Multi-level intelligent overheating protection system and method based on gallium nitride module antenna

Abstract

The invention relates to the technical field of antennas, in particular to a multi-level intelligent overheating protection system and method based on a gallium nitride module antenna. According to the method, a thermal load optimal migration path is analyzed based on a virtual thermal capacity network model, and an input impedance matching network of a target unit is synchronously tuned according to the thermal load optimal migration path to compensate beam phase distortion caused by thermal migration; and according to the boundary constraint condition of the real-time threat scene feature injection virtual thermal capacity network model, dynamically solving a game equilibrium solution of a thermal load migration path and a beam parameter tuning amount, and generating a thermal-radio frequency combined control instruction set. According to the method, full-process closed-loop control from thermal inertia characteristic analysis, thermal network modeling to thermal-radio frequency collaborative optimization is realized, so that the gallium nitride antenna system can intelligently plan an optimal thermal migration path and compensate thermally induced performance distortion in real time in a complex thermal environment and a threat scene, and finally dual optimization of heat dissipation efficiency and radio frequency indexes is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of antennas, and in particular to a multi-level intelligent overheat protection system and method based on a gallium nitride module antenna. Background Art

[0002] With the development of communication systems towards high frequency bands and high power, gallium nitride (GaN) materials have been gradually widely used in high-power and high-frequency radio frequency electronic devices, especially in fields such as radar, communication systems, and power amplifiers. However, the heat accumulation effect generated under high radio frequency power loading easily causes overheating of antenna elements, which in turn leads to beam distortion, efficiency reduction, and even hardware damage. Traditional overheat protection methods mostly use temperature thresholds to trigger power reduction or forced heat dissipation, but it is difficult to balance the real-time nature of thermal management and the stability of radio frequency performance. In the prior art, although phase change energy storage materials can relieve temperature rise by absorbing heat, the dynamic coupling mechanism between their phase change rates and radio frequency power has not been quantitatively modeled; and the thermal migration path planning based on a fixed thermal resistance network lacks the ability to synergistically compensate for beam phase distortion. In addition, electromagnetic interference and thermal radiation will further exacerbate the game conflict between thermal and radio frequency parameters, and existing systems are difficult to dynamically balance the contradiction between thermal load migration and beam pointing accuracy. Therefore, there is an urgent need for a multi-level intelligent protection method that integrates dynamic thermal inertia characteristic analysis, virtual heat capacity network modeling, and thermal-radio frequency joint tuning to achieve adaptive thermal optimization and radio frequency performance guarantee of GaN antennas under complex working conditions. Summary of the Invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a multi-level intelligent overheat protection system and method based on a gallium nitride module antenna.

[0004] The technical solution adopted by the present invention to achieve the above object is as follows: The first aspect of the present invention discloses a multi-level intelligent overheat protection method based on a gallium nitride module antenna, including the following steps: Real-time analyze the coupling relationship between the phase change rate of the phase change energy storage material in the gallium nitride antenna element and the radio frequency power loading time sequence, and generate a dynamic thermal inertia characteristic map; Construct a virtual heat capacity network model based on the dynamic thermal inertia characteristic map, and the virtual heat capacity network model includes the thermal conduction weight coefficients between each unit and the radio frequency power carrying margin parameters; Analyze the optimal thermal load migration path based on the virtual heat capacity network model, and synchronously tune the input impedance matching network of the target unit according to the optimal thermal load migration path to compensate for the beam phase distortion caused by thermal migration; According to the boundary constraint conditions of the real-time threat scenario characteristics injected into the virtual heat capacity network model, dynamically solve the game equilibrium solution of the thermal load migration path and the beam parameter tuning amount, and generate a thermal-radio frequency joint control instruction set.

[0005] Preferably, the coupling relationship between the phase change rate of the phase change energy storage material in the gallium nitride antenna unit and the radio frequency power loading time sequence is analyzed in real time to generate a dynamic thermal inertia characteristic map, specifically as follows: The phase change rate time sequence vectors of multiple preset monitoring points inside the gallium nitride antenna unit and the corresponding radio frequency power loading time sequence pulse parameters are synchronously collected in real time. After aligning the two according to the time stamp, a spatio-temporal correlation data set is formed; Based on a sliding time window, the spatio-temporal correlation data set is segmented. The mutual information analysis method is used to calculate the correlation degree sequences between the phase change rate change gradients and the radio frequency power pulse amplitude and duty cycle within each time window, and the dynamic lag time coefficient between the two is extracted through the grey correlation degree algorithm; The correlation degree sequences and the lag time coefficient are input into a preset three-dimensional feature fusion model, which are respectively mapped into a thermal inertia time delay factor, a phase change thermal response intensity factor, and a radio frequency power cumulative effect index in the time-frequency domain. Combined with a preset phase change rate change rate threshold, piecewise linear interpolation is performed to generate a dynamic thermal inertia characteristic matrix containing the temperature field evolution trend; According to the spatio-temporal distribution of each factor in the dynamic thermal inertia characteristic matrix, through a three-dimensional interpolation algorithm and an isosurface rendering technique, the temperature field evolution trend, the thermal response intensity, and the power cumulative effect are visualized to generate a dynamic thermal inertia characteristic map containing color gradients and contour lines.

[0006] Preferably, a virtual heat capacity network model is constructed based on the dynamic thermal inertia characteristic map, specifically as follows: The thermal inertia time delay factor, the phase change thermal response intensity factor, and the radio frequency power cumulative effect index are extracted from the dynamic thermal inertia characteristic map as the model input parameters; Based on the correlation degree sequences and the lag time coefficient between each antenna unit within a sliding time window, a heat conduction weight coefficient is constructed using a graph theory adjacency matrix, where the correlation degree sequence is used as the edge weight after normalization, and the lag time coefficient is used to correct the weight distribution ratio of adjacent units after exponential smoothing; The radio frequency power cumulative effect index and the phase change thermal response intensity factor are non-linearly superimposed, and combined with the gradient direction of the temperature field evolution trend, the radio frequency power bearing margin parameter of each unit is determined; Based on the real-time updated heat conduction weight coefficient and the radio frequency power bearing margin parameter, the node connection structure of the virtual heat capacity network is optimized using a greedy algorithm, and the heat flow balance constraint condition is synchronously applied to ensure that the heat load migration path and the radio frequency power distribution satisfy the first law of thermodynamics; The convergence of the network model is verified through Monte Carlo random perturbation testing, and the adjustment step size of the weight coefficient and the update threshold of the margin parameter are reversely corrected according to the statistical distribution characteristics of the perturbation response data to achieve the dynamic matching of the model parameters and the physical field changes.

[0007] Preferably, based on the virtual heat capacity network model, analyze the optimal migration path of the thermal load, and synchronously tune the input impedance matching network of the target unit according to the optimal migration path of the thermal load to compensate for the beam phase distortion caused by thermal migration. Specifically: Based on the heat conduction weight coefficient and the radio frequency power bearing margin parameter, use the Dijkstra algorithm to determine the minimum thermal resistance path between each target unit, and generate a set of candidate thermal load migration paths; Combined with the gradient direction of the temperature field evolution trend, by calculating the product of the temperature gradient change rate between adjacent nodes on each candidate migration path and the phase change heat response intensity factor and then performing weighted averaging, obtain the path thermal flow dynamic equilibrium factor; Compare the path thermal flow dynamic equilibrium factor with a preset thermal flow equilibrium threshold, eliminate the candidate migration paths with factor values greater than the preset thermal flow equilibrium threshold, and retain the candidate migration paths with factor values not greater than the preset thermal flow equilibrium threshold to obtain a set of optimal migration paths for the thermal load; According to the thermal flow distribution characteristics on the optimal migration path, extract the change in the thermally induced dielectric constant of the preset nodes, and determine the corresponding beam phase offset compensation value based on the radio frequency power cumulative effect index; Use the change in the thermally induced dielectric constant and the phase offset compensation value to dynamically adjust the input impedance matching network of the target unit to compensate for the beam phase distortion caused by thermal migration.

