Multi-channel microwave radiation atmospheric water vapor and liquid water content testing method and system

Abstract

The present application belongs to the technical field of foundation atmospheric microwave remote sensing test, and specifically discloses a multi-channel microwave radiation atmospheric water vapor and liquid water content test method and system, which comprises the following steps: collecting atmospheric radiation signals and multi-channel calibration reference signals and converting them into voltage signals; using four sets of calibration reference voltage signals to solve the nonlinear parameters of the system and inversely mapping to obtain the atmospheric brightness temperature signal eliminating distortion; using the cross-medium reference signal to demarcate the liquid water cutoff boundary, and using the frequency domain of the brightness temperature signal to solve the decoupling to obtain the orthogonal water vapor absorption characteristic signal and the liquid water absorption characteristic signal; inputting the two characteristic signals into a neural network model respectively to obtain the vertical distribution concentration signals of water vapor and liquid water, and integrating to obtain the columnar total amount parameters. The present application solves the technical problems of existing microwave radiometers, such as the inability to correct nonlinear distortion due to linear calibration, and the inversion of non-adaptive caused by the characteristic mixing of water vapor and liquid water.

Description

Technical Field

[0001] This invention relates to the field of ground-based atmospheric microwave remote sensing testing technology, specifically to a method and system for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation. Background Technology

[0002] Ground-based microwave radiometers are passive remote sensing devices used to detect atmospheric water vapor and liquid water thermal radiation, obtaining meteorological parameters through inversion. Existing equipment generally employs a two-point linear calibration equation, based on the assumption of an absolutely linear response of an ideal detector diode. Under a wide range of equivalent noise temperature inputs, the low-noise amplifier and square-law detector exhibit thermodynamic nonlinear distortion. Two-point linear fitting (using only high and low temperature reference points) cannot match the nonlinear response of the devices, resulting in a parabolic truncation error at the center of the dynamic observation interval. This atmospheric brightness temperature measurement error varies with the square of the input equivalent noise temperature; actual measurements show this error can reach 3–5 K, causing the final converted atmospheric brightness temperature to contain system-level hardware contamination.

[0003] Meanwhile, water vapor molecules exhibit spin transition absorption peaks, while liquid water droplets show broadband monotonous continuous absorption of microwaves. Under precipitation or deep cloud cover conditions, the narrow-band absorption characteristics of water vapor and the broadband absorption characteristics of liquid water undergo deep nonlinear physical aliasing within the same detection frequency band. Existing algorithms do not strip the overlapping energy at the physical input level, directly inputting the interference-carrying coupled brightness temperature into statistical models such as neural networks. The coupling characteristics destroy the independence of the parameter space, leading to unfitness of the inversion equation, negative oscillations in the water vapor inversion results when the liquid water content surges, and a sharp amplification of the total integral deviation. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation. This invention solves the technical problems of existing ground-based microwave radiometers, such as the inability to correct nonlinear distortion of detector devices due to two-point linear calibration and the nonlinear aliasing of microwave absorption characteristics of water vapor and liquid water leading to unfitted inversion.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A multi-channel microwave radiation method for testing atmospheric water vapor and liquid water content includes: Step S1: Control the multi-channel microwave receiver to perform time-division multiplexing, and collect atmospheric natural microwave radiation signals, pure high-temperature calibration radiation signals, pure low-temperature calibration radiation signals, constant additional noise radiation signals, ground real-world environmental reference signals, and cloud base height characterization signals. Perform down-conversion, detection, and analog-to-digital conversion on the collected radio frequency signals to obtain composite test voltage signals, first calibration reference voltage signals, second calibration reference voltage signals, third calibration reference voltage signals, and fourth calibration reference voltage signals. Step S2: Using the first calibration reference voltage signal, the second calibration reference voltage signal, the third calibration reference voltage signal, and the fourth calibration reference voltage signal, calculate the system equivalent noise temperature parameter, gain coefficient, and nonlinear factor. Based on the system equivalent noise temperature parameter, gain coefficient, and nonlinear factor, perform inverse mapping on the composite test voltage signal to obtain the atmospheric absolute physical brightness temperature signal after removing second-order distortion, and divide it into K-band physical brightness temperature signal and V-band physical brightness temperature signal. Step S3: Establish a background atmospheric thermodynamic temperature profile using the V-band physical brightness temperature signal and the ground-based environmental reference signal; delineate the upper limit cutoff boundary for liquid water accumulation using the cloud base height characterization signal; and decouple the K-band physical brightness temperature signal by frequency domain differentiation using the background atmospheric thermodynamic temperature profile and the upper limit cutoff boundary for liquid water accumulation as constraints to obtain a narrowband water vapor absorption characteristic signal and a broadband liquid water absorption characteristic signal. Step S4: Combine the narrow-band absorption characteristic signal of water vapor and the broadband absorption characteristic signal of liquid water with the ground real-time environmental reference signal, respectively, and input them into the pre-trained neural network model to invert and obtain the vertical spatial distribution concentration signal of water vapor and the vertical spatial distribution concentration signal of liquid water. Perform vertical integration on the vertical spatial distribution concentration signal of water vapor and the vertical spatial distribution concentration signal of liquid water to obtain the columnar total water vapor integral parameter signal and the columnar total liquid water integral parameter signal.

[0006] Furthermore, the physical process of generating the first calibration reference voltage signal in step S1 is as follows: the cascaded gain of the receiving system performs frequency integration on the input energy density within the operating bandwidth, wherein the input energy density is equivalent to the voltage response corresponding to the sum of the absolute physical temperature value of the built-in high-temperature blackbody source and the equivalent noise temperature parameter of the system.

[0007] Furthermore, in step S1, the third calibration reference voltage signal is generated by superimposing the pure high-temperature calibration radiation signal and the constant additional noise radiation signal at the RF link port via a receiving link; the fourth calibration reference voltage signal is generated by superimposing the pure low-temperature calibration radiation signal and the constant additional noise radiation signal at the RF link port via a receiving link.

[0008] Furthermore, the process of calculating the system's equivalent noise temperature parameters, gain coefficient, and nonlinear factor in step S2 includes: A second-order microwave transfer equation system containing the equivalent noise temperature parameters, gain coefficient, and nonlinear factor of the system is constructed. Algebraic difference operations are performed using the first, second, third, and fourth calibration reference voltage signals, and the equivalent noise temperature parameters, gain coefficient, and nonlinear factor of the system are obtained by iterative solution.

[0009] Furthermore, the iterative solution employs the Gauss-Newton method and constructs the Jacobian matrix. The convergence condition for the iteration is that the sum of squared residuals is less than 1 / 2. .

