information is quickly introduced to reduce errors, improve the forecasting effect of meteorological factors, and achieve rapid response to changing weather.
Smart Images

Figure CN115146817B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of weather forecasting, and in particular to a method and system for generating a hybrid hot start initial field of a regional numerical forecasting model. Background Art
[0002] Refined spatiotemporal weather forecasts can provide fixed-point, regular, and quantitative meteorological information for civil aviation, logistics, clean energy, water affairs, urban management, and other departments. The method used for refined spatiotemporal weather forecasts is the regional numerical forecast model, that is, in the preset forecast area, a mathematical model of the atmospheric thermal dynamic field changing over time is constructed, and then supercomputing technology is used to perform large-scale numerical simulation integration and solve the constructed mathematical model. At present, the horizontal resolution of the regional numerical forecast model can reach the kilometer to 100-meter level, and the time resolution of the forecast can reach the hour to minute level.
[0003] The inherent requirement of spatiotemporal refined weather forecasting is to continuously improve the accuracy of regional numerical forecast models. However, in the process of long-term integration to solve regional numerical forecast models, model errors, initial value errors, boundary errors and numerical calculation errors will gradually accumulate with the integration time, causing the weather system simulated by the model to be distorted. One of the important sources of error is the accumulated error caused by the introduction of lateral boundary conditions containing errors in the commonly used Newton scaling scheme. On the one hand, this error directly affects the development of the predicted pressure field and temperature field. On a short time scale, this error has little effect on small and medium-scale weather systems, but it has a great distortion on large-scale weather movements, making the regional numerical forecast model unable to effectively reflect the accurate position of the weather system, causing the simulation system to drift; on the other hand, it also leads to deviations in the predicted water vapor distribution, which in turn has a certain degree of impact on the physical process of the regional numerical forecast model. After a long period of numerical integration, this error can further transfer energy between scales, ultimately leading to a rapid reduction in the forecasting ability of the regional numerical forecast model for weather systems of various scales.
[0004] Compared with the regional numerical prediction model, the global model has a lower simulation resolution (tens to hundreds of kilometers), which makes it unable to accurately reflect the changes in the kilometer-level weather system; however, since the global model simulates the atmospheric state of the entire earth's spherical surface and there is no lateral boundary, its simulation accuracy of large-scale weather systems is better than that of the regional numerical prediction model. In order to correct this shortcoming of the regional numerical prediction model, the regional numerical prediction model needs to continuously introduce large-scale information of the global model forecast meteorological field to correct its own lateral boundary error, which is a key technical issue in the current spatiotemporal refined weather forecast.
[0005] The regional numerical forecast model can extract the area of interest from the global model as the initial field data to start running. This method is called cold start. At present, in order to introduce large-scale information from the global model, the traditional idea is to adopt an intermittent cold start method, that is, to use the latest global model forecast field as the initial field to restart the entire regional numerical forecast model forecast system at regular intervals, thereby indirectly introducing large-scale information. This method is easy to implement and only requires minor modifications to the forecast system process. Various regional weather forecast systems in the United States, Europe, and China all operate in an intermittent cold start mode. This method generally restarts the system every 24 hours or 12 hours, and the large-scale fields of the global model are not introduced at the remaining forecast times, so it is also called a partial cold start. Its main process is as follows Figure 1 shown.
[0006] However, the cold start method of regional numerical forecast models also brings some problems:
[0007] 1. The resolution of the global model is generally tens to hundreds of kilometers, which is much lower than that of the regional numerical forecast model. This means that the regional numerical forecast model weather forecast system must run for a long time during the cold start phase to regenerate the small and medium-scale system so that the regional numerical forecast model can regain stability. This process requires the use of a large amount of additional computing resources, resulting in a large waste of computing resources during the cold start phase of the regional numerical forecast model, and it is impossible to achieve a short time interval cold start. The update speed of the regional numerical forecast model forecast initial field cannot adapt to the changing weather changes;
[0008] 2. During the cold start process, the entire small and medium-scale weather system in the regional numerical forecast model is rebuilt, which also leads to poor continuity of the simulated weather system before and after the cold start in the region;
[0009] 3. The time of cold start also needs to be kept in a period of relatively stable weather. However, in some areas, such as tropical and subtropical regions with strong convection, small- and medium-scale weather developments may occur at any time, so it is difficult to determine a stable weather time for cold start.
[0010] Therefore, how to introduce sufficient global model information to achieve hot start of regional numerical prediction models without affecting the stability of regional numerical prediction models is an urgent problem to be solved in the field of spatiotemporal refined weather forecasting. Summary of the invention
[0011] The present invention aims to solve at least one of the technical problems existing in the prior art, and in particular innovatively proposes a method and system for generating a hybrid hot start initial field of a regional numerical forecast model, which can introduce sufficient global model information to achieve hot start of the regional numerical forecast model without affecting the stability of the regional numerical forecast model.
