Method and system for assimilating non-Gaussian distribution data of hyperspectral error of meteorological satellite based on quantum calculation

Abstract

The invention discloses a method and system for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing, and the method carries out the generalized quality control of meteorological satellite hyperspectral data in consideration of an outlier value on the observation brightness temperature and equivalent data of a satellite hyperspectral infrared detector channel. Comprising the steps of channel optimal selection considering quantum information entropy, deviation correction based on artificial intelligence generalized integration generation type deep learning and cloud detection based on a minimum residual method to remove cloud view field point data, so that quality control data is obtained. And performing quantum space solution on the quantum non-Gaussian increment four-dimensional variational assimilation model compatible with error Gaussian distribution and non-Gaussian distribution based on quality control data to obtain analysis field data, and solving the problem that the brightness temperature assimilation of the hyperspectral infrared channel with error non-Gaussian distribution (or obvious non-Gaussian characteristics) is limited in the prior art. Meanwhile, the method is high in solving speed, and the obtained analysis field data has good precision and calculation timeliness.

Description

Technical Field

[0001] The present invention belongs to the field of atmospheric science technology, and in particular relates to a method and system for assimilating high-spectral error non-Gaussian distribution data of meteorological satellites based on quantum computing. Background Art

[0002] Data assimilation is a key component of atmospheric science, enabling the integration of observational data with numerical weather prediction models. The quality of numerical weather forecasts depends heavily on the accuracy of the initial conditions provided by the assimilation system. Satellite data assimilation, particularly hyperspectral infrared data, accounts for a significant portion of operational numerical weather prediction systems. Data assimilation involves effectively integrating all available information with the background field to produce the most accurate estimate possible of the atmospheric state at a given moment, known as the analysis field. The "analysis field" derived from data assimilation provides higher-quality "initial values" for numerical weather forecasts.

[0003] The currently internationally accepted classical variational data assimilation method requires that errors ("errors" belongs to the mathematical field, and is called "bias" in the field of meteorological satellites) follow a "Gaussian distribution." Variational assimilation requires that errors follow a Gaussian distribution. For satellite detector data assimilation, this requires that the channel brightness temperature bias also follow a Gaussian distribution. The so-called "brightness temperature bias" is the difference between the actual observed brightness temperature of the satellite channel and the simulated brightness temperature. The simulated brightness temperature is based on the spectral coefficients of the satellite detector and is obtained by simulating and interpolating the background field using the rapid radiative transfer model. Through theoretical analysis and extensive experimental research, domestic and foreign scholars have demonstrated that the "Gaussian distribution" of errors is only a theoretical assumption. In reality, many data errors exhibit a "non-Gaussian distribution." Therefore, the analysis fields obtained using classical variational data assimilation have certain accuracy issues, affecting subsequent numerical forecast results. For example, the brightness temperature data of the water vapor channel of meteorological satellite hyperspectral infrared detectors are highly non-Gaussian. If the internationally classic variational method is used to assimilate the brightness temperature of the hyperspectral infrared water vapor channel, the brightness temperature of the hyperspectral infrared water vapor channel with strong non-Gaussianity will be "poorly used" or its use will be restricted. A lot of data will be "eliminated" by the quality control before variational assimilation, resulting in the loss of a lot of information brought by the "usable" data, thereby affecting the numerical forecast results of wet weather processes (heavy rainfall and typhoons, etc.). Summary of the Invention

[0004] In order to solve the problems existing in the prior art, the present invention proposes a method and system for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing.

[0005] The technical solutions of the present invention are as follows:

[0006] A method for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing, the method comprising:

[0007] The radiation transfer model is simulated for the meteorological satellite hyperspectral channel data after background field data interpolation processing to obtain the corresponding satellite hyperspectral channel observation brightness temperature equivalent data as the channel simulation brightness temperature;

[0008] Performing generalized quality control of meteorological satellite hyperspectral data considering outliers on the channel simulated brightness temperature to form quality control data, wherein the generalized quality control includes optimal channel selection considering quantum information entropy, deviation correction based on artificial intelligence generalized ensemble generative deep learning, and cloud detection and removal based on the minimum residual method;

[0009] A quantum non-Gaussian incremental four-dimensional variational assimilation model compatible with Gaussian and non-Gaussian error distributions is constructed, wherein a dynamically adjustable weight contribution factor is introduced into the observation term in the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model;

[0010] Based on the quality control data, the quantum non-Gaussian incremental four-dimensional variational assimilation model is solved to obtain analysis field data.

[0011] Furthermore, the specific method for optimal channel selection considering quantum information entropy includes:

[0012] 1) Eliminate satellite channels whose absolute value of the brightness temperature error of the channel simulation in the radiation transfer model exceeds 10K; the channel simulation brightness temperature error is defined as the difference between the channel observation brightness temperature and the channel simulation brightness temperature.

[0013] 2) Based on the peak layer of the satellite channel weight function, that is, the atmospheric layer information at different altitudes detected by different satellite channels, the satellite channels are preliminarily selected to obtain a preliminarily selected channel combination;

[0014] 3) Select the optimal channel for the initially selected channel combination considering quantum information entropy.

[0015] Furthermore, the channel observation brightness temperature is obtained by performing nonlinear brightness temperature reconstruction based on compressed sensing on the brightness temperature data observed by the meteorological satellite hyperspectral infrared channel. The specific method is as follows:

[0016] The brightness temperature data observed by the hyperspectral infrared channel of the meteorological satellite is sparsely represented in the wavelet domain space, and the original brightness temperature signal is reconstructed by decoding. The specific expression is as follows:

[0017]

[0018] Among them, x true is the reconstructed original brightness temperature signal; y signal is the observed compressed brightness temperature signal, that is, the received brightness temperature signal; H CS is the structural operator; R signal is the brightness temperature signal error; W CS represents the transformation matrix; λCS represents the regularization parameter.

[0019] Furthermore, the specific method of performing optimal channel selection considering quantum information entropy on the preliminarily selected channel combination includes:

[0020] 1) Normalizing the observed brightness temperature sequence data corresponding to each channel of the hyperspectral infrared detector in the preliminarily screened channel combination; For the observed brightness temperature sequence data X of the i-th channel in the channel combination i,entropy ={x i,entropy (j), j = 1, 2, ..., N}, the normalized observed brightness temperature sequence data of the i-th channel is recorded as Y i,entropy ={y i,entropy (j), j=1,2,...,N}; where x i,entropy (j) is the jth observed brightness temperature sequence data of the i-th channel; j is the serial number mark of the observed brightness temperature sequence data, and N is the total amount of observed brightness temperature sequence data; y i,entropy (j) is the normalized brightness temperature sequence data of the jth observation of the i-th channel;

[0021] 2) Perform quantum space reconstruction on the normalized sequence data of each channel to obtain the quantum space matrix corresponding to each channel; for the normalized observed brightness temperature sequence data Y of the i-th channel i,entropy , and its corresponding quantum space matrix The expression is as follows:

[0022]

[0023] Where, is the l-th reconstruction component in the quantum space matrix, l=1,2,...,K, K is the total number of reconstruction components, m is the embedding dimension; is the time window;

[0024] 3) Quantize each reconstructed component in the quantum space matrix of each channel and calculate the quantum information entropy of each channel; Each reconstructed component in is quantized as follows:

[0025]

[0026] In the formula, |q b > represents the reconstruction component The qth state vector of l,k Represents the state vector |q b >, k=1,2,...,n, n is the number of state vectors, n=2 m, m is the constituent qubit;

[0027] The probability of the state vector appearing P i,k As the probability of an event occurring, we get the quantum space matrix Quantum information entropy in, k=1,2,...,n;

[0028] 4) Based on the quantum information entropy and variational assimilation background error covariance matrix of each channel, the quantum mutual information of each channel is calculated to form a multidimensional feature space composed of multiple groups of channels, and the satellite channel corresponding to the maximum quantum mutual information is searched in the multidimensional feature space and selected as the optimal channel.

[0029] Furthermore, the specific method of deviation correction based on artificial intelligence generalized integrated generative deep learning includes:

[0030] 1) Selecting bias-corrected predictors;

[0031] 2) For the brightness temperature equivalent data of all field-of-view points of each channel of the meteorological satellite hyperspectral infrared detector that need to be corrected, a deep model generated by combining convolutional neural network and deep neural network models is constructed as the basic model for ensemble learning, and generalized weighted ensemble machine learning is performed to find the optimal or suboptimal ensemble weight for integrating the prediction results of the basic model. The objective minimization function of the weighted ensemble machine learning is:

[0032]

[0033] Where n is the total number of samples; is the actual value of the value i to be corrected. is the predicted value of the to-be-corrected value i by the basic model j, and the value of the predicted value is obtained based on the selected prediction factors as the input of the basic model; is the integration weight corresponding to the base model j.

[0034] Furthermore, the specific method of cloud detection and elimination based on the minimum residual method includes:

[0035] 1) According to the radiation transfer theory, the simulated radiation value of the meteorological satellite hyperspectral infrared detector channel is mathematically modeled. The simulated radiation value of the satellite channel g is Expressed as: Among them, N e Indicates the effective cloud cover at the viewing point; represents clear-sky radiation; The cloud top pressure is p c,g Cloudy radiation simulation at 1000 Hz;

[0036] 2) Calculate the deviation between the observed radiation value and the simulated radiation value of the hyperspectral infrared detector channel. For the observed radiation value of channel g, and simulated radiation values The deviation δ g Expressed as:

[0037] 3) Based on the n channel combinations obtained by the optimal channel selection based on quantum information entropy, the minimum residual method objective function is minimized at each field of view point to obtain the effective cloud amount N at the field of view point e and effective cloud top pressure p c , the minimum residual method objective function is defined as follows:

[0038]

[0039] 4) Based on effective cloud cover N e The cloud detection of the field point is performed with the set threshold. For the effective cloud amount N e For a viewpoint whose brightness temperature is greater than the set threshold, all brightness temperature data of all channel combinations at the viewpoint are discarded.

[0040] Furthermore, the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model is:

[0041]

[0042] w(r i )=(1 / r i )·(dρ(r i ) / dr i )

[0043] Where δx0 is the analysis increment, x0 is the control variable, and the solved x0 is the analysis field; is the state vector predicted by the numerical prediction model; the superscript f indicates forecast; B is the background error covariance matrix; R i is the observation error covariance matrix; superscript T is the matrix transpose; superscript -1 is the matrix inverse; d i is the brightness temperature increment, y i represents the i-th observation value, i = 1, 2, ..., N, N is the total number of observations; represents the brightness temperature value of the ith observation value simulated by the rapid radiative transfer model based on the pattern space, which is also the equivalent data of the ith observation value. H i To integrate the model M i,0 (·) The observation operator that maps the solution from the model space to the observation space; M i,0(x0) is a nonlinear numerical prediction model that integrates the control variable x0 from the initial condition and initial time t = 0 to t = i, where t is the time mark; the matrix H i and M i,0 Respectively represent H i and M i,0 Linear representation of the tangent pattern; r i is the brightness temperature increment amplitude, r i =d i / σ i , σ i is the observation error of channel i; w(r i ) is the weight contribution factor; ρ(r i ) is the M-estimation cost function.

[0044] Furthermore, the specific method for solving the quantum non-Gaussian incremental four-dimensional variational assimilation model includes:

[0045] 1) Using quantum annealing driven incremental four-dimensional non-Gaussian quantum data assimilation method, the real number representation is performed by Q qubits, and the analysis increment δx0 is quantized by the mapping matrix G:

[0046]

[0047] g T =[-2 Q-1 ,2 Q-2 ,2 Q-3 ,...,2 1 ,2 0 ]

[0048]

[0049] Where β is an adjustable scaling parameter; b is a binary vector whose elements are all 0 or 1, 0 is a vector whose elements are all 0; g is the quantization of a real number; Q is a natural number;

[0050] 2) Based on the quantized analytical increments, the cost function of the quantum non-Gaussian increment four-dimensional variational assimilation model is solved in quantum space:

[0051]

[0052] Where,

[0053]

[0054] Where Ham represents the Hamiltonian; w(·) represents the weight function of the constructed M-estimated new norm; A and u are as follows:

[0055] 3) Minimization through quantum annealing method Get the best analysis field.

[0056] Furthermore, it also includes the evaluation of the accuracy and timeliness of the analysis field data and the accuracy of the analysis field in improving the numerical weather forecast of wet processes.

[0057] A system for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing, the system comprising a first data processing module, a second data processing module, a third data processing module, and a fourth data processing module;

[0058] The first data processing module is used to simulate the radiation transmission mode of the meteorological satellite hyperspectral channel data after the background field data interpolation processing, and obtain the corresponding satellite hyperspectral channel observation brightness temperature equivalent data as the channel simulation brightness temperature;

[0059] The second data processing module is used to perform generalized quality control of meteorological satellite hyperspectral data considering outliers on the channel simulated brightness temperature to form quality control data, wherein the generalized quality control includes optimal channel selection considering quantum information entropy, deviation correction based on artificial intelligence generalized integrated generative deep learning, and cloud detection and elimination based on the minimum residual method;

[0060] The third data processing module is used to construct a quantum non-Gaussian incremental four-dimensional variational assimilation model that is compatible with Gaussian and non-Gaussian error distributions, and introduces a dynamically adjustable weight contribution factor into the observation term in the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model;

[0061] The fourth data processing module is used to solve the quantum non-Gaussian incremental four-dimensional variational assimilation model based on the quality control data to obtain analysis field data.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention proposes a method and system for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing. The method performs generalized quality control of the hyperspectral channel observation brightness temperature data considering outliers based on the equivalent data of the satellite hyperspectral infrared detector channel observation brightness temperature, thereby obtaining quality control data and ensuring the quality of data variational assimilation; then, based on the quality control data, a quantum non-Gaussian incremental four-dimensional variational assimilation model compatible with Gaussian and non-Gaussian error distributions is quantum-space solved to obtain analysis field data. The method solves the problem in the prior art that the assimilation of hyperspectral infrared channel brightness temperature with non-Gaussian error distribution (or obvious non-Gaussian characteristics) is limited, effectively assimilates hyperspectral water vapor channel brightness temperature, and improves the numerical forecast accuracy of wet processes such as precipitation and typhoons; at the same time, the method of the present invention has a fast solution speed, and the obtained analysis field data has good accuracy and computational timeliness, which can effectively improve the accuracy of numerical weather forecasts of wet processes.