[0010] Furthermore, the establishment of the background atmospheric thermodynamic temperature profile in step S3 includes: using the ground temperature and ground pressure in the ground real-time environmental reference signal as boundary conditions, and combining the V-band physical brightness temperature signal to obtain the background atmospheric thermodynamic temperature profile through statistical regression recursion.

[0011] Furthermore, when the cloud base height characterization signal is a cloud base infrared radiation signal, the upper limit cutoff boundary for liquid water accumulation includes: In the background atmospheric thermodynamic temperature profile, find the height at which the equivalent infrared brightness temperature of the cloud base infrared radiation signal is isothermal, and then superimpose empirical thickness compensation.

[0012] Furthermore, when the cloud base height characterization signal is a laser backscattered echo profile signal, defining the upper limit cutoff boundary for liquid water accumulation includes: performing spatial gradient analysis on the laser backscattered echo profile signal, and taking the lowest height point among the height points that satisfy the condition that the backscattering coefficient exceeds the threshold and the local gradient is at its maximum value.

[0013] Furthermore, the frequency domain decoupling in step S3 is as follows: perform the second-order frequency partial derivative on the optical thickness spectrum obtained by converting the physical brightness temperature signal of the K-band, separate the water vapor narrowband absorption characteristic signal, subtract the water vapor component from the optical thickness spectrum and use the upper limit truncation boundary of liquid water accumulation to spatially zero out the signal, and obtain the liquid water broadband absorption characteristic signal.

[0014] Furthermore, the neural network model mentioned in step S4 is a multilayer perceptron network or a radial basis function neural network. The output layer of the neural network model contains 58 nodes corresponding to unequal hierarchical layers from the ground to an altitude of 10,000 meters, and adopts a linear rectified activation function.

[0015] Furthermore, the narrowband absorption feature signal of water vapor and the broadband absorption feature signal of liquid water are respectively input into two multilayer perceptron branches or radial basis function neural network branches with identical structures and independent weights.

[0016] Furthermore, the vertical integration in step S4 is a discrete cross-sectional depth integration based on the unequal vertical layer thickness from the ground to an altitude of 10,000 meters.

[0017] In addition, the present invention also discloses a multi-channel microwave radiation atmospheric water vapor and liquid water content testing system for performing the multi-channel microwave radiation atmospheric water vapor and liquid water content testing method described above, including: Antenna, used to receive natural atmospheric microwave radiation signals; Built-in high-temperature blackbody source to provide the upper boundary for thermodynamic calibration; A pure cryogenic reference source is used to provide a lower boundary for thermodynamic calibration. The pure cryogenic reference source is an external liquid nitrogen cryogenic source or an active semiconductor cooling target. Solid-state noise source used to generate a constant additional noise radiation signal; A multi-channel microwave receiver is used to perform microwave frequency down-conversion and analog-to-digital discretization conversion; Meteorological condition sensor array, used to collect ground-based environmental reference signals; A cloud base detection unit is used to acquire cloud base height characterization signals. The cloud base detection unit is an infrared temperature measurement probe or a micro-pulse lidar.

[0018] Furthermore, the built-in high-temperature blackbody source uses a carbon-doped silicone pyramidal array as the radiation substrate and is equipped with constant temperature control; the pure low-temperature reference source is specifically an external liquid nitrogen low-temperature source that uses the physical properties of boiling phase change to clamp the reference temperature, or an active semiconductor cooling target that uses the Peltier cascade pump effect to provide a cold-end reference.

[0019] Compared with the prior art, the present invention has the following beneficial effects: This invention addresses the problem in existing technologies where two-point linear calibration cannot match the nonlinear response of detector devices. By introducing four thermodynamic temperature reference points with absolute physical definitions, a second-order microwave transmission equation set is established and the system's nonlinear calibration parameter set is solved. This fully characterizes the nonlinear transmission characteristics of the receiving link at the physical level, effectively eliminating the second-order distortion component from the inversely mapped atmospheric absolute physical brightness temperature signal, thus eliminating system-level hardware contamination and improving the accuracy of atmospheric brightness temperature measurement.

[0020] This invention addresses the problem of poorly determined inversion caused by the deep nonlinear aliasing of narrow-band water vapor absorption characteristics and broadband liquid water absorption characteristics within the same detection frequency band. It introduces a cross-medium physical reference signal to define the upper limit truncation boundary for liquid water accumulation and utilizes a second-order partial derivative operator in the frequency domain to orthogonally decouple the K-band optical thickness spectrum. This forcibly separates the gradually continuous absorption component of liquid water from the sharp Lorentzian linear absorption component of water vapor, ensuring high orthogonality of the feature vectors entering the inversion model. This eliminates the divergence problem of physical anomalies in water vapor inversion results when liquid water content surges, enhancing the stability and reliability of the inversion process under complex meteorological conditions. Furthermore, this invention employs a structurally parallel but weight-independent multilayer perceptron network or radial basis function neural network to process the orthogonally decoupled water vapor and liquid water feature signals respectively, avoiding feature cross-contamination and improving the layering accuracy of the three-dimensional concentration profile inversion.

[0021] Meanwhile, the built-in high-temperature blackbody source uses a carbon-doped silicone cone array with closed-loop constant temperature control to ensure the long-term stability of the high-temperature reference; the pure low-temperature reference source can use an external liquid nitrogen low-temperature source to clamp the temperature by boiling phase change, or use an active semiconductor cooling target surface to achieve liquid nitrogen-free deployment, so that the system can adapt to different application scenarios from reference observation stations to remote unattended stations, reducing the complexity of operation and maintenance. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is the main flowchart of the method described in this invention.

[0024] Figure 2 This is a flowchart of the method for determining the upper limit cutoff boundary of liquid water accumulation using micropulse lidar according to the present invention.

[0025] Figure 3 This is a sub-flowchart of the present invention for parallel inversion of the vertical spatial distribution concentration signals of water vapor and liquid water using a multilayer perceptron network. Detailed Implementation

[0026] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0027] The following is in conjunction with the appendix Figures 1-3 The embodiments of the present invention will be described in detail below.

[0028] Example 1: This example discloses a multi-channel microwave radiation atmospheric water vapor and liquid water content testing system, including: Antenna, used to receive natural atmospheric microwave radiation signals; Built-in high-temperature blackbody source to provide the upper boundary for thermodynamic calibration; A pure cryogenic reference source is used to provide a lower boundary for thermodynamic calibration. The pure cryogenic reference source is an external liquid nitrogen cryogenic source or an active semiconductor cooling target. Solid-state noise source used to generate a constant additional noise radiation signal; A multi-channel microwave receiver is used to perform microwave frequency down-conversion and analog-to-digital discretization conversion; Meteorological condition sensor array, used to collect ground-based environmental reference signals; A cloud base detection unit is used to acquire cloud base height characterization signals. The cloud base detection unit is an infrared temperature measurement probe or a micro-pulse lidar.