[0012] In order to achieve the above-mentioned object of the present invention, according to a first aspect of the present invention, the present invention provides a method for generating a hybrid hot start initial field of a regional numerical forecast model, comprising the following steps:
[0013] S1, the minimum credible scale l of the global model resolution is calculated based on the effective resolution of the forecast wind field of the global model min ;
[0014] S2, based on the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max The east-west or north-south width of the target area to be forecasted;
[0015] S3, based on the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) of the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model;
[0016] S4, the minimum credible scale l min The largest of the three, namely, the kinetic error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model respectively, and the two are added together to generate the initial field of the hot start of the regional numerical prediction model at the current time.
[0017] Preferably, step S2 comprises:
[0018] S21, the forecast wind field W of the global model at the minimum credible scale l min The target area scale l to the target area that needs forecasting max Fourier decomposition or low-pass filter is used to separate the scales at each scale l between the two, and the large-scale wind speed spectrum distribution W corresponding to the forecast wind field W of the global model at each scale l is obtained. large (l) and the small-scale wind speed spectrum distribution W small (l), the scale separation calculation formula is as follows:
[0019] W=W large (l)+W small (l)
[0020] in,
[0021] S22, Analyze wind field in target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Discrete Fourier decomposition or low-pass filter is used to separate the scales at each scale l to obtain the analytical wind field of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0022]
[0023] in,
[0024] S23, Large-scale wind speed spectrum distribution W of the forecast wind field W based on the global model large (l) and the analytical wind field of the target area Large-scale wind speed spectrum distribution Calculate the forecast wind field W of the global model and the analytical wind field of the target area at each scale l The difference kinetic energy spectrum A(l) is calculated as follows:
[0025]
[0026] S24, based on the difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l), a kinetic energy error tolerance l1 is obtained, and the calculation formula is as follows:
[0027] A(l)=∈1
[0028] Among them, the scale l that satisfies the above formula is the kinetic energy error tolerance scale corresponding to the first difference tolerance ∈1, denoted as l1.
[0029] Preferably, step S3 comprises:
[0030] S31, calculate the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model after the last startup The wind field error D between them is as follows:
[0031]
[0032] S32, the forecast wind field of the regional numerical forecast model after the last startup of the target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting maxScale separation is performed on each scale l between the two to obtain the forecast wind field of the regional numerical forecast model at the last startup of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0033]
[0034] in,
[0035] S33, based on the forecast wind field W of the global model obtained in S21, the corresponding large-scale wind speed spectrum distribution W at each scale l large (l) and the forecast wind field of the regional numerical forecast model at the last startup of the target area obtained by S32 The corresponding large-scale wind speed spectrum distribution at each scale l Calculate the large-scale wind field error d(l) corresponding to each scale l, and the calculation formula is as follows:
[0036]
[0037] S34, based on the forecast wind field W of the global model of the target area obtained in S31 and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) at the small-scale end is calculated by using the wind field error D between the two and the large-scale wind field error d(l) corresponding to each scale l obtained by S33. The calculation formula is as follows:
[0038]
[0039] S35, based on the residual difference kinetic energy B(l) at the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) at the small-scale end, the regional numerical prediction model residual scale l2 is obtained, and the calculation formula is as follows:
[0040] B(l)=∈2
[0041] Among them, the scale l that satisfies the above formula is the residual scale of the regional numerical prediction model corresponding to the second difference tolerance ∈2, denoted as l2.
[0042] Preferably, step S4 comprises:
[0043] S41, based on the minimum credible scale l min , kinetic energy error tolerance scale l1 and regional numerical prediction model residual scale l2 determine the dynamic hybrid scale l b , the formula is as follows:
[0044] l b=max(l min ,l1,l2);
[0045] S42, the dynamic mixing scale l b As a separation scale and using discrete Fourier decomposition or low-pass filter to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model respectively;
[0046] S43, adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model to obtain a mixed field as the initial field for hot starting of the regional numerical forecast model at the current time.
[0047] Preferably, the effective resolution in step S1 refers to the distribution profile K(l) of the high-altitude wind field kinetic energy spectrum K with scale l starting to be significantly lower than the standard -5 / 3 power law distribution The calculation formula is as follows:
[0048]
[0049] And for any l<l min
[0050]
[0051] Where n is the height and η is the specified small kinetic energy difference.
[0052] According to a second aspect of the present invention, the present invention provides a hybrid hot start initial field generation device for a regional numerical forecast model, comprising:
[0053] The minimum credible scale calculation module is used to calculate the minimum credible scale of the global model resolution based on the effective resolution of the forecast wind field of the global model. min ;
[0054] Kinetic error tolerance scale calculation module is used to calculate the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max The east-west or north-south width of the target area to be forecasted;
[0055] Residual scale calculation module, used for the forecast wind field W based on the global model of the target area and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) of the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model;
[0056] The forecast initial field generation module is used to convert the minimum credible scale l min The largest of the three, namely, the kinetic error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model respectively, and the two are added together to generate the initial field of the hot start of the regional numerical prediction model at the current time.