[0064] In generalized quality control, the present method utilizes a three-step approach to optimal channel selection that considers quantum information entropy. This contrasts with the commonly used entropy reduction method in the field, which selects channels by minimizing the "amplitude" of analytical errors after identifying and updating the analytical field. Unlike these methods, the present method utilizes the superposition, coherence, and entanglement of quantum states, enabling it to accurately represent objective laws.

[0065] The channel-observed brightness temperature in the present method is obtained by performing nonlinear brightness temperature reconstruction based on compressed sensing on brightness temperature data observed from meteorological satellite hyperspectral infrared channels. Unlike conventional principal component analysis (PCA), which has difficulty removing "stripe" noise when reconstructing satellite hyperspectral brightness temperature observations, the present method proposes a nonlinear brightness temperature reconstruction based on compressed sensing, which effectively removes "stripe" noise. By introducing a sparse representation in the wavelet domain, better reconstruction results can be achieved with fewer random samples. In generalized quality control, the present method utilizes a bias correction method based on artificial intelligence generalized ensemble generative deep learning. Unlike common "offline" and "online" bias correction methods in the field, the present method proposes a bias correction method based on artificial intelligence generalized ensemble generative deep learning that incorporates the characteristics of hyperspectral channel brightness temperature bias. This method generates a deep ensemble learning base model based on a combination of convolutional neural networks (CNNs) and deep neural networks (DNNs). The prediction results of different base models are weighted and integrated to correct hyperspectral channel brightness temperature bias.

[0066] The method of the present invention performs quantum non-Gaussian incremental four-dimensional variational assimilation based on the new M-estimation norm. Different from the assumption that the observation error obeys the Gaussian distribution in classical variational assimilation, the present invention constructs a new norm based on the classical M-estimation method (e.g., Huber-estimation) and the basic properties of the M-estimation cost function (e.g., continuity, non-negativity, convexity, etc.) and the basic properties of the weight function (e.g., continuity, monotonicity, etc.). The new M-estimation norm is coupled to the classical variational assimilation cost function, and a quantum non-Gaussian variational assimilation method is proposed in the quantum state space. Non-Gaussian variational assimilation solves the defect that the classical variational method can only assimilate data errors obeying the Gaussian distribution. The key core technology of non-Gaussian variational assimilation is that it differs from classical variational assimilation in that it eliminates some "outliers" with large errors before minimization iteration. The non-Gaussian variational assimilation method introduces a weight function of the new M-estimation norm as a weight contribution factor in the observation term of the variational assimilation cost function, reconstructs a new cost function for variational assimilation, uses "outliers", but reduces the contribution of "outliers" to the cost function. The new non-Gaussian variational assimilation method of the present invention solves the problem of limited brightness temperature assimilation of the hyperspectral infrared water vapor channel and improves the accuracy of weather forecasts for wet processes such as typhoons and rainstorms. Different from the Euclidean space metric of the classical variational method, the new metric relationship of the quantum state space can better reveal the essence of the closer relationship between the observed brightness temperature and the simulated brightness temperature obtained by variational assimilation.

[0067] This invention employs quantum annealing-driven incremental four-dimensional non-Gaussian quantum data assimilation. Classical variational assimilation methods typically employ gradient descent and related improved methods to iteratively minimize the cost function to generate the optimal analysis field. However, this iterative minimization process is time-consuming due to the complex iterative computations required to fully reduce the cost function. Consequently, existing variational data assimilation methods require significant computational resources in numerical weather forecasting systems. This invention employs quantum annealing-driven incremental four-dimensional variational data assimilation to solve the quantum non-Gaussian incremental four-dimensional variational data assimilation model. This design accelerates the solution of the variational assimilation cost function and reduces computational resource consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Schematic diagram of the process of the method for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing of the present invention;

[0069] Figure 2 This is a principle block diagram of the method of assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing in the present invention. DETAILED DESCRIPTION

[0070] The present invention is further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent modifications of the present invention by those skilled in the art fall within the scope defined by the claims attached to this application.

[0071] Example 1:

[0072] The present invention provides a method for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing, such as Figure 1 and Figure 2 , the method comprising:

[0073] S1, interpolating the background field data that matches the meteorological satellite hyperspectral channel observation data in time and space to the field of view point of the meteorological satellite hyperspectral channel data;

[0074] Furthermore, in this example, the meteorological satellite hyperspectral channel data can be selected from FY-4A / GIIRS data. Of course, data from the Fengyun-4B hyperspectral channel, other types of polar-orbiting and geostationary satellite hyperspectral infrared detectors can also be selected. The FY-4A / GIIRS data are obtained from the official website of the National Satellite Meteorological Center (http: / / satellite.nsmc.org.cn) and can be obtained by registration. The background field data of the present invention are obtained from the Final Global Data Assimilation System (FNL) data of the National Centers for Environmental Prediction (NCEP) of the United States. The data are obtained by registration on the relevant official website.

[0075] The above-mentioned FY-4A / GIIRS data refers to the detection data of the Geostationary Interferometric Infrared Sounder (GIIRS) carried by the FengYun-4A (FY-4A) satellite.

[0076] The interpolation in this example can be based on the latitude and longitude information of the FY-4A / GIIRS field of view point, and the bilinear interpolation method is used to interpolate the NCEP / FNL data to the FY-4A / GIIRS field of view point.

[0077] S2. Perform radiation transfer model simulation on the temperature profile, humidity profile and other model space data after background field data interpolation processing to obtain the corresponding satellite hyperspectral channel observation brightness temperature equivalent data as the channel simulation brightness temperature;

[0078] S3. Perform generalized quality control of hyperspectral observation brightness temperature data based on channel-simulated brightness temperature, taking into account outliers, to generate quality control data. Generalized quality control includes optimal channel selection based on quantum information entropy, bias correction based on artificial intelligence generalized integrated generative deep learning, and cloud detection based on the minimum residual method to eliminate cloud field point data.

[0079] Furthermore, unlike classical variational assimilation which eliminates “outliers” in the initial quality control process, the generalized quality control process of the present invention retains “outliers” so that they can be used in subsequent data assimilation processes, but reduces the contribution of “outliers” to the cost function.