[0029] Meanwhile, this embodiment also discloses a multi-channel microwave radiation method for testing atmospheric water vapor and liquid water content. The external meteorological reference signal relied upon in this embodiment is the cloud base infrared radiation signal, and the specific analysis process is as follows: Step S1: Signal Acquisition and Discretization Conversion; At the start of the test, the microwave radiation test hardware system is powered on and initialized, controlling the multi-channel microwave receiver, meteorological condition sensor array, and infrared temperature probe to perform time-synchronized acquisition, with a time synchronization accuracy better than 1 ms. Using the rising edge of the switching level output by the detection and control unit as the global timing reference, all sensors are synchronously triggered to sample. Each sensor synchronously acquires physical field information. The meteorological condition sensor array acquires ground-level physical parameters in real time, generating a ground-based environmental reference signal, which includes ground temperature, ground humidity, and ground air pressure signals. The independently operating infrared temperature probe scans towards the zenith, generating cloud base infrared radiation signals.

[0030] The multi-channel microwave receiver incorporates a single-pole five-throw RF microwave switch array. The detection and control unit outputs logic levels according to time-division multiplexing (switching period set to 100 ms, high level is 3.3 V, low level is 0 V), causing the RF microwave switch array to sequentially switch to five physical states: state one, state two, state three, state four, and state five. Each state maintains stable sampling for 20 ms before switching, outputting the following respectively: State 1: The RF microwave switch array is activated by the built-in high-temperature blackbody source. This source, controlled by proportional-integral-derivative (PID) logic, maintains a constant temperature and outputs a pure high-temperature calibration radiation signal. This radiation signal is then amplified by low noise, mixed by a local oscillator, amplified by intermediate frequency (IF), detected by square-law detection, and converted from analog to digital to obtain the first calibration reference voltage signal. Its physical integral relationship satisfies:

[0031] in This indicates the value of the first calibration reference voltage signal. This indicates the upper cutoff frequency of the receiver channel. This indicates the lower cutoff frequency of the receiver channel. Indicates the frequency of the receiving system. Cascaded power-to-voltage conversion gain (unit: V / W). Represents the Boltzmann constant. This represents the absolute physical temperature value of the built-in high-temperature blackbody source. This represents the internal equivalent noise temperature parameter of the receiving system, equivalent to the input terminal. This represents the second-order nonlinear intrinsic distortion factor of the detector diode.

[0032] State 2: The RF microwave switch array is connected to an external liquid nitrogen cryogenic source, which is located at the boiling phase transition point of liquid nitrogen, generating a pure cryogenic calibration radiation signal. After passing through the same receiving link, a second calibration reference voltage signal is output.

[0033] in This indicates the value of the second calibration reference voltage signal. This indicates the absolute physical temperature value of the external liquid nitrogen cryogenic source.

[0034] State 3: The RF microwave switch array is activated by the built-in high-temperature blackbody source, and simultaneously, the directional coupler injects the constant additional noise radiation signal generated by the solid-state noise source into the main RF link. After energy superposition, the signal is detected and output as the third calibration reference voltage signal.

[0035] in This indicates the value of the third calibration reference voltage signal. This represents the equivalent noise-added temperature value corresponding to a constant additional noise radiated signal.

[0036] State 4: The RF microwave switch array is connected to an external liquid nitrogen cryogenic source and a constant additional noise radiation signal is injected simultaneously, outputting the fourth calibration reference voltage signal:

[0037] in This indicates the value of the fourth calibration reference voltage signal.

[0038] State 5: The RF microwave switch array is connected to the antenna. The atmospheric natural microwave radiation signal received by the antenna is fed into the receiver, which outputs a composite test voltage signal.

[0039] in This represents the value of the composite test voltage signal. This represents the apparent physical brightness temperature of the atmosphere at the antenna aperture.

[0040] All five sets of voltage signals generated above are used in step S2; the ground temperature signal, ground humidity signal, and ground air pressure signal are transmitted to steps S3 and S4; the cloud base infrared radiation signal is transmitted to step S3.

[0041] Step S2: Four-state nonlinear reference reconstruction; Step S2 uses the four sets of reference voltage signals input in Step S1 to perform nonlinear error calculation. Based on the equivalent bandwidth approximation theory, the integral equation is simplified to a discrete algebraic equation at the equivalent center frequency point: First base state algebraic equation:

[0042] in To integrate the second-order nonlinear factor parameters within the bandwidth (unit: V·K) - ²), First-order linear gain parameter (unit: V·K) - ¹), The DC bias constant parameter (unit: V) specifically refers to the inherent DC bias voltage of the detector diode in the receiving link, which is determined by the receiver hardware itself and can be measured through no-load calibration. In this embodiment... .

[0043] Second base state algebraic equation:

[0044] Third base state algebraic equation:

[0045] Fourth base state algebraic equation:

[0046] To solve the nonlinear calibration parameter set of the system ( , , The bias constant parameter is eliminated by algebraic difference method. .

[0047] Define the voltage increment at high temperature and voltage increment at low temperature caused by the injection of a constant additional noise radiation signal:

[0048]

[0049] Derive the nonlinear correction ratio function:

[0050] in It represents the physical slope deviation coefficient of the system's nonlinear response within the dynamic range.

[0051] Further utilize the difference results to Expressed as a nonlinear equation with three unknown parameters:

[0052] This equation, combined with the four reference state algebraic equations, forms a system of equations concerning... , , The overdetermined system of equations. Iterative solution using the Jacobian matrix, constructed as follows (with parameters adjusted for the four reference state equations). , , (Find the partial derivative)

[0053] Substitute the known parameters , , And set the initial values ​​for iteration: estimate the initial gain by scaling at the following two points. ;make ; .

[0054] The Gauss-Newton method is used for iteration, and the convergence condition is that the sum of squared residuals is less than 1 / 2. After four iterations, the system converges, yielding the system's nonlinear calibration parameter set: ; ; ; Substitute the above parameters into the first reference state equation and calculate... This is consistent with the measured value.

[0055] After obtaining the system's nonlinear calibration parameter set, extract the composite test voltage signal transmitted in step S1, and substitute it into the inverse mapping square root solution equation:

[0056] in This represents the value of the atmospheric absolute physical brightness temperature signal after removing second-order distortion. Substitute, within the root sign Since the number is positive, it can be solved normally, yielding the following result. If a mandatory order For example, brightness temperature is This stage stripped away The second-order nonlinear hardware intrinsic distortion error.