[0057] Preferably, the kinetic energy error tolerance scale calculation module is specifically used for:
[0058] The global model forecast wind field W at the minimum credible scale l min The target area scale l to the target area that needs forecasting max Fourier decomposition or low-pass filter is used to separate the scales at each scale l between the two, and the large-scale wind speed spectrum distribution W corresponding to the forecast wind field W of the global model at each scale l is obtained. large (l) and the small-scale wind speed spectrum distribution W small (l), the scale separation calculation formula is as follows:
[0059] W=W large (l)+W small (l)
[0060] in,
[0061] Analyze wind field in target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Discrete Fourier decomposition or low-pass filter is used to separate the scales at each scale l to obtain the analytical wind field of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0062]
[0063] in,
[0064] Large-scale wind speed spectrum distribution W of the forecast wind field W based on the global model large (l) and the analytical wind field of the target area Large-scale wind speed spectrum distribution Calculate the forecast wind field W of the global model and the analytical wind field of the target area at each scale l The difference kinetic energy spectrum A(l) is calculated as follows:
[0065]
[0066] Based on the difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l), the kinetic energy error tolerance l1 is obtained, and the calculation formula is as follows:
[0067] A(l)=∈1
[0068] Among them, the scale l that satisfies the above formula is the kinetic energy error tolerance scale corresponding to the first difference tolerance ∈1, denoted as l1.
[0069] Preferably, the residual scale calculation module is specifically used for:
[0070] Calculate the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model after the last startup The wind field error D between them is as follows:
[0071]
[0072] The wind field forecast of the regional numerical forecast model after the last start-up of the target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Scale separation is performed on each scale l between the two to obtain the forecast wind field of the regional numerical forecast model at the last startup of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0073]
[0074] in,
[0075] The large-scale wind speed spectrum distribution W corresponding to the forecast wind field W at each scale l based on the global model large (l) and the forecast wind field of the regional numerical forecast model at the last startup of the target area The corresponding large-scale wind speed spectrum distribution at each scale l Calculate the large-scale wind field error d(l) corresponding to each scale l, and the calculation formula is as follows:
[0076]
[0077] The forecast wind field W of the global model based on the target area and the forecast wind field of the regional numerical prediction model at the last startup The residual difference kinetic energy B(l) at the small-scale end is calculated by the wind field error D between the two scales and the large-scale wind field error d(l) corresponding to each scale l. The calculation formula is as follows:
[0078]
[0079] Based on the residual difference kinetic energy B(l) at the small scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) at the small scale end, the residual scale l2 of the regional numerical prediction model is obtained, and the calculation formula is as follows:
[0080] B(l)=∈2
[0081] Among them, the scale l that satisfies the above formula is the residual scale of the regional numerical prediction model corresponding to the second difference tolerance ∈2, denoted as l2.
[0082] Preferably, the forecast initial field generation module is specifically used for:
[0083] Based on the minimum credible scale l min , kinetic energy error tolerance scale l1 and regional numerical prediction model residual scale l2 determine the dynamic hybrid scale l b , the formula is as follows:
[0084] l b =max(l min ,l1,l2);
[0085] The dynamic mixing scale l b As a separation scale and using discrete Fourier decomposition or low-pass filter to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model respectively;
[0086] The mixed field obtained by adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model is used as the initial field for the hot start of the regional numerical forecast model at the current time.
[0087] Preferably, the effective resolution refers to the distribution profile K(l) of the high-altitude wind field kinetic energy spectrum K with scale l, which begins to be significantly lower than the standard -5 / 3 power law distribution. The calculation formula is as follows:
[0088]
[0089] And for any l<l min
[0090]
[0091] Where n is the height and η is the specified small kinetic energy difference.
[0092] It can be seen from the above technical scheme that the present invention discloses a method and system for generating a hybrid hot start initial field of a regional numerical prediction model, by separating the global model and the regional numerical prediction model into large and small scales, and then merging the large-scale information of the global model and the small-scale information of the regional numerical prediction model to introduce the large-scale information of the global model in real time. The present invention calculates the kinetic energy spectrum and error distribution of the global and regional numerical prediction models in the designated forecast area in real time at the hot start moment of the regional numerical prediction model, specifies the difference tolerance so as to reduce the error of the regional numerical prediction model as much as possible, and then controls the residual difference kinetic energy of the designated small-scale end, thereby maintaining the stability of the regional numerical prediction model in the hot start start period. Because the scheme is fast in calculation, it does not run intermittently during operation, but introduces large-scale information at each forecast time; the scheme can dynamically reduce the error of the regional numerical prediction model, and is conducive to the rapid achievement of model stability after the hot start of the regional weather model, thereby improving the forecast effect of various meteorological elements. Therefore, the present invention can introduce sufficient global model information to achieve the hot start of the regional numerical prediction model without affecting the stability of the regional numerical prediction model.