[0008] Preferably, use the change in the thermally induced dielectric constant and the phase offset compensation value to dynamically adjust the input impedance matching network of the target unit to compensate for the beam phase distortion caused by thermal migration. Specifically: According to the change in the thermally induced dielectric constant of each node on the optimal migration path, calculate its relative deviation rate from the nominal dielectric constant of the antenna unit substrate to generate a dielectric constant perturbation coefficient matrix; Decompose the phase offset compensation value into phase gradient components in the azimuth and elevation dimensions according to the preset beam pointing angle, and calculate the equivalent susceptance adjustment amount required for the matching network in combination with the dielectric constant perturbation coefficient matrix; According to the equivalent susceptance adjustment amount, dynamically adjust the capacitance value distribution of the adjustable capacitor bank so that the reflection phase at the input end of the matching network cancels out the thermally induced phase distortion component; Verify the phase compensation effect by real-time monitoring of the antenna port standing wave ratio. If the residual phase error exceeds the preset error threshold, use the gradient descent method to iteratively optimize the mapping relationship between the dielectric constant perturbation coefficient and the susceptance adjustment amount; Update the power bearing margin parameter in the virtual heat capacity network according to the optimized mapping relationship to form a closed-loop adaptive matching of thermal-electric-phase parameters.

[0009] Preferably, according to the characteristics of the real-time threat scenario, inject the boundary constraint conditions of the virtual heat capacity network model, dynamically solve the game equilibrium solution of the heat load migration path and the beam parameter tuning amount, and generate a thermal-radio frequency joint control instruction set, specifically as follows: Obtain the electromagnetic interference intensity distribution and the heat radiation value of the threat scenario in real time, and quantify them into the heat flux density constraint vector and the radio frequency power limit matrix, which are used as the dynamic boundary conditions of the virtual heat capacity network model; Based on the boundary conditions, perform constraint processing on the heat conduction weight coefficient, generate a constrained thermal resistance topology diagram, and determine the heat flux accommodation space of each node in combination with the radio frequency power carrying margin parameter; Introduce the heat flux density constraint vector on the thermal resistance topology diagram for two-way search, eliminate the path branches that exceed the heat flux accommodation space, and obtain a set of feasible heat load migration paths that meet the boundary conditions; According to the radio frequency power limit matrix of each node on the set of feasible migration paths, determine the maximum allowable beam scanning angle and the power tuning range corresponding to the path, and generate a path-beam coupling parameter table; By solving the multi-objective optimization function of the minimum value of the path thermal resistance and the maximum value of the beam pointing stability, screen out the Pareto optimal solution set in the path-beam coupling parameter table; Linearly weighted and fuse the thermal resistance gradient change rate and the beam phase tuning amount in the Pareto optimal solution set to generate a thermal-radio frequency joint control instruction set including power reallocation instructions, heat flux migration timing, and phase compensation parameters.

[0010] The multi-level intelligent overheat protection method further includes the following steps: Collect the antenna element temperature field distribution data and the beam pointing accuracy deviation value after the execution of the instruction set in real time, extract the offset between the actual heat flux migration path and the theoretical path, and generate a heat conduction error vector; Perform correlation analysis on the heat conduction error vector and the radio frequency power loading timing, obtain the dynamic correction factor of the heat conduction weight coefficient, and adjust the weight distribution ratio of adjacent nodes in combination with the lag time coefficient; According to the beam pointing accuracy deviation value, inversely deduce the actual influence of the heat-induced dielectric constant perturbation amount, and update the attenuation gradient of the radio frequency power carrying margin parameter in combination with the phase change heat response intensity factor; Use the corrected heat conduction weight coefficient and power carrying margin parameter to reconstruct the adjacency matrix of the virtual heat capacity network, and verify the network convergence through the heat flux balance constraint; Based on the residual distribution of the predicted value and the measured value of the node temperature of the reconstructed network, adaptively adjust the update step size of the weight coefficient and the margin parameter.

[0011] The second aspect of the present invention discloses a multi-level intelligent overheat protection system based on a gallium nitride module antenna, which is applied to any of the steps of the multi-level intelligent overheat protection method based on a gallium nitride module antenna, including: The GaN antenna array module uses a GaN-on-SiC broadband antenna array, which contains at least 8 radiating units. Each unit has a built-in temperature sensor and micro phase change energy storage material. The operating frequency band covers 1.2-5.8GHz, and the unit spacing is 0.75 times the center frequency wavelength. The thermal state dynamic monitoring module is integrated with the temperature sensor of the gallium nitride antenna unit, the phase change rate detection circuit of the phase change energy storage material, and the thermal inertia characteristic calculation unit. It outputs a thermal state vector set including the chip junction temperature gradient, substrate heat flux density, and the solid-liquid phase change ratio of the phase change material in real time. A virtual heat capacity network modeling module is used to dynamically calculate the heat conduction weight coefficient and RF power carrying margin parameter between adjacent antenna units according to the thermal state vector set, and generate a virtual heat capacity network topology diagram including the heat migration path priority and beam distortion compensation amount; The RF-thermal coupling compensation module includes an adjustable capacitor array and a distributed phase shifter. Based on the thermal migration path data of the virtual thermal capacitance network topology diagram, the reflection coefficient of the input impedance matching network is dynamically tuned to achieve beam pointing phase pre-distortion compensation within ±0.25dB. The threat scenario adaptive decision module dynamically calculates the Pareto optimal solution of the thermal load migration path and beam synthesis parameters according to the modulation type and pulse width characteristics of the real-time threat signal, and outputs an interference effectiveness enhancement instruction set with a frequency band priority of 1.2-1.6GHz.

[0012] The present invention solves the technical defects existing in the background technology, and has the following beneficial effects: the present invention establishes a dynamic regulation mechanism of thermal-electric coupling to realize closed-loop control of the entire process from thermal inertia characteristic analysis, thermal network modeling to thermal-RF collaborative optimization, so that the antenna system can not only intelligently plan the optimal heat migration path in a complex thermal environment, but also compensate for thermally induced performance distortion in real time, ultimately achieving dual optimization of heat dissipation efficiency and RF indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0014] Figure 1 This is the overall method flow chart of the multi-level intelligent overheat protection method; Figure 2 This is a partial method flowchart of the multi-level intelligent overheat protection method; Figure 3 This is a system block diagram of the multi-level intelligent overheat protection system. Detailed implementation manners

[0015] In order to more clearly understand the above objects, features, and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0016] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0017] As Figure 1 shown, a multi-level intelligent overheat protection method based on a gallium nitride module antenna is disclosed in the first aspect of the present invention, including the following steps: S102. Real-time analyze the coupling relationship between the phase change rate of the phase change energy storage material in the gallium nitride antenna unit and the radio frequency power loading time sequence, and generate a dynamic thermal inertia characteristic map; S104. Construct a virtual heat capacity network model based on the dynamic thermal inertia characteristic map, and the virtual heat capacity network model includes the heat conduction weight coefficients and radio frequency power bearing margin parameters between each unit; S106. Analyze the optimal heat load migration path based on the virtual heat capacity network model, and synchronously tune the input impedance matching network of the target unit according to the optimal heat load migration path to compensate for the beam phase distortion caused by heat migration; S108. Inject the boundary constraint conditions of the virtual heat capacity network model according to the real-time threat scenario characteristics, dynamically solve the game equilibrium solution of the heat load migration path and the beam parameter tuning amount, and generate a thermal-radio frequency joint control instruction set.