[0057] Since the system channels cover the characteristic frequency bands of water vapor and oxygen, the atmospheric absolute physical brightness temperature signal is parsed in parallel into a K-band physical brightness temperature signal containing 22.235 GHz to 30 GHz, and a V-band physical brightness temperature signal containing 51 GHz to 59 GHz.

[0058] The K-band physical brightness temperature signal and V-band physical brightness temperature signal output in step S2 are transmitted to step S3.

[0059] Step S3: Multi-band feature coupling decoupling; Step S3 executes multi-band feature coupling decoupling in the digital signal processor. The microwave energy level transition absorption of water vapor and the continuous spectrum scattering of liquid water droplets exhibit nonlinear aliasing in the frequency domain. Directly using coupling parameters for inversion will lead to unsteady divergence. Therefore, an external cross-medium physical reference signal is introduced as a priori spatial constraint, and orthogonal isolation is achieved in the frequency domain through differentiation.

[0060] Specifically, it includes the following three sub-steps: Sub-step S3.1: Reconstruction of the background atmospheric thermodynamic temperature profile; A digital signal processor extracts the V-band physical brightness temperature signal (51 GHz to 59 GHz) from the atmospheric absolute physical brightness temperature signal. Oxygen molecules exhibit a strong spin absorption band in the V-band, and the mixing ratio of oxygen in the atmosphere is constant. Therefore, the spatial distribution of the V-band physical brightness temperature signal is mainly determined by the atmospheric thermodynamic temperature.

[0061] The digital signal processor simultaneously retrieves the ground temperature signal and ground air pressure signal from the ground real-time environmental reference signal in step S1, in order to... Using the surface value at a given location as the initial boundary condition, and combining it with the V-band physical brightness temperature signal matrix, a background atmospheric thermodynamic temperature profile unaffected by liquid water is reconstructed by recursively applying a statistical regression matrix along the height direction.

[0062] in The vertical space target's altitude is represented as Background atmospheric thermodynamic temperature profile (unit: K) at the location. This represents the ground temperature signal value (unit: K) extracted from the ground real-time environmental reference signal. Indicates the target altitude (unit: m). Indicates atmospheric altitude Location, subject to ground air pressure signal With V-band physical brightness temperature signal matrix The specific expression for the atmospheric ambient temperature lapse rate function (unit: K / m) under joint constraints is as follows:

[0063] In this expression, the coefficients 0.0001 and 0.00005 were obtained by performing multiple linear regression calibration on 1 million sets of radiosonde data from 120 benchmark radiosonde stations between 2015 and 2025 and the corresponding V-band brightness temperature data. The regression determination coefficients are... ; 0.0065 is the standard atmospheric temperature lapse rate (K / m).

[0064] Sub-step S3.2: Delineation of the upper limit cutoff boundary for liquid water accumulation; Microwaves have a long wavelength and strong penetrating power through thin clouds and ice crystals, making it difficult to determine cloud boundaries using the microwave channel itself. The digital signal processor retrieves the cloud base infrared radiation signal (wavelength 8 μm to 14 μm) output from the infrared thermography probe in step S1. Water vapor and liquid water droplets are opaque in this wavelength band, allowing for accurate sensing of the cloud base's physical height.

[0065] Based on the equivalent infrared brightness temperature of the cloud base infrared radiation signal and the background atmospheric thermodynamic temperature profile The digital signal processor searches for the isothermal points of both liquid water and overlays empirical thickness compensation to determine the upper limit cutoff boundary for liquid water accumulation.

[0066] in This represents the calculated absolute height of the upper limit cutoff boundary for liquid water accumulation (unit: m). This represents the background atmospheric thermodynamic temperature profile (unit: K). Indicates the equivalent infrared brightness temperature (unit: K) of the infrared radiation signal at the cloud base. The empirical thickness compensation parameter (unit: m) represents the suspension of liquid water droplets. It is used to correct the cloud base height measurement deviation of the infrared thermography probe, ensuring that the liquid water integral interval completely covers the actual cloud base. It is obtained by statistical calibration of local radiosonde historical data. In this embodiment, it is set in conjunction with radiosonde data from eastern my country. (Error range ±10 m).

[0067] Sub-step S3.3: Orthogonal stripping of absorption lines; A digital signal processor extracts the K-band physical brightness temperature signal (22.235 GHz to 30 GHz), which includes a water vapor absorption peak and a continuous spectrum. Based on radiative transfer theory, a two-component extinction equation with truncated boundary constraints is established:

[0068] in Indicates frequency The total microwave optical thickness in the K-band at that location is determined by the K-band physical brightness temperature signal. Compared with the background atmospheric thermodynamic temperature profile integral average temperature The transformation is obtained, and the transformation relationship is: ,and The calculation formula is:

[0069] The integration limit of 10,000 meters corresponds to the maximum reconstruction height of the background atmospheric thermodynamic temperature profile. This indicates that water vapor molecules are at a certain frequency. and height Narrowband mass absorption coefficient at a given location (unit: m² / g). The value represents the concentration of water vapor in the vertical spatial distribution (unit: g / m³). This represents the absolute height value (in meters) of the upper limit cutoff boundary for liquid water accumulation input in sub-step S3.2. This indicates the frequency of liquid water droplets. and height Broadband quality absorption coefficient at the location (unit: m² / g). This represents the concentration of liquid water in the vertical spatial distribution (unit: g / m³).

[0070] Water vapor absorption coefficient It exhibits a sharp Lorentz line at 22.235 GHz, while the absorption coefficient of liquid water... It increases monotonically and gradually with frequency (consistent with the Rayleigh approximation). Digital signal processor for K-band total optical thickness. By calculating the second-order partial derivative of the frequency, the liquid horizontal component approaches zero, thus separating the water vapor characteristic component:

[0071] in, This is the narrowband absorption characteristic signal of water vapor output after decoupling. This signal is a second derivative spectrum in the frequency domain, still defined on the original frequency grid of the K-band.

[0072] The water vapor characteristics are then transformed back to the optical thickness domain for energy stripping.

[0073] The specific process of the inverse feature transformation is as follows: For the frequency domain second derivative spectrum of length 1024 points Perform a 1024-point inverse fast Fourier transform to obtain the frequency domain water vapor optical thickness component; By mapping this component back to 34 equally spaced sampling frequency grids from 22.235 GHz to 30 GHz in the K-band using linear interpolation, the water vapor optical thickness spectrum was obtained. .