[0093] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0094] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:
[0095] Figure 1 It is a flow chart of a method for generating a hybrid hot start initial field of a regional numerical forecast model in a preferred embodiment of the present invention;
[0096] Figure 2 It is a structural schematic diagram of a hybrid hot start initial field generation system of a regional numerical forecast model in a preferred implementation manner of the present invention. DETAILED DESCRIPTION
[0097] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0098] Those skilled in the art will appreciate that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined.
[0099] like Figure 1 As shown, the present invention provides a method for generating a hybrid hot start initial field of a regional numerical forecast model, and the method may include the following steps:
[0100] S1, the minimum credible scale l of the global model resolution is calculated based on the effective resolution of the forecast wind field of the global model min ;
[0101] Since the atmospheric model will maintain equilibrium at the minimum credible scale l min Artificial kinetic energy attenuation is performed below, so the initial field generation method of the present invention does not trust the wind field part below this scale.
[0102] Specifically, the effective resolution in this step refers to the distribution profile K(l) of the high-altitude wind field kinetic energy spectrum K with scale l, which begins to be significantly lower than the standard -5 / 3 power law distribution. The calculation formula is as follows:
[0103]
[0104] And for any l<l min
[0105]
[0106] Wherein, n is the height, and η is the specified small kinetic energy difference. Specifically, in this embodiment, n is 250; η is specified according to actual needs, generally about 4-10 times the grid spacing of the atmospheric model, and in this embodiment, η is 6 times the grid spacing of the atmospheric model.
[0107] S2, based on the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max The east-west or north-south width of the target area to be forecasted;
[0108] Specifically, in this embodiment, step S2 may include the following steps:
[0109] S21, the forecast wind field W of the global model at the minimum credible scale l min The target area scale l to the target area that needs forecasting max Fourier decomposition or low-pass filter is used to separate the scales at each scale l between the two, and the large-scale wind speed spectrum distribution W corresponding to the forecast wind field W of the global model at each scale l is obtained. large (l) and the small-scale wind speed spectrum distribution W small (l), the scale separation calculation formula is as follows:
[0110] W=W large (l)+W small (l)
[0111] in,
[0112] From this we can get the global model forecast wind field W in Decomposed wind speed results at these scales.
[0113] Specifically, in this embodiment, a Raymond low-pass filter is used to perform scale separation.
[0114] S22, Analyze wind field in target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Discrete Fourier decomposition or low-pass filter is used to separate the scales at each scale l to obtain the analytical wind field of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0115]
[0116] in,
[0117] This can give the analytical wind field of the target area. exist Decomposed wind speed results at these scales.
[0118] Specifically, in this embodiment, a Raymond low-pass filter is used to perform scale separation.
[0119] It should be noted that the analysis wind field of the target area in this embodiment is It refers to the wind field adjusted by observation and is generally considered to be the true value.
[0120] S23, Large-scale wind speed spectrum distribution W of the forecast wind field W based on the global model large (l) and the analytical wind field of the target area Large-scale wind speed spectrum distribution Calculate the forecast wind field W of the global model and the analytical wind field of the target area at each scale l The difference kinetic energy spectrum A(l) is calculated as follows:
[0121]
[0122] S24, based on the difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l), a kinetic energy error tolerance l1 is obtained, and the calculation formula is as follows:
[0123] A(l)=∈1
[0124] The scale l satisfying the above formula is the kinetic energy error tolerance scale corresponding to the first difference tolerance ∈ 1, which is denoted as l 1. Specifically, the first difference tolerance ∈ 1 is preset according to specific circumstances, and in this embodiment, its value range is 0.05≤∈1≤0.1.
[0125] S3, based on the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) of the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model;
[0126] Specifically, in this embodiment, step S3 may include the following steps:
[0127] S31, calculate the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model after the last startup The wind field error D between them is as follows:
[0128]
[0129] S32, the forecast wind field of the regional numerical forecast model after the last startup of the target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Scale separation is performed on each scale l between the two to obtain the forecast wind field of the regional numerical forecast model at the last startup of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0130]
[0131] in,
[0132] From this, we can get the forecast wind field of the regional numerical forecast model at the last startup of the target area. exist Decomposed wind speed results at these scales.