[0080] Satellite hyperspectral data includes accurate data, erroneous data, and outliers. Erroneous data are directly discarded. However, a significant proportion of outlier data consists of accurate but extreme observations of important weather processes or events. Assimilating such data often improves the quality of the assimilated analysis field, thereby enhancing the accuracy of numerical weather forecasts (e.g., precipitation intensity and location, typhoon paths).

[0081] S4. Construct a quantum non-Gaussian incremental four-dimensional variational assimilation model that is compatible with Gaussian and non-Gaussian error distributions. In the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model, a dynamically adjustable weight contribution factor is introduced into the observation term.

[0082] Furthermore, the method of the present invention solves the problem that the cost function minimization iteration of the classical variational assimilation method is difficult to converge or fail due to the periodic fluctuations of the brightness temperature deviation of the hyperspectral infrared water vapor channel, thereby improving the accuracy of weather forecasts for wet processes such as typhoons and heavy rains.

[0083] S5. Based on the quality control data, the quantum non-Gaussian incremental four-dimensional variational assimilation model is solved to obtain the analytical field data.

[0084] Example 2:

[0085] This embodiment is further designed on the basis of the first embodiment in that: the optimal selection of satellite channels is a key technology for variational assimilation or inversion of hyperspectral atmospheric vertical sounders, which can reduce the ill-posedness of variational assimilation or inversion caused by redundant information observed by hyperspectral sounders, and further solve the channel "dimensionality curse" problem. The present invention proposes a "three-step method" for the optimal selection of satellite hyperspectral channels using quantum information entropy. The core method is based on the superposition, coherence, and entanglement of quantum states in quantum theory, and can accurately express the laws of objective existence. The specific methods for optimal channel selection considering quantum information entropy in this example include:

[0086] 1) Eliminate satellite channels with an absolute value of brightness temperature error exceeding 10 K in the radiative transfer model simulation. The simulated brightness temperature error is defined as the difference between the channel's observed brightness temperature and the channel's simulated brightness temperature.

[0087] 2) Based on the peak layer of the satellite channel weight function, that is, the atmospheric layer information at different altitudes detected by different satellite channels, the satellite channels are preliminarily selected to ensure that the selected channels can detect atmospheric information at different atmospheric altitude layers, and that channels are selected for each atmospheric altitude layer, thereby obtaining a preliminarily selected channel combination;

[0088] The weighting function (WF) characterizes the contribution of the atmosphere at different altitudes to the radiation of different satellite channels, and its peak position reflects the effective detection layer of the channel. WF is defined as:

[0089]

[0090] Wherein, τ represents the channel transmittance; v represents the channel center frequency; θ represents the satellite zenith angle; p represents the air pressure; represents partial derivative.

[0091] The channel initial selection steps based on the channel weight function are as follows:

[0092] 2.1) Eliminate multi-peak channels and retain single-peak channels to ensure the “singularity” of channel detection information.

[0093] 2.2) Layered channel selection. Within atmospheric altitude layers or pressure layers with the same radiation transmission pattern (e.g., 850hPa), prioritize channels with the largest weight function peak and steepest weight function to minimize channel detection uncertainty. Meteorological information from adjacent altitude layers is not included.

[0094] 2.3) Considering issues such as surface emissivity, channels with peak layers of the weight function located on the surface are removed to eliminate the influence of surface signals on the variational assimilation results.

[0095] 3) Select the optimal channel for the initially selected channel combination considering quantum information entropy.

[0096] Example 3:

[0097] This embodiment further enhances the first embodiment by implementing brightness temperature reconstruction based on the potential for interference during observation or signal transmission, thus potentially containing noise. This requires denoising the noise by performing brightness temperature reconstruction on the relevant data. Unlike the prior art's "linear transformation" principal component analysis, this embodiment proposes a "nonlinear" sampling, compressed sensing (CS) method for reconstructing the brightness temperature observed in the hyperspectral infrared channel. In this example, the channel-observed brightness temperature is obtained by performing nonlinear brightness temperature reconstruction based on compressed sensing on the brightness temperature data observed in the meteorological satellite's hyperspectral infrared channel.

[0098] The brightness temperature data observed by the hyperspectral infrared channel of the meteorological satellite is sparsely represented in the wavelet domain space, and the original brightness temperature signal is reconstructed by decoding. The specific expression is as follows:

[0099]

[0100] Among them, x true is the reconstructed original brightness temperature signal; y signal is the observed compressed brightness temperature signal, that is, the received brightness temperature signal; H CS is the structural operator; R signal is the brightness temperature signal error; W CS represents the transformation matrix; λ CS Represents the regularization parameter, which is set to 0.1 in the present invention.

[0101] Furthermore, the main idea of ​​nonlinear sampling compressed sensing is to compress and sample sparse signals, and select an appropriate reconstruction method to restore the compressed signal to the original signal after transmission, allowing the sparse state vector to be reconstructed through a small number of random measurements. In the specific implementation process, by introducing sparse representation in the wavelet domain space, better reconstruction results can be obtained with fewer random sampling samples.

[0102] Example 4:

[0103] This embodiment further enhances the first embodiment by adopting the classical theory of information entropy, i.e., from the perspective of information theory, the channel selection principle is to ensure that, given the number of selected channels, a good analysis field is obtained after variational assimilation of the selected channel subset (also called a channel combination). Channel selection is terminated when the entropy change amplitude no longer increases significantly or the number of channels in the selected channel combination reaches an initially set value (e.g., 100 channels).

[0104] Due to the superposition, coherence and entanglement of quantum states, it can accurately express the laws of objective existence, which leads to significant differences between quantum theory and traditional information expression methods, and "quantum" has excellent performance.

[0105] In this example, the specific method for selecting the optimal channel combination considering quantum information entropy includes:

[0106] 1) Normalize the brightness temperature sequence data corresponding to each channel of the hyperspectral infrared detector in the initially selected channel combination; for the brightness temperature sequence data X of the i-th channel in the channel combination i,entropy ={x i,entropy (j), j = 1, 2, ..., N}, the normalized observed brightness temperature sequence data of the i-th channel is recorded as Y i,entropy ={y i,entropy (j), j=1,2,...,N}; where xi,entropy (j) is the jth observed brightness temperature sequence data of the i-th channel; j is the serial number mark of the observed brightness temperature sequence data, and N is the total amount of observed brightness temperature sequence data; y i,entropy (j) is the normalized brightness temperature sequence data of the jth observation of the i-th channel;

[0107] 2) Perform quantum space reconstruction on the normalized sequence data of each channel to obtain the quantum space matrix corresponding to each channel; for the normalized observed brightness temperature sequence data Y of the i-th channel i,entropy , and its corresponding quantum space matrix The expression is as follows:

[0108]

[0109] Where, is the l-th reconstruction component in the quantum space matrix, l=1,2,...,K, K is the total number of reconstruction components, m is the embedding dimension; As the time window, the present invention selects 6 hours, which is also the time window of four-dimensional variational assimilation.