[0074] From the total optical thickness Subtract and using a step function in The above height is set to zero to obtain the broadband absorption characteristic signal of liquid water:

[0075] in This refers to the broadband absorption characteristic signal of liquid water at the decoupled output. The step function is specifically defined as follows:

[0076] When height Greater than the upper limit cutoff boundary for liquid water accumulation At that time, the function value is forced to zero to ensure that the liquid water feature is only valid below the cloud base.

[0077] After this processing, the orthogonality between the narrow-band absorption characteristic signal of water vapor and the broadband absorption characteristic signal of liquid water reaches over 99.7%, effectively eliminating the ill-posedness of the downstream inversion. Step S3 outputs the two mutually orthogonal characteristic signals and transmits them to step S4.

[0078] Step S4: Spatial distribution mapping and total integral calculation; Step S4 is executed by a digital signal processor, whose inputs include the water vapor narrowband absorption characteristic signal and the liquid water broadband absorption characteristic signal output from step S3, as well as the ground real-time environmental reference signals (ground temperature signal, ground humidity signal, and ground air pressure signal) from step S1.

[0079] The inversion model employs a backpropagation neural network based on a multilayer perceptron topology.

[0080] The network structure is as follows: The input layer has 38 nodes, of which 34 nodes correspond to the normalized water vapor or liquid water characteristic spectrum values ​​of 34 equally spaced sampling frequency points in the K-band range of 22.235 GHz to 30 GHz, and the remaining 4 nodes correspond to the ground temperature signal, ground humidity signal, ground air pressure signal and cloud base height reference parameters, respectively. The two hidden layers contain 128 and 64 nodes respectively; The output layer has 58 nodes, corresponding to unequal-spacing slices from the ground to an altitude of 10,000 meters. The vertical spatial resolution is 25 meters from 0 to 500 meters, 50 meters from 500 to 2000 meters, and 250 meters from 2000 to 10,000 meters.

[0081] Sub-step S4.1: Calculate the feature forward mapping; Since the absorption characteristics of water vapor and liquid water have been orthogonally decoupled in step S3, in order to avoid the unsteadiness caused by cross-contamination of features, two multilayer perceptron branches with the same structure but independent weights are run in parallel in the digital signal processor for water vapor and liquid water inversion, respectively.

[0082] Taking the water vapor inversion branch as an example, the digital signal processor merges the input signals into a water vapor-specific input feature vector. ,in These are the narrowband absorption characteristic signal values ​​of water vapor at 34 frequency points. This is the ground humidity signal value. This serves as the baseline parameter for cloud base height. The data undergoes non-linear activation propagation through the first hidden layer:

[0083] in Indicates the first hidden layer. The output value of each neuron This represents the Sigmoid non-linear activation function. Indicates the input layer's first... The node is connected to the first hidden layer. The connection weight matrix of each node is pre-trained and solidified using historical radar echo sets from radiosonde stations. The training data comes from radiosonde data (including water vapor and liquid water concentration profiles) and corresponding microwave radiation brightness temperature data from three benchmark radiosonde stations (Beijing, Shanghai, and Guangzhou) over the past five years. The time synchronization error between the radiosonde data profiles and the corresponding microwave radiation brightness temperature data is ≤5 min. The training batch is 10,000 batches with a learning rate of 0.001. Training stops when the loss function is less than 0.0001. The first feature vector representing the water vapor-specific input feature vector. Each dimension component Indicates the first hidden layer. Bias term parameters for each neuron.

[0084] The information continues to propagate forward to the second hidden layer and the output layer.

[0085] The output layer uses a linear rectified function as the final nonlinear activation function to hard-cut off negative output values.

[0086] in Indicates the output layer number The inversion results of the physical stratification nodes, i.e., the water vapor vertical spatial distribution concentration signal at the th Discrete node values ​​of the layer (unit: g / m³). Represents the linear rectified activation function. Indicates the second hidden layer The node to the output layer The connection weight matrix of each node Indicates the second hidden layer The output value of each neuron denotes the bias parameter of the physical layer node in the output layer.

[0087] The liquid water inversion branch adopts the same equation for parallel forward mapping, and the feature vector is replaced with , and finally outputs the concentration signal of the vertical spatial distribution of liquid water at 58 layered height nodes .

[0088] Sub-step S4.2: Discrete cross-section depth integration; The digital signal processor retrieves the concentration signal of the vertical spatial distribution of water vapor, performs discrete numerical integration according to the vertical height slicing interval, and obtains the total water vapor holding capacity of the entire atmosphere layer, that is, the columnar total water vapor integral parameter signal:

[0089] where denotes the generated columnar total water vapor integral parameter signal (unit: mm), denotes the discrete node value of the concentration signal of the vertical spatial distribution of water vapor at the layer (unit: g / m³), denotes the physical thickness of the vertical space between the layer and the layer (unit: m).

[0090] Similarly, the digital signal processor performs the same integration operation on the concentration signal of the vertical spatial distribution of liquid water, and outputs the columnar total liquid water integral parameter signal:

[0091] where denotes the generated columnar total liquid water integral parameter signal (unit: mm), denotes the discrete node value of the concentration signal of the vertical spatial distribution of liquid water at the layer (unit: g / m³).

[0092] So far, the concentration signal of the vertical spatial distribution of water vapor, the concentration signal of the vertical spatial distribution of liquid water, the columnar total water vapor integral parameter signal, and the columnar total liquid water integral parameter signal are all generated, and the data closed loop is completed.

[0093] In actual use, the microwave radiation testing hardware system matched with this embodiment adopts an integrated isolation cabin design, which is specifically as follows: Energy harvesting employs a shared offset parabolic reflector antenna (1.2 m aperture, 0.6 m focal length), configured with a dual-corrugated horn feed confocal design. The K-band feed frequency of the dual-corrugated horn feed is 22.235 GHz–30 GHz with a gain of 25 dB; the V-band feed frequency is 51 GHz–59 GHz with a gain of 30 dB. The curvature of the metallic reflector layer of the offset parabolic reflector is optimized by optical ray tracing, achieving a reflection efficiency ≥95%.

[0094] The dual-corrugated horn feed converts the captured spatial electromagnetic waves into waveguide master modes, which are then fed into the back-end microwave RF front-end module.

[0095] The microwave RF front-end module, the built-in high-temperature blackbody source, and the solid-state noise source are integrated into a high-precision constant-temperature sealed chamber. The built-in high-temperature blackbody source uses a carbon-doped silicone pyramid array as the radiation substrate, and the surface of the pyramid is covered with an electric heating film and a high-precision platinum resistance temperature sensor.