[0018] It should be noted that since the traditional GaN antenna system is difficult to simultaneously consider effective heat dissipation and stable radio frequency performance under high temperature and high power conditions, the thermal management efficiency is low and the beam quality deteriorates. The present invention realizes a full-process closed-loop control from thermal inertia characteristic analysis, thermal network modeling to thermal-radio frequency collaborative optimization by establishing a dynamic regulation mechanism of thermal-electric coupling, enabling the antenna system to not only intelligently plan the optimal heat migration path in a complex thermal environment, but also compensate for the heat-induced performance distortion in real time, and finally achieve the dual optimization of heat dissipation efficiency and radio frequency indicators.

[0019] Preferably, the coupling relationship between the phase change rate of the phase change energy storage material in the gallium nitride antenna unit and the RF power loading time sequence is analyzed in real time to generate a dynamic thermal inertia characteristic map, such as Figure 2 shown, specifically as follows: S202. Synchronously collect the phase change rate time sequence vectors and the corresponding RF power loading time sequence pulse parameters at multiple preset monitoring points inside the gallium nitride antenna unit in real time, and form a spatio-temporal correlation data set after aligning the two according to the time stamp; Among them, the phase change rate time sequence vector refers to a dynamic data sequence composed of the change rates of the solid-liquid phase change ratios of the phase change material (such as paraffin) sampled and recorded at different time points; the RF power loading time sequence pulse parameters refer to the set of characteristic parameters such as the power amplitude, pulse width, and duty cycle of the RF signal in the time dimension.

[0020] S204. Segment the spatio-temporal correlation data set based on a sliding time window, calculate the correlation degree sequences between the phase change rate change gradients and the RF power pulse amplitude and duty cycle in each time window using the mutual information analysis method, and extract the dynamic lag time coefficient between the two through the grey correlation degree algorithm; S206. Input the correlation degree sequence and the lag time coefficient into a preset three-dimensional feature fusion model, map them to a thermal inertia time delay factor, a phase change thermal response intensity factor, and a RF power cumulative effect index respectively in the time-frequency domain, and perform piecewise linear interpolation in combination with a preset phase change rate change rate threshold to generate a dynamic thermal inertia characteristic matrix including the temperature field evolution trend; Among them, the preset three-dimensional feature fusion model is an intelligent analysis model based on multi-dimensional data fusion obtained through pre-training, which is used to perform collaborative mapping and feature extraction on the thermal inertia time delay factor, the phase change thermal response intensity factor, and the RF power cumulative effect index in the time domain, frequency domain, and spatial domain. This model uses wavelet transform to decompose the time-frequency features, combines the long short-term memory (LSTM) network to capture the dynamic lag effect, and enhances the features of the spatial thermal field distribution through the convolutional neural network (CNN), and finally outputs a dynamic thermal inertia characteristic matrix including the temperature field gradient, the thermal response intensity, and the power cumulative effect.

[0021] S208. According to the spatio-temporal distribution of each factor in the dynamic thermal inertia characteristic matrix, visualize the temperature field evolution trend, the thermal response intensity, and the power cumulative effect through a three-dimensional interpolation algorithm and an isosurface rendering technique, and generate a dynamic thermal inertia characteristic map including color gradients and contour lines.

[0022] Taking a GaN antenna array with a working frequency band of 2.4 GHz and a unit peak power of 150 W as an example, in specific implementation: Paraffin-based phase change materials (melting point 60 °C) and distributed temperature sensors are respectively embedded in 4 radiation units, and the time series data of the liquid phase ratio of the phase change materials (such as the phase change rate of unit A is 0.8% / s at t = 1.2 s) and the radio frequency power pulse parameters (pulse width 10 ms, duty cycle 30%) are synchronously collected at a sampling rate of 100 Hz; The aligned spatio-temporal data set is segmented by a 5 ms sliding time window, and the mutual information analysis method is used to obtain the correlation degree sequence between the phase change gradient and the power amplitude (such as the maximum mutual information value in the window is 0.75), and the lag time coefficient (such as 2.3 ms) is extracted by combining the grey correlation degree algorithm; The above parameters are input into a three-dimensional feature fusion model (including a wavelet transform layer and an LSTM network), and the thermal inertia time delay factor (such as level 1.8), the phase change heat response intensity factor (such as level 2.4), and the power cumulative effect index (such as level 3.1) are output. After piecewise linear interpolation, a dynamic thermal inertia feature matrix (resolution 0.1 °C / pixel) is generated; Finally, a three-dimensional thermal field map is generated through the OpenGL rendering engine, and the temperature gradient between units (such as the highest temperature on the surface of unit A is 62.5 °C) and the heat flow diffusion direction (such as the decreasing rate along the positive Y-axis is 1.2 °C / mm) are intuitively displayed with a red-blue gradient color scale.

[0023] In summary, the present invention accurately characterizes the evolution trend of the temperature field by real-time analyzing the spatio-temporal correlation between the phase change rate and the radio frequency power, and fusing the thermal inertia, thermal response and power cumulative effect, and generates a visual dynamic thermal inertia feature map, so as to provide high-precision data support for subsequent intelligent thermal regulation.

[0024] Preferably, a virtual heat capacity network model is constructed based on the dynamic thermal inertia feature map, specifically: The thermal inertia time delay factor, the phase change heat response intensity factor and the radio frequency power cumulative effect index are extracted from the dynamic thermal inertia feature map as the model input parameters; Based on the correlation degree sequence and the lag time coefficient between each antenna unit within the sliding time window, a heat conduction weight coefficient is constructed by using a graph theory adjacency matrix, where the correlation degree sequence is used as the edge weight after normalization, and the lag time coefficient is used to correct the weight distribution ratio of adjacent units after exponential smoothing. It should be noted that when constructing the heat conduction weight coefficient, first, the heat correlation degree values between each antenna element (such as the mutual information value between 0 and 1) are normalized so that all values are unified within the range of 0 - 1, serving as the initial weight of the edges connecting each element in the adjacency matrix. Then, the measured lag time coefficient (such as a 3.5 ms heat transfer delay between two elements) is subjected to exponential smoothing to obtain a more stable correction coefficient. Finally, this correction coefficient is used to adjust the initial weight distribution ratio between adjacent elements. For example, the original weight from A to B is 0.8, and after the lag time correction, it may be adjusted to 0.7, which can more accurately reflect the actual heat conduction situation. Through this method, a weight network model that better conforms to the actual heat transfer characteristics can be established.

[0025] Nonlinearly superimpose the radio frequency power cumulative effect index and the phase change heat response intensity factor, and combine the gradient direction of the temperature field evolution trend to determine the radio frequency power bearing margin parameter of each unit; It should be noted that when calculating the radio frequency power bearing margin parameter of each antenna element, the index representing the power accumulation degree (such as level 3.1) and the intensity factor reflecting the heat absorption capacity of the phase change material (such as level 2.4) are superimposed and calculated through a nonlinear function (such as an S-shaped curve) to obtain a comprehensive heat load index. Then, in combination with the diffusion direction of the real-time temperature field (such as a gradient change of 1.2 °C / mm from the high-temperature area to the low-temperature area), the comprehensive index is directionally corrected: when the heat flow direction is consistent with the power distribution, the margin is appropriately increased, and vice versa. Finally, the percentage of the radio frequency power that each unit can safely withstand under the current working conditions is output (such as the bearing margin of unit A is 85%), providing a basis for subsequent power distribution.