[0110] 3) Perform quantization on each reconstructed component in the quantum space matrix of each channel and calculate the quantum information entropy of each channel; Each reconstructed component in is quantized as follows:

[0111]

[0112] In the formula, |q b > represents the reconstruction component The qth state vector of l,k Represents the state vector |q b >, k=1,2,...,n, n is the number of state vectors, n=2 m , m is the constituent qubit;

[0113] The probability of the state vector appearing P i,k As the probability of an event occurring, we get the quantum space matrix Quantum information entropy in, k=1,2,...,n;

[0114] 4) Based on the quantum information entropy and variational assimilation background error covariance matrix of each channel, the quantum mutual information of each channel is calculated to form a multidimensional feature space composed of multiple groups of channels. The satellite channel corresponding to the maximum quantum mutual information is searched in the multidimensional feature space and selected as the optimal channel.

[0115] Furthermore, the quantum information entropy of the background error covariance matrix B after the variational assimilation of the transformed data is S(σ B )=-tr(σ B logσ B ), that is, σ B logσ B The trace of σ B represents the density matrix of |B>. Assuming that any channel brightness temperature series data (for example, ) is marked as c, and the quantum state is marked as |c>. Then the quantum information entropy of |c> is S(σ c )=-tr(σ c logσ c ). σ c represents the density matrix of |c>. We further obtain the density matrix ρ of the composite quantum composed of |c> and |B> cB Then, in quantum space, the quantum mutual information between the brightness temperature sequence data |c> observed by the hyperspectral infrared detector channel and the variational assimilation background error covariance matrix |B> is:

[0116]

[0117] The optimal channel selection problem of hyperspectral infrared detector based on quantum information entropy is transformed into the problem of finding the best or optimal feature subspace. An optimal d-dimensional feature set is searched in the original m-dimensional feature space composed of m channels. * The process is:

[0118] S(|c> * :|B)=max i S(|c>:|B).

[0119] Embodiment 5:

[0120] This embodiment further enhances the first embodiment by identifying that the brightness temperature deviations of satellite hyperspectral infrared detector channels contain various systematic errors. The effectiveness of bias correction determines the quality of the analysis field obtained after assimilating satellite data and influences the forecast accuracy of the final numerical weather forecast model. For example, these errors affect precipitation location and intensity, typhoon paths, and other factors. Satellite observation data itself also contains errors due to instrument sensitivity, calibration, and other factors. Therefore, bias correction must be performed before assimilating the brightness temperature of satellite hyperspectral infrared detector channels to eliminate or reduce the aforementioned systematic errors.

[0121] The international classic "offline" method of air mass bias correction (BC) is based on a set of forecast factors Correction is performed. If the brightness temperature deviation of each satellite channel j is B j , then we have the following relationship:

[0122]

[0123] Among them, the coefficient A ji and C j It is obtained by least squares fitting method from a large number of samples.

[0124] Different from the least squares fitting method used in "offline" deviation correction, this invention proposes a deviation correction method based on artificial intelligence generalized integrated generative deep learning. This invention rewrites the above formula. Assume that the brightness temperature deviation of a channel is denoted as y BC , the corresponding predictor is x BC , then the two can be expressed as the following relationship:

[0125] y BC =f(x BC )+v

[0126] Where f represents the forward mapping; f is a linear combination in the "offline" bias correction method; v represents a constant. BC represents the dependent variable; x BC Represents the independent variable.

[0127] In this example, the specific methods for bias correction based on artificial intelligence generalized integrated generative deep learning include:

[0128] 1) Select the prediction factors for bias correction, which include the model background field 1000-300hPa thickness, 200-50hPa thickness, model surface temperature, total water vapor, longitude and latitude information of the field of view point, and can also include surface wind field, terrain characteristics, tropopause height, cloud liquid water content and the square of temperature lapse rate as "prediction factors".

[0129] 2) For the brightness temperature equivalent data of all field-of-view points of each channel of the meteorological satellite hyperspectral infrared detector that need to be corrected, a deep model generated by combining convolutional neural network and deep neural network models is constructed as the basic model for ensemble learning, and generalized weighted ensemble machine learning is performed to find the optimal or suboptimal ensemble weights of the prediction results of the ensemble basic model. The objective minimization function of weighted ensemble machine learning is:

[0130]

[0131] Where n is the total number of samples; is the actual value of the value i to be corrected. The predicted value of the to-be-corrected value i by the basic model j is obtained based on the selected prediction factors as the input of the basic model. The method of using the basic model to predict the to-be-corrected value by the selected prediction factors is a prior art method and will not be described in detail here. is the integration weight corresponding to the base model j.

[0132] Example 6:

[0133] This embodiment is further designed on the basis of the first embodiment in that: in this embodiment, the specific method of eliminating the brightness temperature observation values ​​of all channels of the cloud field point based on the minimum residual method of cloud detection includes:

[0134] 1) According to the radiation transfer theory, the simulated radiation value of the meteorological satellite hyperspectral infrared detector channel is mathematically modeled. The simulated radiation value of the satellite channel g is Expressed as: Among them, N e Indicates the effective cloud cover at the viewing point; represents clear-sky radiation; The cloud top pressure is p c,g Cloudy radiation simulation; the channel radiation value can be directly converted into brightness temperature value through Planck function. This method is an existing technology and will not be described here;

[0135] 2) Calculate the deviation between the observed radiation value and the simulated radiation value of the hyperspectral infrared detector channel. For the observed radiation value of channel g, and simulated radiation values The deviation δ g Expressed as:

[0136] 3) Based on the n channel combinations obtained by the optimal channel selection based on quantum information entropy, the minimum residual method objective function is minimized at each field of view point to obtain the effective cloud amount N at the field of view point e and effective cloud top pressure p c , effective cloud top pressure p c The cloud top pressure is obtained by combining the n channels. w g is the comprehensive proportion. The minimum residual method objective function is defined as follows:

[0137]

[0138] 4) Based on effective cloud cover N e The cloud detection of the field point is performed with the set threshold (the set threshold is generally 0.1), and the effective cloud amount N e For viewpoints with brightness temperatures greater than a set threshold, all brightness temperature data of all channel combinations at that viewpoint are discarded. The present invention only performs assimilation research on the brightness temperature data of the retained viewpoint channel combinations.

[0139] Furthermore, the effective cloud cover N is minimized e and effective cloud top pressure pc The process is divided into two steps, as follows:

[0140] Step 1: According to the model pressure layer of the adopted radiation transfer model, the effective cloud top pressure is given and substituted into the minimum residual method objective function to perform minimization calculation to obtain the effective cloud amount N e :

[0141]

[0142] Step 2: Set the effective cloud cover N e After substituting the minimum residual method objective function for minimization calculation, the effective cloud top pressure p is obtained c , p c It is the cloud top pressure of n channel combinations, that is, the effective cloud top pressure value of a certain field of view point that is finally set together.