[0096] The detection and control unit operates a variable proportional integral-derivative closed-loop control to stabilize the physical temperature of the built-in high-temperature blackbody source at 353.15 K, with a fluctuation range not exceeding ±0.1 K, so as to eliminate the influence of high-temperature reference drift on the calibration equation.

[0097] The solid-state noise source generates broadband white noise using a reverse-biased avalanche diode. The equivalent noise temperature is controlled to 200.0 K by a precision coaxial attenuator network and injected non-contactly into the main RF link via a directional coupler. The external liquid nitrogen cryogenic source is a microwave-transparent foam inverted cone-shaped absorbing target filled with liquid nitrogen. The radiation temperature is clamped at 77.15 K using the boiling phase transition of liquid nitrogen, serving as a cold junction reference.

[0098] The multi-channel microwave receiver adopts a superheterodyne frequency conversion architecture.

[0099] The radio frequency (RF) energy output from the RF microwave switch array is first amplified by a low-noise amplifier to reduce its noise floor, and then down-converted to a local oscillator signal generated by a double-balanced mixer and a voltage-controlled oscillator, shifting the 22 GHz to 59 GHz microwave signal to an intermediate frequency (IF) of 0 to 2 GHz. The IF signal is then divided into channels by an IF amplifier network and a bandpass filter bank, and sent to a square-law detector diode array to convert the IF power envelope into a DC voltage analog quantity.

[0100] The analog-to-digital converter uses a 24-bit high-resolution sampling chip to discretize the DC voltage at a 100 kHz oversampling rate. It then performs digital low-pass filtering and decimation noise reduction through a field-programmable gate array (FPGA), and finally outputs a high signal-to-noise ratio composite test voltage signal, a first calibration reference voltage signal, a second calibration reference voltage signal, a third calibration reference voltage signal, and a fourth calibration reference voltage signal to the digital signal processor, thus completing the physical tracing from microwave photons to digital voltage.

[0101] Example 2: This example is basically the same as Example 1, except that Example 1 requires periodic replenishment of liquid nitrogen, which is inconvenient for unattended remote stations. This example uses active semiconductor cooling instead of liquid nitrogen cooling without changing the core logic.

[0102] Specifically as follows: Liquid nitrogen-free nonlinear reference reconstruction (corresponding to the evolution of steps S1 and S2): The pure cryogenic calibration radiation signal source is replaced by an external liquid nitrogen cryogenic source with an active semiconductor-cooled target. The active semiconductor-cooled target utilizes the Peltier effect (using a TEC1-12706 Peltier element, cascaded in two stages). By controlling the drive current of the cascaded semiconductors (adjustable range 0-5A), the target surface temperature is pumped to 233.15 K. The precision temperature measuring bridge inside the target surface, model PT1000, has a measurement accuracy of ±0.1K and provides real-time feedback of the target surface temperature, forming a closed-loop control.

[0103] In step S1, the generation sources of the second and fourth calibration reference voltage signals are changed. In step S2, the second reference state algebraic equation in the second-order microwave transfer equations is changed to:

[0104] in This represents the value of the second calibration reference voltage signal captured under the architecture of Example 2. This indicates the absolute physical temperature value output in real time by the active semiconductor cooling target surface, which is obtained through real-time feedback from a precision temperature-sensing bridge embedded inside the target surface. This represents the second-order nonlinear factor parameter in the state of Example 2. This represents the first-order linear gain parameter in the state of Example 2. This represents the voltage DC bias constant parameter under the conditions of Example 2. The system equivalent noise temperature parameter represents the state of Example 2.

[0105] Because the temperature of the cooled target surface is higher than that of liquid nitrogen, the temperature range of the calibration reference point is reduced, and the curvature detection sensitivity is decreased.

[0106] To compensate for rounding errors in solving the equations, the digital signal processor performs rounding operations when solving the system's nonlinear calibration parameter set. , and Kalman filtering is introduced to apply smooth recursive constraints. The state equations for each parameter are as follows: ; in This represents process noise, characterizing the time-series fluctuation uncertainty of the parameters themselves; all observation equations are... ,in To observe noise, the random error of a single calibration measurement is characterized. The process noise variance is set to The observation noise variance is set to ; The process noise variance is set to The observation noise variance is set to ; The process noise variance is set to The observation noise variance is set to The covariance matrix of historical calibration data was calculated using data from nearly 100 calibrations. This observer was used to recursively smooth the parameters of a single calculation to restore the accuracy of the parameter calculation and ensure that the atmospheric absolute physical brightness temperature signal output to step S3 has no baseline drift.

[0107] In some preferred embodiments, a preferred alternative is provided: in step S4 of embodiment 2, the multilayer perceptron network is replaced with a radial basis function (RBF) neural network to avoid the gradient vanishing problem.

[0108] Sub-step S4.1 forward mapping is modified as follows: The digital signal processor combines the narrowband absorption characteristic signal of water vapor and the broadband absorption characteristic signal of liquid water, which are orthogonally output in step S3, with the ground-based environmental reference signal from step S1 to construct a feature vector matrix. The non-linear distance metric between the input layer and the hidden layer uses a Gaussian kernel function. The hidden layer's... The output activation value of each node is:

[0109] in Indicates the hidden layer of the RBF network. Each node targets the feature vector matrix. The response output quantity, This represents the total number of Gaussian kernel functions in the hidden layer, as shown in this embodiment. ; Indicates the first The center cluster vectors of the Gaussian function are used to extract and solidify the training data using the K-Means unsupervised clustering algorithm (with 128 clusters). Indicates the first The width smoothing parameter of the Gaussian function is set to 0.8 times the average distance between cluster centers. (Calculation...) The value range is 0.5–1.2.

[0110] The output layer uses linear weighted summation: ; in Indicates the output layer number The substance concentration values ​​obtained by inverting from each physical stratification node. This represents the total number of Gaussian kernel functions in the hidden layers. Indicates the hidden layer number 1 The node and the output layer The mapping weight matrix between physical hierarchical nodes.

[0111] In practical use, only the source hardware acquisition method and the end mapping equation were replaced. The overall process still follows the causal chain of discretization transformation, reconstruction of transfer equation system, orthogonal decoupling of boundary constraints, and nonlinear mapping in three-dimensional space, avoiding redundant and isolated data.

[0112] Example 3: This example discloses a technical solution for delineating liquid water boundaries using a micro-pulse lidar instead of an infrared temperature probe. In practical applications: Based on Example 2, this example replaces the source of the cloud base height characterization signal from an infrared thermometer to a micropulse lidar (MPL). The MPL emits a 532 nm pulsed laser towards the zenith and receives backscattered signals from clouds and aerosols.

[0113] The cloud base height characterization signal collected in step S1 is concretized into a laser backscatter echo profile signal.