[0133] Specifically, in this embodiment, a Raymond low-pass filter is used for scale separation;
[0134] S33, based on the forecast wind field W of the global model obtained in S21, the corresponding large-scale wind speed spectrum distribution W at each scale l large (l) and the forecast wind field of the regional numerical forecast model at the last startup of the target area obtained by S32 The corresponding large-scale wind speed spectrum distribution at each scale l Calculate the large-scale wind field error d(l) corresponding to each scale l, and the calculation formula is as follows:
[0135]
[0136] S34, based on the forecast wind field W of the global model of the target area obtained in S31 and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) at the small-scale end is calculated by using the wind field error D between the two and the large-scale wind field error d(l) corresponding to each scale l obtained by S33. The calculation formula is as follows:
[0137]
[0138] S35, based on the residual difference kinetic energy B(l) at the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) at the small-scale end, the regional numerical prediction model residual scale l2 is obtained, and the calculation formula is as follows:
[0139] B(l)=∈2
[0140] Among them, the scale l that satisfies the above formula is the residual scale of the regional numerical forecast model corresponding to the second difference tolerance ∈2, which is recorded as l2. l2 represents the retention amount of the CCP weather system in the regional numerical forecast model at the next startup, which to a certain extent predicts the integral stability of the regional weather forecast model system during the hot start process. Specifically, the second difference tolerance ∈2 is pre-set according to the specific situation. In this embodiment, its value range is 0.05≤∈2≤0.1.
[0141] S4, the minimum credible scale lmin The largest of the three, namely, the kinetic error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model respectively, and the two are added together to generate the initial field of the hot start of the regional numerical prediction model at the current time.
[0142] Specifically, in this embodiment, step S4 may include the following steps:
[0143] S41, based on the minimum credible scale l min , kinetic energy error tolerance scale l1 and regional numerical prediction model residual scale l2 determine the dynamic hybrid scale l b , the formula is as follows:
[0144] l b =max(l min ,l1,l2);
[0145] Specifically, because the smaller the scale, the greater the error that may be introduced, therefore, we choose l min The largest of l1 and l2 is taken as the dynamic mixing scale l b , ensuring that the introduced error is relatively small, thereby ensuring the introduction of a sufficiently trusted global model to reduce the error while ensuring the stability of the hot start.
[0146] S42, the dynamic mixing scale l b As a separation scale and using discrete Fourier decomposition or low-pass filter to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model respectively;
[0147] Specifically, the dynamic mixing scale l is calculated using b The Raymond low-pass filter is used as a separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model.
[0148] Specifically, the forecast field includes at least one of the following: wind field, temperature field, water vapor field, pressure field, and potential height field.
[0149] S43, adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model to obtain a mixed field as the initial field for hot starting of the regional numerical forecast model at the current time.
[0150] When hot starting the regional numerical prediction model for the current time period, the mixed field obtained by adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model is used as the initial field to hot start the regional numerical prediction model.
[0151] Specifically, taking the wind field as an example, let the initial field obtained after mixing be W b ,but:
[0152]
[0153] In summary, this embodiment discloses a method for generating a hybrid hot start initial field of a regional numerical forecast model. First, the minimum credible scale l of the global model resolution is calculated based on the effective resolution of the forecast wind field of the global model. min ; Then based on the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max is the east-west or north-south width of the target area to be forecasted; then the forecast wind field W based on the global model of the target area and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) of the small-scale end and the second difference tolerance ∈2 of the preset residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model; finally, the minimum credible scale l min The largest of the three, namely, the kinetic error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model respectively, and the two are added together to generate the initial field of the hot start of the regional numerical prediction model at the current time.
[0154] The hybrid hot start initial field generation method of the regional numerical prediction model of this embodiment separates the global model and the regional numerical prediction model into large and small scales, and then merges the large-scale information of the global model and the small-scale information of the regional numerical prediction model to introduce the large-scale information of the global model in real time. The present invention calculates the kinetic energy spectrum and error distribution of the global and regional numerical prediction models in the specified forecast area in real time at the hot start moment of the regional numerical prediction model, specifies the difference tolerance so as to reduce the error of the regional numerical prediction model as much as possible, and then controls the residual difference kinetic energy of the specified small-scale end, thereby maintaining the stability of the regional numerical prediction model in the hot start start period. Because the scheme is fast in calculation, it does not run intermittently during operation, but instead introduces large-scale information at each forecast time; the scheme can dynamically reduce the error of the regional numerical prediction model, and is conducive to the rapid achievement of model stability after the hot start of the regional weather model, thereby improving the forecast effect of various meteorological elements.
[0155] This scheme is different from the previous fixed-scale hybrid technology scheme that was arbitrarily set by humans. It adopts a dynamic scale hybrid technology based on kinetic error judgment. When introducing large-scale information from the global model, it fully considers controlling the error of the introduced band, so it can effectively reduce the error of the regional numerical prediction model; at the same time, this scheme controls the difference between the two modes at the small-scale end so that the regional numerical prediction model can quickly reach model stability during hot start.