[0026] Based on the heat conduction weight coefficient and the radio frequency power bearing margin parameter updated in real time, use the greedy algorithm to optimize the node connection structure of the virtual heat capacity network, and simultaneously apply the heat flow balance constraint condition to ensure that the heat load migration path and the radio frequency power distribution satisfy the first law of thermodynamics; Verify the convergence of the network model through Monte Carlo random perturbation testing, and inversely correct the adjustment step of the weight coefficient and the update threshold of the margin parameter according to the statistical distribution characteristics of the perturbation response data to achieve the dynamic matching of the model parameters and the physical field changes.

[0027] Among them, the Monte Carlo random perturbation test refers to a method of verifying the model stability by inputting a series of randomly simulated interference factors (such as temperature fluctuations, power mutations, etc.) into the system; the perturbation response data refers to the recorded data set of the actual reactions (such as temperature change curves, thermal resistance fluctuations, etc.) of the system to these random interferences during the test.

[0028] Based on the dynamic thermal inertia characteristic map of the aforementioned GaN antenna array (operating frequency band 2.4 GHz) (including the thermal inertia time delay factor of 1.8 levels, the phase change heat response intensity factor of 2.4 levels, and the power accumulation effect index of 3.1 levels for unit A), in specific implementation: extract the correlation degree sequence (such as the mutual information value of 0.62) and the lag time coefficient (such as 3.5 ms) of adjacent units A - B with a 200 - ms sliding time window, construct a 5×5 graph - theory adjacency matrix, where the normalized correlation degree (0.62 / maximum value 0.75 ≈ 0.83) is used as the edge weight, and the lag coefficient adjusted by exponential smoothing (α = 0.3) is used to adjust the A - B weight distribution to 0.72; non - linearly superimpose the power accumulation effect index (3.1 levels) and the heat response intensity factor (2.4 levels) of unit A through the Sigmoid function (output value 0.68), combine with the temperature field gradient direction (1.2 °C / mm in the positive Y - axis direction), and calculate its radio - frequency power - bearing margin parameter; use the greedy algorithm to optimize the network structure, reduce the thermal resistance from unit A to unit C by 12%, and at the same time ensure that the error between the total input power (50 W) and the heat dissipation power is < 3% through the heat - flow balance constraint; finally, verify the model convergence through Monte Carlo perturbation testing (such as 1000 samplings). For example, when the temperature prediction residual exceeds 0.5 °C, automatically reduce the weight coefficient adjustment step size from 0.1 to 0.05 to achieve dynamic matching with the measured thermal field.

[0029] In summary, since the traditional thermal management model is difficult to dynamically characterize the coupling relationship between the heat conduction characteristics and the power - bearing capacity among the units in the gallium nitride antenna array, it leads to a lack of precise guidance for heat load distribution and radio - frequency performance optimization. The present invention constructs a virtual heat capacity network model, dynamically quantifies the heat conduction weights and the power - bearing margins among the units, and uses an optimization algorithm to ensure heat - flow balance, thereby realizing the intelligent planning and dynamic adjustment of the heat load migration path, effectively improving the thermal distribution uniformity and radio - frequency working stability of the antenna array.

[0030] Preferably, analyze the optimal heat load migration path based on the virtual heat capacity network model, and synchronously tune the input impedance matching network of the target unit according to the optimal heat load migration path to compensate for the beam phase distortion caused by heat migration. Specifically: Based on the heat conduction weight coefficient and the radio - frequency power - bearing margin parameter, use the Dijkstra algorithm to determine the minimum thermal resistance path among the target units, and generate a set of candidate heat load migration paths; It should be noted that when planning the thermal load migration path, each unit in the antenna array is regarded as a network node, and the thermal conduction weight coefficient (such as 0.72) between units is converted into the path thermal resistance value. Then, using Dijkstra's shortest path algorithm, starting from the overheated unit (such as unit A at 62.5 °C), the total thermal resistance values of all possible paths to other units (especially units with good heat dissipation conditions) are gradually calculated. Then, by comparing the cumulative thermal resistance values of different paths, the path with the minimum thermal resistance from the heat source unit to the target heat dissipation unit is found (such as the total thermal resistance of the path from A to B to D is 1.8 K / W). Finally, several (such as 3) candidate paths with relatively small thermal resistance are output for subsequent evaluation of heat flow balance and selection of the optimal path.

[0031] Combined with the gradient direction of the temperature field evolution trend, by calculating the weighted average after multiplying the temperature gradient change rate between adjacent nodes on each candidate migration path by the phase change heat response intensity factor, the path heat flow dynamic balance factor is obtained; Compare the path heat flow dynamic balance factor with the preset heat flow balance threshold, eliminate the candidate migration paths with factor values greater than the preset heat flow balance threshold, and retain the candidate migration paths with factor values not greater than the preset heat flow balance threshold to obtain the optimal thermal load migration path set; According to the heat flow distribution characteristics on the optimal migration path, extract the change in the thermally induced dielectric constant of the preset nodes, and determine the corresponding beam phase offset compensation value based on the radio frequency power cumulative effect index; Among them, the heat flow distribution characteristics refer to the dynamic change characteristics such as the temperature gradient and heat flux density of the heat on the transmission path; the preset nodes are the key unit positions in the antenna array selected in advance for monitoring and control; the change in the thermally induced dielectric constant represents the offset value of the dielectric characteristics (such as the dielectric constant) of the antenna material caused by temperature changes.

[0032] Use the change in the thermally induced dielectric constant and the phase offset compensation value to dynamically adjust the input impedance matching network of the target unit to compensate for the beam phase distortion caused by thermal migration.

[0033] Among them, the target unit refers to a specific antenna unit in the gallium nitride antenna array that currently needs thermal load migration and impedance matching tuning (such as unit A and unit B that need to compensate for phase offset mentioned above).

[0034] Similarly, taking the aforementioned GaN antenna array (the temperature of unit A is 62.5 °C and the load margin is 85%) as an example, in specific implementation: Based on the virtual heat capacity network model (the edge weight of the adjacency matrix is 0.72), the Dijkstra algorithm is used to calculate the minimum thermal resistance path from unit A to the heat dissipation unit D (such as the total thermal resistance of the path from A to B to D is 1.8 K / W), and 3 candidate paths are generated; Combining the temperature field gradient direction (1.2 °C / mm along the Y-axis), calculate the thermal flow dynamic equilibrium factor of each path (such as the factor value of the path from A to B to D is 0.65, and the preset threshold is 0.7), and screen out the optimal path A to B to D; Extract the change in the thermally induced dielectric constant (Δε r = 0.3) of unit B on this path, and calculate the phase shift compensation value (such as 5.2°) according to its power accumulation effect index (level 2.8); Adjust the capacitance value of the matching network of unit B from 3.5 pF to 3.2 pF through the adjustable capacitor array to achieve beam pointing deviation compensation (residual error < 0.5°), while maintaining the power output of unit A stable at 48 W (the load margin consumption is 12%).