[0114] Steps S3 and S3.2 no longer rely on temperature fitting to find isothermal points, but instead perform spatial gradient analysis on the laser backscattered echo profile signal to obtain the truncation boundary.

[0115] The specific process is as follows: Extract all that satisfy... And local gradient The height point that is the maximum value is taken as the point with the minimum height among the maximum values, and then the upper limit cutoff boundary for liquid water accumulation is selected.

[0116] in This represents the absolute height of the upper limit cutoff boundary for liquid water accumulation calculated using the laser medium, which directly and equivalently replaces the upper limit cutoff boundary for liquid water accumulation in Example 1. ; This indicates the high-resolution image extracted from the laser backscattered echo profile signal. The backscattering coefficient at the location, in units of ; The threshold constant representing the scattering coefficient that distinguishes clear-sky background from solid cloud droplets is set based on statistical observation data from a micropulse lidar (model MPL-4B). ,when At that time, it was determined that a physical cloud droplet existed.

[0117] Calculated Directly used as the upper limit of liquid water integration in step S3.3 With the downstream processing unchanged, the data stream seamlessly merges into the original microwave orthogonal link after crossing the laser medium, achieving a complete replacement of the detection method.

[0118] In practical implementation, the invention will be further illustrated through a numerical simulation of a severe convective weather event: At the leading edge of a strong convective squall line system, the atmospheric boundary layer is extremely unstable, causing a rapid accumulation of water vapor in the lower atmosphere and a surge in liquid water content within the clouds. Traditional linear single-channel radiometers, due to the deep overlap between water vapor characteristic peaks and cloud droplet scattering bands, will inevitably cause their inversion algorithms to collapse and diverge at this moment. The forced orthogonal decoupling scheme of this invention is introduced at this critical juncture.

[0119] Hardware benchmark parameters: Built-in high-temperature blackbody source closed-loop locking at Equivalent Additional Temperature of Solid-State Noise Source The external liquid nitrogen cryogenic source is in .

[0120] Environmental situational awareness: ground temperature signal Ground humidity signal Ground pressure signal The infrared thermometer measures the equivalent infrared brightness temperature of the infrared radiation signal at the cloud base. .

[0121] Signal Acquisition and Discretization (Verification Step S1): At the microwave frequency of 23.834 GHz, a multi-channel microwave receiver sequentially drives an RF microwave switch array, and the analog-to-digital converter outputs five digitized analog quantities: First calibration reference voltage signal ; Second calibration reference voltage signal ; Third calibration reference voltage signal ; Fourth calibration reference voltage signal ; Composite test voltage signal .

[0122] The five sets of voltage signals and four sets of meteorological boundary signals output in step S1 are all used downstream.

[0123] Nonlinear curvature deconstruction and brightness-temperature reconstruction: A digital signal processor retrieves five sets of voltage signals to inverse the equations. High-temperature voltage increment. Low-temperature voltage increment If the detector diode is absolutely linear, the voltage increments at high and low temperatures should be equal when equal amounts of 200.0 K solid-state noise are injected. This confirms that the detector experiences compression saturation in a high-temperature region.

[0124] Calculate the nonlinear correction ratio function .

[0125] The nonlinear calibration parameter set of the system is obtained by iteratively solving the Jacobian matrix. , , .

[0126] Will , , and Substituting the values ​​into the inverse mapping square root solution equation, we obtain the atmospheric absolute physical brightness temperature signal value at 23.834 GHz. If a mandatory order Pseudo-brightness temperature is This stage involved the stripping away of [something]. The second-order nonlinear hardware intrinsic distortion error is eliminated. Step S2 consumes all reference voltage signals and outputs a pure atmospheric absolute physical brightness temperature signal matrix to step S3.

[0127] Forced orthogonality of features under strong convection boundaries (verification step S3): The digital signal processor calls the V-band physical brightness temperature signal matrix and combines it with surface parameters. and The background atmospheric thermodynamic temperature profile, which decreases along the height, is reconstructed. .

[0128] because Searching upwards along the background atmospheric thermodynamic temperature profile, the isothermal physical altitude is located at 1850 meters, which is locked as the upper limit cutoff boundary for liquid water accumulation. .

[0129] The K-band physical brightness temperature signal matrix was extracted, and second-order partial derivatives with respect to frequency were calculated for the continuous optical thickness curve from 22.235 GHz to 30 GHz. The smooth cloud droplet microwave scattering broadband substrate was zeroed out under the second-order partial derivative operator, and the Lorentz-shaped peaks of water vapor microwave energy level transitions were isolated. The prominent peak energy characteristics were inversely transformed to generate a narrowband absorption characteristic signal of water vapor. This narrowband absorption characteristic signal was subtracted from the total optical thickness, and the energy distribution of liquid water above 1850 meters was forcibly zeroed out to generate a broadband absorption characteristic signal of liquid water. The orthogonality of the input features reached over 99.7%, effectively eliminating the divergence risk in downstream inversion.

[0130] Three-dimensional meteorological field reconstruction and total discrete integral (verification step S4): Two parallel multilayer perceptron networks receive orthogonal eigenvector matrices.

[0131] The water vapor profile reconstruction branch processes the narrowband absorption characteristic signal of water vapor, outputting 58 discrete node values, including the concentration at a height of 200 meters. .

[0132] The liquid water profile reconstruction branch processes the broadband absorption characteristic signal of liquid water. Constrained by the upper limit truncation boundary of liquid water accumulation at 1850 meters, the output nodes exhibit high concentration of liquid water distribution below 1850 meters, while the node values ​​above 1850 meters are forced to zero by the ReLU activation function, thus avoiding the generation of spurious negative values ​​or abnormally high values ​​that penetrate the cloud top.

[0133] By integrating the discrete cross-section depth, the concentration signal of water vapor in the vertical spatial distribution is accumulated layer by layer to output a columnar total water vapor integral parameter signal. (Calculation example:) Corresponding layer thickness The contribution of this layer is (The total integral value is obtained by summing all 58 layers). Similarly, in the liquid water integral calculation, layers below 1850 meters are summed to output a columnar total liquid water integral parameter signal. .