[0156] This solution is used separately from the weather model, so it does not depend on the specific weather model and data source, and can be used in conjunction with any of the original hybrid technologies, so it has strong flexibility. This solution changes the traditional partial cold start method, greatly reducing the amount of calculation when introducing large-scale information into the regional numerical forecast model, thereby enabling multiple times of regional weather forecast hot start continuous operation, which can better reflect the requirements of the changeable weather system changes on the initial field of the regional numerical forecast model.
[0157] Therefore, the present invention can introduce sufficient global model information to achieve hot start of the regional numerical forecast model without affecting the stability of the regional numerical forecast model.
[0158] like Figure 2 As shown, an embodiment of the present invention further provides a hybrid hot start initial field generation system for a regional numerical forecast model, which may include:
[0159] The minimum credible scale calculation module 201 is used to calculate the minimum credible scale l of the global model resolution based on the effective resolution of the forecast wind field of the global model. min ;
[0160] Kinetic energy error tolerance scale calculation module 202, used for the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max The east-west or north-south width of the target area to be forecasted;
[0161] The residual scale calculation module 203 is used to calculate the wind field W based on the global model of the target area and the wind field W of the regional numerical forecast model at the last startup. The residual difference kinetic energy B(l) of the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model;
[0162] The forecast initial field generation module 204 is used to convert the minimum credible scale l min The largest of the three, namely, the kinetic error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model respectively, and the two are added together to generate the initial field of the hot start of the regional numerical prediction model at the current time.
[0163] In this embodiment, the kinetic energy error tolerance scale calculation module 202 is specifically used for:
[0164] The global model forecast wind field W at the minimum credible scale l min The target area scale l to the target area that needs forecasting max Fourier decomposition or low-pass filter is used to separate the scales at each scale l between the two, and the large-scale wind speed spectrum distribution W corresponding to the forecast wind field W of the global model at each scale l is obtained. large (l) and the small-scale wind speed spectrum distribution W small (l), the scale separation calculation formula is as follows:
[0165] W=W large (l)+W small (l)
[0166] in,
[0167] Analyze wind field in target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Discrete Fourier decomposition or low-pass filter is used to separate the scales at each scale l to obtain the analytical wind field of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0168]
[0169] in,
[0170] Large-scale wind speed spectrum distribution W of the forecast wind field W based on the global model large (l) and the analytical wind field of the target area Large-scale wind speed spectrum distribution Calculate the forecast wind field W of the global model and the analytical wind field of the target area at each scale l The difference kinetic energy spectrum A(l) is calculated as follows:
[0171]
[0172] Based on the difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l), the kinetic energy error tolerance l1 is obtained, and the calculation formula is as follows:
[0173] A(l)=∈1
[0174] Among them, the scale l that satisfies the above formula is the kinetic energy error tolerance scale corresponding to the first difference tolerance ∈1, denoted as l1.
[0175] In this embodiment, the residual scale calculation module 203 is specifically used for:
[0176] Calculate the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model after the last startup The wind field error D between them is as follows:
[0177]
[0178] The wind field forecast of the regional numerical forecast model after the last start-up of the target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Scale separation is performed on each scale l between the two to obtain the forecast wind field of the regional numerical forecast model at the last startup of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows:
[0179]
[0180] in,
[0181] The large-scale wind speed spectrum distribution W corresponding to the forecast wind field W at each scale l based on the global model large (l) and the forecast wind field of the regional numerical forecast model at the last startup of the target area The corresponding large-scale wind speed spectrum distribution at each scale l Calculate the large-scale wind field error d(l) corresponding to each scale l, and the calculation formula is as follows:
[0182]
[0183] The forecast wind field W of the global model based on the target area and the forecast wind field of the regional numerical prediction model at the last startup The residual difference kinetic energy B(l) at the small-scale end is calculated by the wind field error D between the two scales and the large-scale wind field error d(l) corresponding to each scale l. The calculation formula is as follows:
[0184]
[0185] Based on the residual difference kinetic energy B(l) at the small scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) at the small scale end, the residual scale l2 of the regional numerical prediction model is obtained, and the calculation formula is as follows:
[0186] B(l)=∈2
[0187] Among them, the scale l that satisfies the above formula is the residual scale of the regional numerical prediction model corresponding to the second difference tolerance ∈2, denoted as l2.
[0188] In this embodiment, the forecast initial field generation module 204 is specifically used for:
[0189] Based on the minimum credible scale l min , kinetic energy error tolerance scale l1 and regional numerical prediction model residual scale l2 determine the dynamic hybrid scale l b , the formula is as follows:
[0190] l b =max(l min ,l1,l2);
[0191] The dynamic mixing scale l b As a separation scale and using discrete Fourier decomposition or low-pass filter to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model respectively;
[0192] The mixed field obtained by adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model is used as the initial field for the hot start of the regional numerical forecast model at the current time.