[0035] In summary, the present invention intelligently plans the optimal heat migration path through the virtual heat capacity network model, and synchronously adjusts the impedance matching network, realizing the coordinated compensation of heat load migration and beam phase distortion, which not only ensures the efficient evacuation of heat, but also maintains the stability of the radio frequency performance of the antenna array, effectively solving the problem of thermoelectric coupling interference.

[0036] Preferably, the input impedance matching network of the target unit is dynamically adjusted by using the change in the thermally induced dielectric constant and the phase shift compensation value to compensate for the beam phase distortion caused by heat migration, specifically: According to the change in the thermally induced dielectric constant of each node on the optimal migration path, calculate its relative deviation rate from the nominal dielectric constant of the antenna element substrate, and generate a dielectric constant perturbation coefficient matrix; Decompose the phase shift compensation value into phase gradient components in the azimuth and elevation dimensions according to the preset beam pointing angle, and calculate the required equivalent susceptance adjustment amount of the matching network in combination with the dielectric constant perturbation coefficient matrix; It should be noted that when calculating the equivalent susceptance adjustment amount of the matching network, the change in the thermally induced dielectric constant is converted into a corresponding perturbation coefficient matrix (such as the Δε r = 0.3 of unit B corresponds to a perturbation coefficient of 0.85). Then, the phase shift value to be compensated (such as 5.2°) is decomposed into phase components in the azimuth and elevation directions (such as azimuth 3.1° + elevation 2.1°). Next, these phase components are matched with the corresponding unit data in the perturbation coefficient matrix to obtain the susceptance value that needs to be adjusted for each unit's matching network (such as unit B needs to increase the susceptance by 0.15 S). Finally, a list of the required equivalent susceptance adjustment amounts for each unit is output to guide the parameter setting of the subsequent adjustable capacitor bank.

[0037] Dynamically adjust the capacitance value distribution of the tunable capacitor bank according to the equivalent susceptance adjustment amount, so that the reflection phase at the input end of the matching network cancels out the thermally induced phase distortion component; Verify the phase compensation effect by real-time monitoring of the antenna port standing wave ratio. If the residual phase error exceeds the preset error threshold, the mapping relationship between the dielectric constant perturbation coefficient and the susceptance adjustment amount is iteratively optimized by the gradient descent method; Update the power carrying margin parameter in the virtual heat capacity network according to the optimized mapping relationship to form a closed-loop adaptive matching of thermal-electric-phase parameters.

[0038] For the thermally induced dielectric constant change Δε of element B (temperature 58 °C) in the aforementioned GaN antenna array r = 0.3 (nominal ε r = 9.8), the relative deviation rate of 3.06% is determined and a perturbation coefficient matrix (coefficient of element B is 0.97) is generated; the 5.2° phase shift to be compensated is decomposed into 3.1° in the azimuth dimension and 2.1° in the elevation dimension. Combining with the perturbation coefficient matrix, the electromagnetic field simulation inversion calculation is carried out, and it is obtained that the equivalent susceptance of the matching network of element B needs to be increased by 0.15 S; accordingly, the capacitance value of the tunable capacitor bank of element B is adjusted from 3.5 pF to 3.8 pF (step accuracy 0.1 pF), so that the reflection phase at the input end cancels out the thermal distortion phase in the opposite direction, and the measured standing wave ratio drops from 1.8 to 1.2; when the residual phase error of 0.6° exceeds the threshold of 0.5°, the mapping relationship is optimized by iterating 3 times using the gradient descent method. Finally, the power carrying margin parameter of element B is updated from 82% to 79%, and at the same time, the thermal conduction weight coefficient from adjacent element A to B in the virtual heat capacity network is corrected from 0.72 to 0.68 to form a closed-loop regulation.

[0039] In summary, the present invention realizes the adaptive collaborative regulation of thermal deformation and electrical performance by dynamically calculating the correspondence between dielectric constant perturbation and phase shift, adjusting the impedance matching network parameters in real time to actively cancel the thermally induced phase distortion by the reflection phase, and continuously correcting the compensation accuracy through closed-loop optimization, effectively ensuring the beam pointing stability of the antenna array under the change of thermal load.

[0040] Preferably, according to the characteristics of the real-time threat scenario, inject the boundary constraint conditions into the virtual heat capacity network model, dynamically solve the game equilibrium solution of the thermal load migration path and the beam parameter tuning amount, and generate a thermal-radio frequency joint control instruction set, specifically: Obtain the electromagnetic interference intensity distribution and thermal radiation value of the threat scenario in real time, and quantify them into a heat flux density constraint vector and a radio frequency power limit matrix as the dynamic boundary conditions of the virtual heat capacity network model; Based on the boundary conditions, perform constraint processing on the thermal conduction weight coefficient, generate a constrained thermal resistance topology diagram, and determine the heat flux accommodation space of each node in combination with the radio frequency power carrying margin parameter; It should be noted that when generating the constrained thermal resistance topology map, the original thermal conduction weight coefficients are adjusted according to the constraint conditions of the threat scenario (such as the power upper limit of unit C being 30 W and the heat flux upper limit of unit D being 50 W). Specifically, when a certain unit is restricted by power or heat flux, the thermal conduction weight between it and adjacent units will be reduced (such as the weight from unit C to D being adjusted from 0.65 to 0.58), which is equivalent to "setting up roadblocks" on these restricted paths. Then, the equivalent thermal resistance values between units are recalculated using the adjusted weight coefficients, and a new thermal resistance topology map is drawn. The constrained high thermal resistance paths will be clearly marked on the thermal resistance topology map (such as the path from unit C to D being shown in red), ensuring that these restricted areas are automatically avoided during subsequent path planning, just like a navigation system avoiding congested roads.

[0041] It should be noted that when determining the heat flux accommodation space of each node, the current radio frequency power bearing margin parameter of each antenna unit is obtained (such as the bearing margin of unit D being 75%). Then, based on the maximum theoretical heat dissipation capacity of the unit (such as 80 W) and its current working state, the additional heat flux upper limit that the unit can withstand under the current margin is determined (such as 80 W × 75% = 60 W). Next, in combination with the thermal conduction weight coefficients between adjacent units, this heat flux upper limit is distributed to each possible heat transfer path (such as 42 W being distributed to the path from unit C to D). Finally, by real-time monitoring the actual heat flux values of each path, this distribution scheme is dynamically adjusted to ensure that the calculated heat flux accommodation space is not exceeded at any time, thereby avoiding local overheating.

[0042] A two-way search is performed by introducing a heat flux density constraint vector on the thermal resistance topology map to eliminate the path branches that exceed the heat flux accommodation space, obtaining a set of feasible migration paths for the heat load that meet the boundary conditions; According to the radio frequency power limit matrix of each node on the set of feasible migration paths, the maximum allowable beam scanning angle and power tuning range corresponding to the path are determined, and a path-beam coupling parameter table is generated; By solving the multi-objective optimization function of the minimum path thermal resistance and the maximum beam pointing stability, the Pareto optimal solution set is selected from the path-beam coupling parameter table; It should be noted that when screening the optimal solution set, a coupling parameter table is established that includes all feasible paths (such as from A to C to D), their corresponding thermal resistance values (such as 2.1 K / W), and beam stability parameters (such as a pointing error of 0.8°). Then, two optimization goals are set: to minimize the path thermal resistance as much as possible (high heat transfer efficiency), and at the same time to minimize the beam pointing error as much as possible (stable RF performance). Then, these path schemes are analyzed to find a compromise solution (i.e., the Pareto optimal solution) that cannot be comprehensively surpassed by other solutions - for example, a solution with a non-minimal thermal resistance but the most stable beam, or the smallest thermal resistance but a slightly worse beam. Finally, a set of these optimal candidate solutions is output for the system to select the most suitable balance solution according to real-time requirements (such as preferentially selecting a solution with a 15% reduction in thermal resistance and a beam error ≤ 0.8°).