[0134] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0135] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-channel microwave radiation method for testing atmospheric water vapor and liquid water content, characterized in that, include: Step S1: Control the multi-channel microwave receiver to perform time-division multiplexing to collect atmospheric natural microwave radiation signals, pure high-temperature calibration radiation signals, pure low-temperature calibration radiation signals, constant additional noise radiation signals, ground real-time environmental reference signals, and cloud base height characterization signals. The cloud base height characterization signals are cloud base infrared radiation signals or laser backscattered echo profile signals. The collected radio frequency signals are down-converted, detected, and analog-to-digital converted to obtain composite test voltage signals, first calibration reference voltage signals, second calibration reference voltage signals, third calibration reference voltage signals, and fourth calibration reference voltage signals. Step S2: Using the first calibration reference voltage signal, the second calibration reference voltage signal, the third calibration reference voltage signal, and the fourth calibration reference voltage signal, calculate the system equivalent noise temperature parameter, gain coefficient, and nonlinear factor. Based on the system equivalent noise temperature parameter, gain coefficient, and nonlinear factor, perform inverse mapping on the composite test voltage signal to obtain the atmospheric absolute physical brightness temperature signal after removing second-order distortion, and divide it into K-band physical brightness temperature signal and V-band physical brightness temperature signal. Step S3: Establish a background atmospheric thermodynamic temperature profile using the V-band physical brightness temperature signal and the ground-based environmental reference signal; delineate the upper limit cutoff boundary for liquid water accumulation using the cloud base height characterization signal; and decouple the K-band physical brightness temperature signal by frequency domain differentiation using the background atmospheric thermodynamic temperature profile and the upper limit cutoff boundary for liquid water accumulation as constraints to obtain a narrowband water vapor absorption characteristic signal and a broadband liquid water absorption characteristic signal. Step S3 involves establishing the background atmospheric thermodynamic temperature profile by using the ground temperature and ground pressure from the ground real-time environmental reference signal as boundary conditions, and combining the V-band physical brightness temperature signal to obtain the background atmospheric thermodynamic temperature profile through statistical regression. When the cloud base height characterization signal is a cloud base infrared radiation signal, the upper limit cutoff boundary for liquid water accumulation includes: Find the height in the background atmospheric thermodynamic temperature profile that is isothermal to the equivalent infrared brightness temperature of the cloud base infrared radiation signal, and superimpose empirical thickness compensation. When the cloud base height characterization signal is a laser backscattered echo profile signal, the upper limit cutoff boundary for liquid water accumulation is defined by: performing spatial gradient analysis on the laser backscattered echo profile signal and taking the lowest height point among the height points that satisfy the condition that the backscattering coefficient exceeds the threshold and the local gradient is at its maximum value. The frequency domain decoupling in step S3 is as follows: Perform the second-order frequency partial derivative on the optical thickness spectrum obtained by converting the K-band physical brightness temperature signal to separate the narrowband absorption characteristic signal of water vapor. Then, subtract the water vapor component from the optical thickness spectrum and spatially zero it using the upper limit truncation boundary of liquid water accumulation to obtain the broadband absorption characteristic signal of liquid water. Step S4: Combine the narrowband absorption characteristic signal of water vapor and the broadband absorption characteristic signal of liquid water with the ground real-time environmental reference signal, input them into a pre-trained neural network model, and invert to obtain the vertical spatial distribution concentration signal of water vapor and the vertical spatial distribution concentration signal of liquid water. Perform vertical integration on the vertical spatial distribution concentration signal of water vapor and the vertical spatial distribution concentration signal of liquid water to obtain the columnar total water vapor integral parameter signal and the columnar total liquid water integral parameter signal.

2. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 1, characterized in that, The physical process of generating the first calibration reference voltage signal in step S1 is as follows: the cascaded gain of the receiving system performs frequency integration on the input energy density within the working bandwidth. The input energy density is equivalent to the voltage response corresponding to the sum of the absolute physical temperature value of the built-in high-temperature blackbody source and the equivalent noise temperature parameter of the system.

3. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 2, characterized in that, The third calibration reference voltage signal in step S1 is generated by superimposing the pure high-temperature calibration radiation signal and the constant additional noise radiation signal at the RF link port and then passing it through the receiving link; the fourth calibration reference voltage signal is generated by superimposing the pure low-temperature calibration radiation signal and the constant additional noise radiation signal at the RF link port and then passing it through the receiving link.

4. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 1, characterized in that, The process of calculating the system's equivalent noise temperature parameters, gain coefficient, and nonlinear factor in step S2 includes: A second-order microwave transfer equation system containing the equivalent noise temperature parameters, gain coefficient, and nonlinear factor of the system is constructed. Algebraic difference operations are performed using the first, second, third, and fourth calibration reference voltage signals, and the equivalent noise temperature parameters, gain coefficient, and nonlinear factor of the system are obtained by iterative solution.

5. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 4, characterized in that, The iterative solution employs the Gauss-Newton method and constructs the Jacobian matrix. The convergence condition for the iteration is that the sum of squared residuals is less than 1 / 2. .

6. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 1, characterized in that, Step S3, establishing the background atmospheric thermodynamic temperature profile, includes: using the ground temperature and ground pressure in the ground real-time environmental reference signal as boundary conditions, and combining the V-band physical brightness temperature signal to obtain the background atmospheric thermodynamic temperature profile through statistical regression recursion.

7. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 6, characterized in that, When the cloud base height characterization signal is a cloud base infrared radiation signal, the upper limit cutoff boundary for liquid water accumulation includes: In the background atmospheric thermodynamic temperature profile, find the height at which the equivalent infrared brightness temperature of the cloud base infrared radiation signal is isothermal, and then superimpose empirical thickness compensation.

8. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 1, characterized in that, The neural network model mentioned in step S4 is a multilayer perceptron network or a radial basis function neural network. The output layer of the neural network model contains 58 nodes corresponding to unequal hierarchical layers from the ground to an altitude of 10,000 meters.

9. The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to claim 1, characterized in that, The vertical integration in step S4 is a discrete cross-sectional depth integration based on the unequal vertical layer thickness from the ground to an altitude of 10,000 meters.

10. A multi-channel microwave radiation atmospheric water vapor and liquid water content testing system, characterized in that, The method for testing atmospheric water vapor and liquid water content using multi-channel microwave radiation according to any one of claims 1-9 is characterized by comprising: Antenna, used to receive natural atmospheric microwave radiation signals; Built-in high-temperature blackbody source to provide the upper boundary for thermodynamic calibration; A pure cryogenic reference source is used to provide a lower boundary for thermodynamic calibration. The pure cryogenic reference source is an external liquid nitrogen cryogenic source or an active semiconductor cooling target. Solid-state noise source used to generate a constant additional noise radiation signal; A multi-channel microwave receiver is used to perform microwave frequency down-conversion and analog-to-digital discretization conversion; Meteorological condition sensor array, used to collect ground-based environmental reference signals; A cloud base detection unit is used to acquire cloud base height characterization signals. The cloud base detection unit is an infrared temperature measurement probe or a micro-pulse lidar.

Images