[0143] Embodiment seven:

[0144] This embodiment further enhances the first embodiment by emphasizing that the fundamental assumption of the classic four-dimensional variational method is that the distributions of the observation errors and control variables follow a Gaussian distribution. Bayes' theorem can be used to derive a cost function for Gaussian-distributed variables. The total cost function of the classic four-dimensional variational assimilation method is defined as follows:

[0145]

[0146] The cost functions of the background term and the observation term are defined as follows:

[0147]

[0148] Where, “background” represents the cost function of the background term; “observation” represents the cost function of the observation term; and the matrix B represents the background (including different meteorological variables such as temperature and humidity) error covariance matrix. i represents the observation error covariance matrix (the observation error here is the channel observation brightness temperature error of the satellite hyperspectral infrared detector). The superscripts -1 and T represent the matrix inversion and transposition marks respectively. x0 is the control variable, and the solved x0 is the analysis field; x b,0 represents the background initial conditions, including temperature, humidity, wind field, surface air temperature and other meteorological variables; y i represents the i-th observation value, i = 1, 2, ..., N, N is the total number of observations; H i To integrate the model M i,0 (·) The observation operator that maps the solution from the model space to the observation space; M i,0 (x0) is the nonlinear numerical prediction model that integrates the control variable x0 from the initial condition and initial time t = 0 to t = i, t is the time mark;

[0149] Furthermore, the total cost function of the classic four-dimensional variational assimilation is changed to the meteorological element incremental form, which is defined as follows:

[0150]

[0151] Where δx0 is the analysis increment, x0 is the control variable; is the state vector predicted by the numerical prediction model; the superscript f indicates forecast; B is the background error covariance matrix; R i is the observation error covariance matrix; d i is the brightness temperature increment, y i represents the i-th observation value; represents the brightness temperature value of the ith observation value simulated by the rapid radiative transfer model based on the pattern space, which is also the equivalent data of the ith observation value. H i To integrate the model M i,0 (·) The observation operator that maps the solution from the model space to the observation space; M i,0 (x0) is a nonlinear numerical prediction model that integrates the control variable x0 from the initial condition and initial time t = 0 to t = i, where t is the time mark; the matrix H i and M i,0 Respectively represent H i and M i,0 Linear representation of the tangent pattern.

[0152] The cost function J is a quadratic cost function in the incremental four-dimensional space δx0.

[0153] Through Taylor expansion and approximation, the calculation of δx0 is simplified to the following problem:

[0154] Aδx0=q

[0155] in, Represents the Hessian matrix in the incremental four-dimensional variational assimilation cost function.

[0156] In actual atmospheric science numerical weather prediction problems, the dimension of the A matrix usually exceeds 10 8 ~10 9 , making it difficult or impossible to calculate the inverse of the A matrix. However, using an iterative optimization process can avoid directly solving the matrix inverse. The conjugate gradient descent algorithm is an optimization algorithm suitable for quadratic cost functions. Classical incremental four-dimensional variational methods use a gradient-based quasi-Newton method to reduce the iterative cost of the cost function to update the analytical increment δx0.

[0157] The quantum non-Gaussian incremental four-dimensional variational assimilation proposed in this invention can take into account both the characteristics of deviation Gaussian and non-Gaussian, and uses quantum computing to accelerate the speed of solving the cost function and reduce the calculation time. In this example, the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model is:

[0158]

[0159] w(r i )=(1 / r i )·(dρ(r i ) / dr i )

[0160] Where δx0 is the analysis increment, x0 is the control variable; is the state vector predicted by the numerical prediction model; the superscript f indicates forecast; B is the background error covariance matrix; R i is the observation error covariance matrix; superscript T is the matrix transpose; superscript -1 is the matrix inverse; d i is the brightness temperature increment, y i represents the i-th observation value, i = 1, 2, ..., N, N is the total number of observations; represents the brightness temperature value of the ith observation value simulated by the rapid radiative transfer model based on the pattern space, that is, the equivalent data of the ith observation value, also known as the channel simulated brightness temperature; H i To integrate the model M i,0 (·) The observation operator that maps the solution from the model space to the observation space; for satellite data, H i Also known as fast radiation transfer mode. i Realize the projection mapping or conversion of pattern space variables (e.g., temperature and other meteorological variables) to observation space (e.g., brightness temperature observed by satellite hyperspectral infrared detector channel); M i,0 (x0) is a nonlinear numerical prediction model that integrates the control variable x0 from the initial condition and initial time t = 0 to t = i, where t is the time mark; the matrix H i and M i,0 Respectively represent H i and M i,0 Linear representation of the tangent pattern; r i is the brightness temperature increment amplitude, r i =d i / σ i , σ i is the observation error of channel i; w(r i ) is the weight contribution factor; ρ(r i ) is the M-estimation cost function.

[0161] Furthermore, the core of the new M-estimation norm lies in the construction of the cost function. This new M-estimation cost function must satisfy the following nine basic properties. Furthermore, the weight function of the M-estimation cost function must satisfy the following five basic properties, as shown in Table 1. These properties serve as the foundation for constructing the new M-estimation norm.

[0162] Table 1M-Basic properties of estimated cost function and weight function

[0163]

[0164]

[0165] According to the basic properties of Table 1, the process of constructing the new M-estimation norm in the present invention through mathematical equations and integration is as follows:

[0166]

[0167] The superscript ' represents the derivative mark.

[0168] Integrate both sides of the above equation separately, according to its basic properties (see Table 1), we have:

[0169]

[0170] It should be noted that when constructing a new norm of the M-estimation method, it is necessary to count and analyze the "adjustment scale" based on actual conditions.

[0171] In order to facilitate the calculation using quantum models, the cost function of the M-estimate quantum non-Gaussian incremental four-dimensional variational assimilation method is approximated. The specific expression is as follows:

[0172]

[0173] in,

[0174]

[0175] in, The tilde indicates a linear approximation. After derivation and analysis, this optimization only operates on the analysis increment, as shown below:

[0176]

[0177] Embodiment 8:

[0178] This embodiment is further designed on the basis of Example 7 in that: compared with the classical incremental four-dimensional variational assimilation method that requires the calculation of the cost function and its gradient, the quantum annealing based on quantum computing only needs to calculate the cost function. However, for quantum annealing, the cost function is represented by a binary variable (i.e., 0 or 1). The basic principle of the quantum annealing algorithm is to search the solution space through the quantum superposition and quantum entanglement of quantum bits. In quantum computing, quantum bits can be in a superposition state of multiple states and can process multiple possible solutions at the same time. Quantum entanglement is the interaction between quantum bits, which can make the optimization of the search solution space more efficient. The quantum annealing algorithm uses the quantum tunneling effect to enable quantum to penetrate potential barriers with higher energy than its own, thereby enabling the algorithm to get rid of local extreme values ​​and approach the global optimum with a higher probability.