[0193] In this embodiment, the effective resolution refers to the distribution profile K(l) of the high-altitude wind field kinetic energy spectrum K with scale l, which begins to be significantly lower than the standard -5 / 3 power law distribution. The calculation formula is as follows:
[0194]
[0195] And for any l<l min
[0196]
[0197] Where n is the height and η is the specified small kinetic energy difference.
[0198] The hybrid hot start initial field generation system of the regional numerical forecast mode in this embodiment is implemented based on the hybrid hot start initial field generation method of the regional numerical forecast mode in the above method embodiment, and its working principle is the same as that of the above method embodiment, which will not be repeated here.
[0199] In this embodiment, if the module / unit integrated in the hybrid hot start initial field generation system of the regional numerical forecast mode is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.
[0200] Those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.
[0201] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for generating a hybrid hot start initial field for a regional numerical forecast model, characterized in that: include: S1, the minimum credible scale l of the global model resolution is calculated based on the effective resolution of the forecast wind field of the global model min The effective resolution in step S1 refers to the distribution profile K(l) of the high-altitude wind field kinetic energy spectrum K with scale l, which begins to be significantly lower than the standard -5 / 3 power law distribution. The calculation formula is as follows: And for any l<l min Where n is the height and η is the specified small kinetic energy difference; S2, based on the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max is the east-west or north-south width of the target area to be forecasted; step S2 comprises: S21, the forecast wind field W of the global model at the minimum credible scale l min The target area scale l to the target area that needs forecasting max Fourier decomposition or low-pass filter is used to separate the scales at each scale l between the two, and the large-scale wind speed spectrum distribution W corresponding to the forecast wind field W of the global model at each scale l is obtained. large (l) and the small-scale wind speed spectrum distribution W small (l), the scale separation calculation formula is as follows: W=W large (l)+W small (l) in, S22, Analyze wind field in target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Discrete Fourier decomposition or low-pass filter is used to separate the scales at each scale l to obtain the analytical wind field of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows: in, S23, Large-scale wind speed spectrum distribution W of the forecast wind field W based on the global model large (l) and the analytical wind field of the target area Large-scale wind speed spectrum distribution Calculate the forecast wind field W of the global model and the analytical wind field of the target area at each scale l The difference kinetic energy spectrum A(l) is calculated as follows: S24, based on the difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l), a kinetic energy error tolerance l1 is obtained, and the calculation formula is as follows: A(l)=∈1 Among them, the scale l that satisfies the above formula is the kinetic energy error tolerance scale corresponding to the first difference tolerance ∈1, denoted as l1; S3, based on the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) of the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model; step S3 includes: S31, calculate the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model after the last startup The wind field error D between them is as follows: S32, the forecast wind field of the regional numerical forecast model after the last startup of the target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Scale separation is performed on each scale l between the two to obtain the forecast wind field of the regional numerical forecast model at the last startup of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows: in, S33, based on the forecast wind field W of the global model obtained in S21, the corresponding large-scale wind speed spectrum distribution W at each scale l large (l) and the forecast wind field of the regional numerical forecast model at the last startup of the target area obtained by S32 The corresponding large-scale wind speed spectrum distribution at each scale l Calculate the large-scale wind field error d(l) corresponding to each scale l, and the calculation formula is as follows: S34, based on the forecast wind field W of the global model of the target area obtained in S31 and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) at the small-scale end is calculated by using the wind field error D between the two and the large-scale wind field error d(l) corresponding to each scale l obtained by S33. The calculation formula is as follows: S35, based on the residual difference kinetic energy B(l) at the small scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) at the small scale end, the regional numerical prediction model residual scale l2 is obtained, and the calculation formula is as follows: B(l)=∈2 Among them, the scale l that satisfies the above formula is the residual scale of the regional numerical prediction model corresponding to the second difference tolerance ∈2, denoted as l2; S4, the minimum credible scale l min The largest of the three, the kinetic energy error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to respectively extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model, and the two are added to generate the initial field of the hot start of the regional numerical prediction model at the current time; step S4 comprises: S41, based on the minimum credible scale l min , kinetic energy error tolerance scale l1 and regional numerical prediction model residual scale l2 determine the dynamic mixing scale l b , the formula is as follows: the b =max(l min ,l1,l2); S42, the dynamic mixing scale l b As a separation scale and using discrete Fourier decomposition or low-pass filter to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model respectively; S43, adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model to obtain a mixed field as the initial field for hot starting of the regional numerical forecast model at the current time.