[0043] Linearly weighted fusion is performed on the thermal resistance gradient change rate and the beam phase tuning amount in the Pareto optimal solution set to generate a thermal-RF joint control instruction set that includes power reallocation instructions, thermal flow migration timing, and phase compensation parameters.

[0044] Among them, the threat scenarios include complex combat environments under the combined action of enemy electromagnetic interference (such as directional RF suppression) and environmental thermal radiation (such as solar radiation or high-temperature heat sources).

[0045] For example, in a threat scenario of encountering enemy electromagnetic interference in the 1.5 GHz frequency band (field strength 20 V / m) and solar radiation (heat flux density 800 W / m²), first, the characteristics of the interference signal are quantified into an RF power limit matrix (such as setting the power upper limit of unit C in the 1.4 - 1.6 GHz frequency band to 30 W), and at the same time, the thermal radiation value is converted into a heat flux density constraint vector (such as setting the heat flux upper limit on the surface of unit D to 50 W). Based on this constraint condition, the thermal conduction weight coefficient from unit C to D in the virtual heat capacity network is corrected from 0.65 to 0.58, a constrained thermal resistance topology map is generated, and the current heat flux accommodation space of unit D is determined to be 42 W. After excluding 3 over-limit paths through two-way path search, the feasible path from A to C to D (thermal resistance 2.1 K / W) is retained, and the maximum allowable beam scanning angle of this path is determined to be ±25° according to the power limit matrix. After multi-objective optimization, the Pareto optimal solution is selected: a 15% reduction in thermal resistance and a beam pointing error ≤ 0.8°. Finally, an instruction set is generated - reducing the power of unit C from 35 W to 28 W, starting the cooling fan of unit D with a delay of 2 ms, and compensating a 3.2° phase shift for the matching network of unit C.

[0046] In summary, since it is difficult for the GaN antenna array to simultaneously meet the thermal management requirements and optimize the RF performance in a complex threat environment, the thermal load migration and beam parameter tuning restrict each other. By quantifying the threat scenario features as network constraint conditions, the present invention intelligently balances the multi-objective requirements of minimizing thermal resistance and beam stability, dynamically generates cooperative control instructions, enables the antenna system to ensure efficient heat dissipation and maintain the optimal RF operating state under electromagnetic interference and thermal radiation threats, and realizes the comprehensive optimization of thermal and electrical performance.

[0047] The multi-level intelligent overheat protection method further includes the following steps: Real-time collect the temperature field distribution data of the antenna element and the beam pointing accuracy deviation value after the execution of the instruction set, extract the offset between the actual migration path and the theoretical path of the heat flow, and generate a heat conduction error vector; Perform a correlation analysis on the heat conduction error vector and the RF power loading timing sequence, obtain the dynamic correction factor of the heat conduction weight coefficient, and adjust the weight distribution ratio of adjacent nodes in combination with the lag time coefficient; According to the beam pointing accuracy deviation value, inversely deduce the actual influence of the heat-induced dielectric constant perturbation amount, and update the attenuation gradient of the RF power bearing margin parameter in combination with the phase change heat response intensity factor; Use the corrected heat conduction weight coefficient and power bearing margin parameter to reconstruct the adjacency matrix of the virtual heat capacity network, and verify the network convergence through the heat flow balance constraint; Based on the residual distribution of the predicted value and the measured value of the node temperature of the reconstructed network, adaptively adjust the update step sizes of the weight coefficient and the margin parameter.

[0048] Taking the antenna array after executing the thermal-RF joint control instruction set as an example (the measured temperature of unit D is 61.2 °C / the theoretical value is 60.5 °C, and the beam pointing deviation is 0.9 °), in specific implementation: collect the temperature field data through an infrared thermal imager, and find that the actual path of the heat flow from A to C to D deviates from the theoretical path by 15% (generate an error vector [0, 0.15, 0]); analyze its correlation coefficient with the RF power pulse (duty cycle 40%) of unit C reaches 0.78, and calculate the dynamic correction factor 0.92 accordingly, and correct the heat conduction weight coefficient from unit C to D from 0.58 to 0.53; inversely deduce the actual change amount Δε of the dielectric constant of unit D according to the 0.9 ° beam deviation r = 0.35 (the original prediction is 0.3), combined with its phase change heat response intensity factor of 2.1 level, adjust the power bearing margin attenuation gradient from 12% / min to 15% / min; reconstruct the adjacency matrix with the corrected parameters, and after verifying that the heat flow balance error < 5%, adjust the weight coefficient update step size from 0.05 to 0.03 according to the temperature residual of 0.7 °C of unit D to improve the prediction accuracy of the next cycle.

[0049] It should be noted that, to solve the problem that the traditional thermal management system cannot correct the error of the heat conduction model in real time under dynamic working conditions, resulting in a deviation between the theoretical thermal management strategy and the actual execution effect. The present invention dynamically corrects the heat conduction weight and power margin parameters by real-time feedback of the temperature field and beam performance data, enables the virtual heat capacity network to continuously approximate the actual physical system, forms a closed-loop thermal-electric collaborative regulation mechanism with self-learning ability, and effectively improves the real-time accuracy and environmental adaptability of the thermal management strategy.

[0050] In this embodiment, the multi-level intelligent overheat protection method further includes the following steps: Real-time collect the difference between the measured value and the theoretical value of the temperature gradient on the heat load migration path, generate a heat flow path offset vector, synchronously extract the residual error of beam phase compensation, and form a phase residual sequence; fuse the two into a bimodal error feature matrix; Perform time-frequency decomposition on the bimodal error feature matrix, extract the heat flow offset spectrum feature and the time-domain correlation of the phase residual, and output the heat conduction weight correction factor and the dielectric constant perturbation sensitivity coefficient; Based on the heat conduction weight correction factor, perform sliding window weighted adjustment on the original heat conduction weight coefficient in the virtual heat capacity network, and use the dielectric constant perturbation sensitivity coefficient to adjust the mapping relationship between the phase residual sequence and the dielectric constant perturbation, and generate the updated heat conduction weight coefficient and the dielectric constant perturbation mapping relationship; Substitute the updated heat conduction weight coefficient and the dielectric constant perturbation mapping relationship into the preset heat flow balance model for solution, and obtain a new heat flow path offset vector and a phase residual sequence; According to the newly generated heat flow path offset vector and phase residual sequence, dynamically adjust the sliding window size of the heat conduction weight correction factor and the scaling ratio of the dielectric constant perturbation sensitivity coefficient to form a closed-loop optimization.

[0051] Among them, the preset heat flow balance model refers to a mathematical model used to describe the dynamic balance relationship between heat flow input, conduction and dissipation in the antenna system, ensuring that the heat accumulation and dissipation of each node during the heat load migration process always satisfy the law of conservation of energy.

[0052] It should be noted that, to solve the problem that the traditional thermal management solution is difficult to correct the heat flow path offset and the residual error of phase compensation in real time under dynamic working conditions, resulting in the mismatch of thermal-electric parameters. This embodiment realizes the closed-loop collaborative optimization of the heat flow migration path and the beam compensation parameters by fusing the bimodal feature analysis of the temperature gradient and the phase error, and enables the system to always maintain the optimal heat distribution and radio frequency performance under dynamic heat load changes.