[0179] The specific methods for solving the quantum non-Gaussian increment four-dimensional variational assimilation model in this example include:

[0180] 1) Using quantum annealing driven incremental four-dimensional non-Gaussian quantum data assimilation method, the real number representation is performed by Q qubits, and the analysis increment δx0 is quantized by the mapping matrix G:

[0181]

[0182] g T =[-2 Q-1 ,2 Q-2 ,2 Q-3 ,...,2 1 ,2 0 ]

[0183]

[0184] Where β is an adjustable scaling parameter with a value of 0.2; b is a binary vector whose elements are all 0 or 1, and 0 is a vector whose elements are all 0; g is the quantization of a real number; Q is a natural number;

[0185] 2) Based on the quantized analytical increments, the cost function of the quantum non-Gaussian increment four-dimensional variational assimilation model is solved in quantum space:

[0186]

[0187] Where,

[0188]

[0189] Where Ham represents the Hamiltonian; w(·) represents the weight function of the constructed M-estimated new norm; A and u are as follows:

[0190] 3) Further transformation or conversion of the above variables Then, the quantum annealing method is used to minimize Get the best analysis field.

[0191] Quantum Annealing (QA) is an optimization algorithm based on the principles of quantum mechanics. It aims to solve complex optimization problems by simulating the annealing process of quantum systems. Its core idea is to use quantum tunneling and quantum superposition to enable the system to quickly traverse high-energy barriers, thereby finding the global optimal solution to the problem.

[0192] Basic Principle: The quantum annealing algorithm simulates the annealing process in solid-state physics. By gradually changing the system's Hamiltonian, it slowly evolves the initial state to the target state, thereby finding the optimal solution to the problem. Unlike the classic simulated annealing algorithm, quantum annealing utilizes the quantum tunneling effect, allowing the system to directly pass through high-energy barriers, avoiding the local optimal solution that may be encountered in traditional annealing methods.

[0193] Furthermore, unlike existing variational data assimilation methods, which use gradient optimization to reduce the cost function to obtain the optimal analysis field, variational data assimilation methods require significant computational resources and are relatively time-consuming within numerical weather forecasting models due to the high number of iterations required to sufficiently reduce the cost function in actual numerical weather forecasting. To address this, the present invention utilizes a quantum annealing-driven solution, which can accelerate the solution of the variational data assimilation cost function, due to quantum effects (e.g., tunneling, superposition, and entanglement).

[0194] Embodiment 9:

[0195] This embodiment is further designed based on the first embodiment in that: this embodiment also includes the evaluation of the accuracy of the analysis field data, the timeliness of the solution, and the accuracy of the analysis field for improving the wet process numerical weather forecast.

[0196] Embodiment 10:

[0197] The present invention provides a system for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing, the system comprising a first data processing module, a second data processing module, a third data processing module, and a fourth data processing module;

[0198] The first data processing module is used to simulate the radiation transmission mode of the meteorological satellite hyperspectral channel data after the background field data interpolation processing, and obtain the corresponding satellite hyperspectral channel observation brightness temperature equivalent data as the channel simulation brightness temperature;

[0199] The second data processing module is used to perform generalized quality control of the channel simulated brightness temperature based on meteorological satellite hyperspectral data, taking into account outliers, to generate quality control data. Generalized quality control includes optimal channel selection considering quantum information entropy, bias correction based on artificial intelligence generalized ensemble generative deep learning, and cloud detection and removal based on the minimum residual method.

[0200] The third data processing module is used to construct a quantum non-Gaussian incremental four-dimensional variational assimilation model that is compatible with Gaussian and non-Gaussian error distributions. The observation term in the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model introduces a dynamically adjustable weight contribution factor;

[0201] The fourth data processing module is used to solve the quantum non-Gaussian incremental four-dimensional variational assimilation model based on quality control data to obtain analysis field data.

[0202] Example 11:

[0203] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the method of any of the above embodiments.

[0204] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above embodiments are implemented.

[0205] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing, characterized in that: The method comprises: The radiation transfer model is simulated for the meteorological satellite hyperspectral channel data after background field data interpolation processing to obtain the corresponding satellite hyperspectral channel observation brightness temperature equivalent data as the channel simulation brightness temperature; Performing generalized quality control of meteorological satellite hyperspectral data considering outliers on the channel simulated brightness temperature to form quality control data, wherein the generalized quality control includes optimal channel selection considering quantum information entropy, deviation correction based on artificial intelligence generalized ensemble generative deep learning, and cloud detection and removal based on the minimum residual method; A quantum non-Gaussian incremental four-dimensional variational assimilation model compatible with Gaussian and non-Gaussian error distributions is constructed, wherein a dynamically adjustable weight contribution factor is introduced into the observation term in the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model; Based on the quality control data, the quantum non-Gaussian incremental four-dimensional variational assimilation model is solved to obtain analysis field data.

2. The method for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing according to claim 1 is characterized in that: The specific method for optimal channel selection considering quantum information entropy includes: 1) Eliminate satellite channels whose absolute value of the brightness temperature error of the channel simulation in the radiative transfer model exceeds 10K; the channel simulation brightness temperature error is defined as the difference between the channel observation brightness temperature and the channel simulation brightness temperature; 2) Based on the peak layer of the satellite channel weight function, that is, the atmospheric layer information at different altitudes detected by different satellite channels, the satellite channels are preliminarily selected to obtain a preliminarily selected channel combination; 3) Select the optimal channel for the initially selected channel combination considering quantum information entropy.

3. The method for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing according to claim 2 is characterized in that: The channel observation brightness temperature is obtained by performing nonlinear brightness temperature reconstruction based on compressed sensing on the brightness temperature data observed by the meteorological satellite hyperspectral infrared channel. The specific method is as follows: The brightness temperature data observed by the hyperspectral infrared channel of the meteorological satellite is sparsely represented in the wavelet domain space, and the original brightness temperature signal is reconstructed by decoding. The specific expression is as follows: Among them, x true is the reconstructed original brightness temperature signal; y signal is the observed compressed brightness temperature signal, that is, the received brightness temperature signal; H CS is the structural operator; R signal is the brightness temperature signal error; W CS represents the transformation matrix; λ CS represents the regularization parameter.