2. A hybrid hot start initial field generation system for a regional numerical forecast model, characterized in that: include: The minimum credible scale calculation module is used to calculate the minimum credible scale of the global model resolution based on the effective resolution of the forecast wind field of the global model. min ; Kinetic error tolerance scale calculation module is used to calculate the minimum credible scale l min The target area scale l max The predicted wind field W of the global model and the analyzed wind field of the target area at each scale l between The difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l) are used to obtain the kinetic energy error tolerance scale l1, where the target area scale l max The east-west or north-south width of the target area to be forecasted; Residual scale calculation module, used for the forecast wind field W based on the global model of the target area and the forecast wind field of the regional numerical forecast model at the last startup The residual difference kinetic energy B(l) of the small-scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) of the small-scale end are used to obtain the residual scale l2 of the regional numerical prediction model; The forecast initial field generation module is used to convert the minimum credible scale l min The largest of the three scales, i.e., the kinetic energy error tolerance scale l1 and the regional numerical prediction model residual scale l2, is used as the separation scale to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical prediction model respectively, and the two are added together to generate the initial field of the hot start of the regional numerical prediction model at the current time; The kinetic energy error tolerance scale calculation module is specifically used for: The global model forecast wind field W at the minimum credible scale l min The target area scale l to the target area that needs forecasting max Fourier decomposition or low-pass filter is used to separate the scales at each scale l between the two, and the large-scale wind speed spectrum distribution W corresponding to the forecast wind field W of the global model at each scale l is obtained. large (l) and the small-scale wind speed spectrum distribution W small (l), the scale separation calculation formula is as follows: W=W large (l)+W small (l) in, Analyze wind field in target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Discrete Fourier decomposition or low-pass filter is used to separate the scales at each scale l to obtain the analytical wind field of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows: in, Large-scale wind speed spectrum distribution W of the forecast wind field W based on the global model large (l) and the analytical wind field of the target area Large-scale wind speed spectrum distribution Calculate the forecast wind field W of the global model and the analytical wind field of the target area at each scale l The difference kinetic energy spectrum A(l) is calculated as follows: Based on the difference kinetic energy spectrum A(l) and the preset first difference tolerance ∈1 of the difference kinetic energy spectrum A(l), the kinetic energy error tolerance l1 is obtained, and the calculation formula is as follows: A(l)=∈1 Among them, the scale l that satisfies the above formula is the kinetic energy error tolerance scale corresponding to the first difference tolerance ∈1, denoted as l1; The residual scale calculation module is specifically used for: Calculate the forecast wind field W of the global model in the target area and the forecast wind field of the regional numerical forecast model after the last startup The wind field error D between them is as follows: The wind field forecast of the regional numerical forecast model after the last start-up of the target area At the minimum credible scale l min The target area scale l to the target area that needs forecasting max Scale separation is performed on each scale l between the two to obtain the forecast wind field of the regional numerical forecast model at the last startup of the target area. The corresponding large-scale wind speed spectrum distribution at each scale l And the small-scale wind speed spectrum distribution The scale separation calculation formula is as follows: in, The large-scale wind speed spectrum distribution W corresponding to the forecast wind field W at each scale l based on the global model large (l) and the forecast wind field of the regional numerical forecast model at the last startup of the target area The corresponding large-scale wind speed spectrum distribution at each scale l Calculate the large-scale wind field error d(l) corresponding to each scale l, and the calculation formula is as follows: The forecast wind field W of the global model based on the target area and the forecast wind field of the regional numerical prediction model at the last startup The residual difference kinetic energy B(l) at the small-scale end is calculated by the wind field error D between the two scales and the large-scale wind field error d(l) corresponding to each scale l. The calculation formula is as follows: Based on the residual difference kinetic energy B(l) at the small scale end and the preset second difference tolerance ∈2 of the residual difference kinetic energy B(l) at the small scale end, the residual scale l2 of the regional numerical prediction model is obtained, and the calculation formula is as follows: B(l)=∈2 Among them, the scale l that satisfies the above formula is the residual scale of the regional numerical prediction model corresponding to the second difference tolerance ∈2, denoted as l2; The forecast initial field generation module is specifically used for: Based on the minimum credible scale l min , kinetic energy error tolerance scale l1 and regional numerical prediction model residual scale l2 determine the dynamic mixing scale l b , the formula is as follows: the b =max(l min ,l1,l2); The dynamic mixing scale l b As a separation scale and using discrete Fourier decomposition or low-pass filter to extract the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model respectively; The mixed field obtained by adding the large-scale information of the forecast field of the global model and the small-scale information of the forecast field of the regional numerical forecast model is used as the initial field for the hot start of the regional numerical forecast model at the current time; The effective resolution refers to the distribution profile K(l) of the high-altitude wind field kinetic energy spectrum K with scale l, which begins to be significantly lower than the standard -5 / 3 power law distribution. The calculation formula is as follows: And for any l<l min Where n is the height and η is the specified small kinetic energy difference.