[0053] In this embodiment, the multi-level intelligent overheat protection method further includes the following steps: Based on the currently updated dielectric constant disturbance mapping relationship, the phase residual sequence is decomposed according to the frequency component to generate the initial capacitance adjustment value of the adjustable capacitor group of each antenna unit; According to the heat conduction weight coefficient of each node in the virtual heat capacity network, the initial capacitance adjustment amount is weightedly corrected; After the capacitance adjustment is performed, the antenna port scattering parameters are collected in real time, the characteristic deviation of the mapping relationship between the antenna port scattering parameters and the dielectric constant perturbation is extracted, and the scattering parameter residual vector is generated. The covariance matrix of Mahalanobis distance is constructed using the heat conduction weight coefficient, and the statistical distance between the scattering parameter residual vector and the phase residual sequence is calculated to quantify the compensation effect of thermoelectric performance distortion. According to the distribution characteristics of Mahalanobis distance and the spatial correlation of heat conduction weight coefficient, the self-consistency verification index including path consistency factor and dielectric compensation efficiency is output; When the self-consistency index exceeds the limit, the sensitivity coefficient of the dielectric constant perturbation mapping relationship is corrected in the direction of the Mahalanobis distance gradient, and the sliding window size of the heat conduction weight coefficient is adjusted synchronously.

[0054] It should be noted that in order to solve the problem that it is difficult for traditional GaN antenna systems to accurately coordinate capacitor tuning and thermal conduction characteristics in a dynamic thermal environment, resulting in a mismatch between dielectric compensation efficiency and thermal management requirements, this embodiment dynamically optimizes the synergistic relationship between dielectric compensation and thermal conduction by establishing a self-consistent verification mechanism for thermal-electric parameters, so that capacitor tuning can effectively offset thermally induced phase distortion while maintaining spatial consistency with the heat flow path, thereby achieving dual optimization of thermal management and electrical performance compensation.

[0055] like Figure 3 As shown, the second aspect of the present invention discloses a multi-level intelligent overheating protection system based on a gallium nitride module antenna, which is applied to any of the steps of the multi-level intelligent overheating protection method based on a gallium nitride module antenna, including: The GaN antenna array module 1011 uses a GaN-on-SiC broadband antenna array, which includes at least 8 radiating units. Each unit has a built-in temperature sensor and a micro phase-change energy storage material. The operating frequency band covers 1.2-5.8 GHz, and the unit spacing is 0.75 times the center frequency wavelength. The thermal state dynamic monitoring module 1022, which is integrated with the temperature sensor of the gallium nitride antenna unit, the phase change rate detection circuit of the phase change energy storage material and the thermal inertia characteristic solving unit, outputs in real time a thermal state vector set including the chip junction temperature gradient, the substrate heat flux density and the solid-liquid phase change ratio of the phase change material; A virtual heat capacity network modeling module 1033 is used to dynamically calculate the heat conduction weight coefficient and the radio frequency power carrying margin parameter between adjacent antenna units according to the thermal state vector set, and generate a virtual heat capacity network topology diagram including the heat migration path priority and the beam distortion compensation amount; The radio frequency-thermal coupling compensation module 1044 includes an adjustable capacitor array and a distributed phase shifter. Based on the thermal migration path data of the virtual heat capacity network topology diagram, it dynamically tunes the reflection coefficient of the input impedance matching network to achieve beam pointing phase pre-distortion compensation within ±0.25 dB. The threat scenario adaptive decision-making module 1055 dynamically calculates the Pareto optimal solution of the thermal load migration path and beam synthesis parameters according to the modulation type and pulse width characteristics of the real-time threat signal, and outputs an interference effectiveness enhancement instruction set with a frequency band priority of 1.2 - 1.6 GHz.

[0056] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A multi-level intelligent overheat protection method based on a gallium nitride module antenna, characterized in that It includes the following steps: Real-time analyze the coupling relationship between the phase change rate of the phase change energy storage material in the gallium nitride antenna unit and the radio frequency power loading timing, and generate a dynamic thermal inertia characteristic map; Construct a virtual heat capacity network model based on the dynamic thermal inertia characteristic map. The virtual heat capacity network model includes the heat conduction weight coefficient between each unit and the radio frequency power bearing margin parameter; Analyze the optimal heat load migration path based on the virtual heat capacity network model, and synchronously tune the input impedance matching network of the target unit according to the optimal heat load migration path to compensate for the beam phase distortion caused by heat migration; Inject the boundary constraint conditions of the virtual heat capacity network model according to the real-time threat scenario characteristics, dynamically solve the game equilibrium solution of the heat load migration path and the beam parameter tuning amount, and generate a thermal-radio frequency joint control instruction set.

2. The multi-level intelligent overheat protection method based on a gallium nitride module antenna according to claim 1, wherein Real-time analyze the coupling relationship between the phase change rate of the phase change energy storage material in the gallium nitride antenna unit and the radio frequency power loading timing, and generate a dynamic thermal inertia characteristic map, specifically: Real-time synchronously collect the phase change rate time series vector and the corresponding radio frequency power loading timing pulse parameters at multiple preset monitoring points inside the gallium nitride antenna unit, and form a spatio-temporal correlation data set after aligning the two according to the time stamp; Segment the spatio-temporal correlation data set based on a sliding time window, use the mutual information analysis method to calculate the correlation degree sequence between the phase change rate change gradient and the radio frequency power pulse amplitude and duty cycle within each time window, and extract the dynamic lag time coefficient between the two through the grey correlation degree algorithm; Input the correlation degree sequence and the lag time coefficient into a preset three-dimensional feature fusion model, map them to the thermal inertia time delay factor, the phase change heat response intensity factor, and the radio frequency power cumulative effect index respectively in the time-frequency domain, and perform piecewise linear interpolation in combination with a preset phase change rate change rate threshold to generate a dynamic thermal inertia characteristic matrix including the temperature field evolution trend; According to the spatio-temporal distribution of each factor in the dynamic thermal inertia characteristic matrix, visualize the temperature field evolution trend, the heat response intensity, and the power cumulative effect through a three-dimensional interpolation algorithm and an isosurface rendering technique, and generate a dynamic thermal inertia characteristic map including color gradients and contour lines.

3. The multi-level intelligent overheat protection method based on a gallium nitride module antenna according to claim 1, wherein Construct a virtual heat capacity network model based on the dynamic thermal inertia characteristic map, specifically: Extract the thermal inertia time delay factor, the phase change heat response intensity factor, and the radio frequency power cumulative effect index from the dynamic thermal inertia characteristic map as the model input parameters; Based on the correlation degree sequence and the lag time coefficient between each antenna unit within a sliding time window, construct the heat conduction weight coefficient using a graph theory adjacency matrix, where the correlation degree sequence is used as the edge weight after normalization, and the lag time coefficient is used to correct the weight distribution ratio of adjacent units after exponential smoothing; Nonlinearly superimpose the radio frequency power cumulative effect index and the phase change heat response intensity factor, and determine the radio frequency power bearing margin parameter of each unit in combination with the gradient direction of the temperature field evolution trend; Based on the real-time updated heat conduction weight coefficient and the radio frequency power bearing margin parameter, use the greedy algorithm to optimize the node connection structure of the virtual heat capacity network, and synchronously apply the heat flow balance constraint condition to ensure that the heat load migration path and the radio frequency power distribution satisfy the first law of thermodynamics; Verify the convergence of the network model through Monte Carlo random perturbation testing, and reverse correct the adjustment step of the weight coefficient and the update threshold of the margin parameter according to the statistical distribution characteristics of the perturbation response data to achieve the dynamic matching between the model parameters and the physical field changes.