4. The method for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing according to claim 2 is characterized in that: The specific method for performing optimal channel selection considering quantum information entropy on the initially selected channel combination includes: 1) Normalizing the observed brightness temperature sequence data corresponding to each channel of the hyperspectral infrared detector in the preliminarily screened channel combination; For the observed brightness temperature sequence data X of the i-th channel in the channel combination i,entropy ={x i,entropy (j), j = 1, 2, ..., N}, the normalized observed brightness temperature sequence data of the i-th channel is recorded as Y i,entropy ={y i,entropy (j), j=1,2,...,N}; where x i,entropy (j) is the jth observed brightness temperature sequence data of the i-th channel; j is the serial number mark of the observed brightness temperature sequence data, and N is the total amount of observed brightness temperature sequence data; y i,entropy (j) is the normalized brightness temperature sequence data of the jth observation of the i-th channel; 2) Perform quantum space reconstruction on the normalized sequence data of each channel to obtain the quantum space matrix corresponding to each channel; for the normalized observed brightness temperature sequence data Y of the i-th channel i,entropy , and its corresponding quantum space matrix The expression is as follows: Where, is the l-th reconstruction component in the quantum space matrix, l=1,2,...,K, K is the total number of reconstruction components, m is the embedding dimension; is the time window; 3) Quantize each reconstructed component in the quantum space matrix of each channel and calculate the quantum information entropy of each channel; Each reconstructed component in is quantized as follows: In the formula, |q b > represents the reconstruction component The qth state vector of l,k Represents the state vector |q b >, k=1,2,...,n, n is the number of state vectors, n=2 m , m is the constituent qubit; The probability of the state vector appearing P i,k As the probability of an event occurring, we get the quantum space matrix Quantum information entropy in, 4) Based on the quantum information entropy and variational assimilation background error covariance matrix of each channel, the quantum mutual information of each channel is calculated to form a multidimensional feature space composed of multiple groups of channels, and the satellite channel corresponding to the maximum quantum mutual information is searched in the multidimensional feature space and selected as the optimal channel.

5. The method for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing according to claim 2 is characterized in that: The specific method of deviation correction based on artificial intelligence generalized integrated generative deep learning includes: 1) Selecting bias-corrected predictors; 2) For the brightness temperature equivalent data of all field-of-view points of each channel of the meteorological satellite hyperspectral infrared detector that need to be corrected, a deep model generated by combining convolutional neural network and deep neural network models is constructed as the basic model for ensemble learning, and generalized weighted ensemble machine learning is performed to find the optimal or suboptimal ensemble weight for integrating the prediction results of the basic model. The objective minimization function of the weighted ensemble machine learning is: Where n is the total number of samples; is the actual value of the value to be corrected i; is the predicted value of the to-be-corrected value i by the basic model j, and the value of the predicted value is obtained based on the selected prediction factors as the input of the basic model; is the integration weight corresponding to the base model j.

6. The method for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing according to claim 3 is characterized in that: The specific method of cloud detection and elimination based on the minimum residual method includes: 1) According to the radiation transfer theory, the simulated radiation value of the meteorological satellite hyperspectral infrared detector channel is mathematically modeled. The simulated radiation value of the satellite channel g is Expressed as: Among them, N e Indicates the effective cloud cover at the viewing point; represents clear-sky radiation; The cloud top pressure is p c,g Cloudy radiation simulation at 1000 Hz; 2) Calculate the deviation between the observed radiation value and the simulated radiation value of the hyperspectral infrared detector channel. For the observed radiation value of channel g, and simulated radiation values The deviation δ g Expressed as: 3) Based on the n channel combinations obtained by the optimal channel selection based on quantum information entropy, the minimum residual method objective function is minimized at each field of view point to obtain the effective cloud amount N at the field of view point e and effective cloud top pressure p c , the minimum residual method objective function is defined as follows: 4) Based on effective cloud cover N e The cloud detection of the field point is performed with the set threshold. For the effective cloud amount N e For a viewpoint whose brightness temperature is greater than the set threshold, all brightness temperature data of all channel combinations at the viewpoint are discarded.

7. The method for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing according to claim 1 is characterized in that: The cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model is: w(r i )=(1 / r i )·(dρ(r i ) / dr i ) Where δx0 is the analysis increment, x0 is the control variable, and the solved x0 is the analysis field; is the state vector predicted by the numerical prediction model; the superscript f indicates forecast; B is the background error covariance matrix; R i is the observation error covariance matrix; superscript T is the matrix transpose; superscript -1 is the matrix inverse; d i is the brightness temperature increment, y i represents the i-th observation value, i = 1, 2, ..., N, N is the total number of observations; represents the brightness temperature value of the ith observation value simulated by the rapid radiative transfer model based on the pattern space, which is also the equivalent data of the ith observation value. H i To integrate the model M i,0 (·) The observation operator that maps the solution from the model space to the observation space; M i,0 (x0) is a nonlinear numerical prediction model that integrates the control variable x0 from the initial condition and initial time t = 0 to t = i, where t is the time mark; the matrix H i and M i,0 Respectively represent H i and M i,0 Linear representation of the tangent pattern; r i is the brightness temperature increment amplitude, r i =d i / σ i , σ i is the observation error of channel i; w(r i ) is the weight contribution factor; ρ(r i ) is the M-estimation cost function.

8. The method for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing according to claim 7, characterized in that: The specific method for solving the quantum non-Gaussian incremental four-dimensional variational assimilation model includes: 1) Using quantum annealing driven incremental four-dimensional non-Gaussian quantum data assimilation method, the real number representation is performed by Q qubits, and the analysis increment δx0 is quantized by the mapping matrix G: g T ＝[-2 Q-1 ,2 Q-2 ,2 Q-3 ,...,2 1 ,2 0 ] Where β is an adjustable scaling parameter; b is a binary vector whose elements are all 0 or 1, 0 is a vector whose elements are all 0; g is the quantization of a real number; Q is a natural number; 2) Based on the quantized analytical increments, the cost function of the quantum non-Gaussian increment four-dimensional variational assimilation model is solved in quantum space: Where, Where Ham represents the Hamiltonian; w(·) represents the weight function of the constructed M-estimated new norm; A and u are as follows: 3) Minimization through quantum annealing method Get the best analysis field.

9. The method for assimilating meteorological satellite hyperspectral error non-Gaussian distribution data based on quantum computing according to claim 1, characterized in that: It also includes the evaluation of the accuracy and timeliness of the analysis field data and the impact of the analysis field on improving the accuracy of numerical weather prediction for wet processes.

10. A system for assimilating non-Gaussian distribution data of meteorological satellite hyperspectral errors based on quantum computing, characterized in that: The system includes a first data processing module, a second data processing module, a third data processing module, and a fourth data processing module; The first data processing module is used to simulate the radiation transmission mode of the meteorological satellite hyperspectral channel data after the background field data interpolation processing, and obtain the corresponding satellite hyperspectral channel observation brightness temperature equivalent data as the channel simulation brightness temperature; The second data processing module is used to perform generalized quality control of meteorological satellite hyperspectral data considering outliers on the channel simulated brightness temperature to form quality control data, wherein the generalized quality control includes optimal channel selection considering quantum information entropy, deviation correction based on artificial intelligence generalized integrated generative deep learning, and cloud detection and elimination based on the minimum residual method; The third data processing module is used to construct a quantum non-Gaussian incremental four-dimensional variational assimilation model that is compatible with Gaussian and non-Gaussian error distributions, and introduces a dynamically adjustable weight contribution factor into the observation term in the cost function of the quantum non-Gaussian incremental four-dimensional variational assimilation model; The fourth data processing module is used to solve the quantum non-Gaussian incremental four-dimensional variational assimilation model based on the quality control data to obtain analysis field data.

Images