4. The multi-level intelligent overheat protection method based on a gallium nitride module antenna according to claim 1, characterized in that Analyze the optimal heat load migration path based on the virtual heat capacity network model, and synchronously tune the input impedance matching network of the target unit according to the optimal heat load migration path to compensate for the beam phase distortion caused by heat migration. Specifically: Based on the heat conduction weight coefficient and the radio frequency power carrying margin parameter, use the Dijkstra algorithm to determine the minimum thermal resistance path between each target unit, and generate a candidate heat load migration path set; Combined with the gradient direction of the temperature field evolution trend, calculate the weighted average after multiplying the temperature gradient change rate between adjacent nodes on each candidate migration path by the phase change heat response intensity factor to obtain the path heat flow dynamic equilibrium factor; Compare the path heat flow dynamic equilibrium factor with the preset heat flow equilibrium threshold, eliminate the candidate migration paths with factor values greater than the preset heat flow equilibrium threshold, and retain the candidate migration paths with factor values not greater than the preset heat flow equilibrium threshold to obtain the optimal heat load migration path set; Extract the heat-induced dielectric constant change amount of the preset node according to the heat flow distribution characteristics on the optimal migration path, and determine the corresponding beam phase offset compensation value based on the radio frequency power cumulative effect index; Dynamically adjust the input impedance matching network of the target unit using the heat-induced dielectric constant change amount and the phase offset compensation value to compensate for the beam phase distortion caused by heat migration.

5. The multi-level intelligent overheat protection method based on a gallium nitride module antenna according to claim 4, wherein, Dynamically adjust the input impedance matching network of the target unit using the heat-induced dielectric constant change amount and the phase offset compensation value to compensate for the beam phase distortion caused by heat migration. Specifically: According to the heat-induced dielectric constant change amount of each node on the optimal migration path, calculate its relative deviation rate from the nominal dielectric constant of the antenna unit substrate to generate a dielectric constant perturbation coefficient matrix; Decompose the phase offset compensation value into the phase gradient components in the azimuth dimension and the elevation dimension according to the preset beam pointing angle, and calculate the equivalent susceptance adjustment amount required for the matching network in combination with the dielectric constant perturbation coefficient matrix; Dynamically adjust the capacitance value distribution of the adjustable capacitor bank according to the equivalent susceptance adjustment amount to cancel out the reflection phase at the input end of the matching network and the heat-induced phase distortion component; Verify the phase compensation effect by real-time monitoring the antenna port standing wave ratio. If the residual phase error exceeds the preset error threshold, use the gradient descent method to iteratively optimize the mapping relationship between the dielectric constant perturbation coefficient and the susceptance adjustment amount; Update the power carrying margin parameter in the virtual heat capacity network according to the optimized mapping relationship to form a closed-loop adaptive matching of the thermal-electric-phase parameters.

6. The multi-level intelligent overheat protection method based on a gallium nitride module antenna according to claim 1, wherein Inject the boundary constraint conditions of the virtual heat capacity network model according to the real-time threat scenario characteristics, dynamically solve the game equilibrium solution of the heat load migration path and the beam parameter tuning amount, and generate a thermal-radio frequency joint control instruction set. Specifically: Obtain the electromagnetic interference intensity distribution and the heat radiation value of the threat scenario in real time, and quantify them into a heat flux density constraint vector and a radio frequency power limit matrix as the dynamic boundary conditions of the virtual heat capacity network model; Based on the boundary conditions, the heat conduction weight coefficient is constrained to generate a constrained thermal resistance topology map, and the heat flow accommodation space of each node is determined in combination with the RF power carrying margin parameters; A heat flux density constraint vector is introduced into the thermal resistance topology map for bidirectional search, and the path branches that exceed the heat flux accommodation space are eliminated to obtain a set of feasible heat load migration paths that meet the boundary conditions. According to the RF power limitation matrix of each node on the feasible migration path set, the maximum allowable beam scanning angle and power tuning range corresponding to the path are determined, and a path-beam coupling parameter table is generated; By solving the multi-objective optimization function of minimum path thermal resistance and maximum beam pointing stability, the Pareto optimal solution set is screened out in the path-beam coupling parameter table; The thermal resistance gradient change rate in the Pareto optimal solution set is linearly weighted fused with the beam phase tuning amount to generate a thermal-RF joint control instruction set including power redistribution instructions, heat flow migration timing and phase compensation parameters.

7. The multi-level intelligent overheat protection method based on a gallium nitride module antenna according to claim 1, characterized in that, The multi-level intelligent overheat protection method further comprises the following steps: Collect the temperature field distribution data of the antenna unit and the beam pointing accuracy deviation value after the instruction set is executed in real time, extract the offset between the actual heat flow migration path and the theoretical path, and generate the heat conduction error vector; The correlation between the heat conduction error vector and the RF power loading timing is analyzed to obtain the dynamic correction factor of the heat conduction weight coefficient, and the weight distribution ratio of adjacent nodes is adjusted in combination with the hysteresis time coefficient; The actual influence of the thermally induced dielectric constant disturbance is inferred based on the beam pointing accuracy deviation value, and the attenuation gradient of the RF power carrying margin parameter is updated in combination with the phase change thermal response intensity factor; The adjacency matrix of the virtual heat capacity network is reconstructed using the modified heat conduction weight coefficient and power carrying margin parameter, and the network convergence is verified by heat flow balance constraints. Based on the residual distribution of the predicted and measured values ​​of the node temperatures of the reconstructed network, the update step size of the weight coefficients and margin parameters is adaptively adjusted.

8. A multi-level intelligent overheat protection system based on a gallium nitride module antenna, which is applied to the steps of the multi-level intelligent overheat protection method based on the gallium nitride module antenna according to any one of claims 1 to 7, characterized in that, include: The GaN antenna array module uses a GaN-on-SiC broadband antenna array, which contains at least 8 radiating units. Each unit has a built-in temperature sensor and micro phase change energy storage material. The operating frequency band covers 1.2-5.8GHz, and the unit spacing is 0.75 times the center frequency wavelength. The thermal state dynamic monitoring module is integrated with the temperature sensor of the gallium nitride antenna unit, the phase change rate detection circuit of the phase change energy storage material, and the thermal inertia characteristic calculation unit. It outputs a thermal state vector set including the chip junction temperature gradient, substrate heat flux density, and the solid-liquid phase change ratio of the phase change material in real time. A virtual heat capacity network modeling module is used to dynamically calculate the heat conduction weight coefficient and RF power carrying margin parameter between adjacent antenna units according to the thermal state vector set, and generate a virtual heat capacity network topology diagram including the heat migration path priority and beam distortion compensation amount; The RF-thermal coupling compensation module includes an adjustable capacitor array and a distributed phase shifter. Based on the thermal migration path data of the virtual thermal capacitance network topology diagram, the reflection coefficient of the input impedance matching network is dynamically tuned to achieve beam pointing phase pre-distortion compensation within ±0.25dB. The threat scenario adaptive decision-making module dynamically calculates the Pareto optimal solutions of the thermal load migration path and beam synthesis parameters according to the modulation type and pulse width characteristics of real-time threat signals, and outputs an interference effectiveness enhancement instruction set with a frequency band priority of 1.2 - 1.6 GHz.

Images