A microwave assembly performance self-adaptive scheduling method and system

Abstract

The present application relates to the technical field of phase amplitude adjustment, in particular to a microwave assembly performance adaptive scheduling method and system, comprising the following steps: based on target frequency power and predicted temperature, solving predicted transmission characteristics with a digital twin model, analyzing a curve to determine key distortion points to generate a non-uniform digital compensation vector, interpolating and smoothing and adding a timestamp to synthesize a compensation configuration file, and finally collecting an error map of actual output and ideal target to correct the model. In the present application, by pre-solving the transmission characteristics of the assembly and taking into account environmental factors such as temperature, key distortion frequency points are identified and frequency sub-bands are divided, non-uniform digital compensation is generated to accurately counteract the loss of each frequency band, and task timestamp synchronization loading is combined to ensure real-time accuracy of scheduling, and the output and target error are continuously compared, the working parameters are associated to build a correction set to iteratively optimize the prediction accuracy, significantly improving the stability and consistency of microwave signals under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of phase amplitude adjustment technology, and in particular to a microwave component performance adaptive scheduling method and system. Background Technology

[0002] The field of phase and amplitude adjustment technology mainly involves waveform modulation of microwave signals during transmission and processing. Core aspects include the precise adjustment of the phase and amplitude of microwave signals to achieve matching, stability, and efficient transmission in complex electromagnetic environments. This technology systematically covers microwave signal frequency characteristic management, power distribution control, and waveform distortion correction, and is widely used in communication systems, radar systems, and other high-frequency electronic devices. Traditional microwave component performance adaptive scheduling methods refer to the dynamic scheduling of the amplitude and phase of the output channel by introducing a preset reference signal for comparison during microwave component operation to cope with environmental changes or fluctuations in device parameters. This is typically achieved by combining a power splitter and a phase shifter to quantify and feedback the output energy. This method generally relies on a power detector to quantify the output energy and uses a phase shift adjustment circuit for real-time compensation, thus structurally achieving adaptive scheduling of component performance through fixed circuit components.

[0003] Traditional adaptive scheduling methods for microwave components heavily rely on quantized feedback of the output signal. This compensation mechanism, based on fixed circuit elements and preset reference signals, exhibits significant response lag when facing rapidly changing task instructions or dynamic electromagnetic environments. The unified scheduling of power splitters and phase shifters makes it difficult to finely correct non-uniform distortions occurring at different frequency points, resulting in poor compensation effects. For example, when a sudden change in operating temperature causes device parameter drift, the system can only initiate compensation after detecting a change in output energy, failing to proactively avoid signal distortion and thus affecting the stability and reliability of communication or radar systems in critical applications. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a microwave component performance adaptive scheduling method and system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a microwave component performance adaptive scheduling method, comprising the following steps:

[0006] S1: Obtain the target frequency and power of the task instruction set at the next moment, monitor the real-time temperature sequence and calculate the predicted temperature, and use a digital twin model to solve the transmission characteristic curve of the prediction component based on the target frequency, the power and the predicted temperature.

[0007] S2: Analyze the transmission characteristic curve of the prediction component, calculate the loss change rate to determine the key distortion frequency point, divide the frequency sub-band with the key distortion frequency point and calculate the average prediction loss, and invert the average prediction loss to generate a non-uniform digital compensation vector.

[0008] S3: Smooth the non-uniform digital compensation vector using an interpolation algorithm to synthesize a compensation configuration file, obtain the execution timestamp of the task instruction set at the next moment and append it to the compensation configuration file to generate a synchronously loaded compensation configuration file;

[0009] S4: Collect the actual output spectrum and operating parameters of the microwave component, obtain the ideal target spectrum, calculate the error spectrum between the actual output spectrum and the ideal target spectrum, associate the operating parameters with the error spectrum to construct a model correction parameter set, and adjust the digital twin model in conjunction with the synchronous loading compensation configuration file.

[0010] As a further aspect of the present invention, the transmission characteristic curve of the prediction component includes frequency response, phase characteristics and gain flatness, the non-uniform digital compensation vector specifically includes frequency sub-band division basis, sub-band compensation value and vector dimension, the synchronous loading compensation configuration file includes smoothing compensation curve data, task synchronization time code and file loading instruction, and the model correction parameter set specifically refers to the spectrum error matrix, parameter correlation weight and model correction coefficient.

[0011] As a further aspect of the present invention, the step of obtaining the transmission characteristic curve of the prediction component specifically includes:

[0012] S101: Obtain the instruction set of the task to be executed at the next moment, and parse the target frequency and power from it through instruction decoding operation. At the same time, call the temperature sensor array to monitor the real-time temperature of several key nodes on the surface of the microwave component with a preset sampling period to form a real-time temperature sequence. Aggregate the target frequency, the power and the real-time temperature sequence into a structured data entity to generate a set of component operating status parameters.

[0013] S102: Based on the set of component operating status parameters, extract the real-time temperature sequence, perform time-series difference operation on the temperature data at multiple time points within the real-time temperature sequence, construct an autoregressive model to fit the temperature change pattern, and extrapolate the temperature at the next prediction time based on the autoregressive model to obtain the predicted temperature.

[0014] S103: For the target frequency, the power and the predicted temperature in the component operating state parameter set, these three parameters are input as boundary conditions into the digital twin model to drive the model to perform multi-physics coupling solution. The electromagnetic response and heat distribution under the combined action of the target frequency, the power and the predicted temperature are solved by the finite element method. The transmission loss values ​​corresponding to the differential frequency points in the solution results are connected to establish the predicted component transmission characteristic curve.

[0015] As a further aspect of the present invention, the step of obtaining the non-uniform digital compensation vector specifically includes:

[0016] S201: Perform a first-order difference operation along the frequency axis of the transmission characteristic curve of the prediction component to calculate the loss change rate, set a reference value for the loss change rate, traverse all frequency points, compare the loss change rate corresponding to each point with the reference value for the loss change rate, filter points whose change rate values ​​exceed the reference value for the loss change rate, and generate key distortion frequency points.

[0017] S202: Call the key distortion frequency points and sort them in ascending order. Using the adjacent key distortion frequency points as interval boundaries, divide the frequency axis of the transmission characteristic curve of the prediction component into multiple frequency sub-bands, establish a frequency sub-band set, and for each frequency sub-band in the frequency sub-band set, extract all the predicted loss values ​​and calculate the arithmetic mean to obtain the average predicted loss.

[0018] S203: For the value corresponding to each frequency sub-band in the average predicted loss, perform an inversion operation one by one, and combine all the new values ​​obtained after the operation according to the original frequency order of the frequency sub-band set to establish a non-uniform digital compensation vector.

[0019] As a further aspect of the present invention, the step of obtaining the synchronous loading compensation configuration file specifically includes:

[0020] S301: For the discrete data points arranged in time sequence in the non-uniform digital compensation vector, a predetermined number of new compensation values ​​are calculated between each pair of adjacent original data points according to their sequence index positions using a preset interpolation algorithm. All the calculated new compensation values ​​are then inserted between the original data points of the non-uniform digital compensation vector in index order to reconstruct and expand the data sequence and generate a compensation configuration file.

[0021] S302: Obtain the next moment task instruction set, retrieve the timestamp field in the instruction set data structure, extract the future execution time information recorded in Coordinated Universal Time format, without performing any format conversion or time zone adjustment operation, and directly establish the extracted raw time data as the execution timestamp;

[0022] S303: Call the compensation configuration file and the execution timestamp, take the execution timestamp as the current data element, append it to the end of the compensation configuration file data structure according to the preset binary encapsulation rules, establish a combined data structure, and generate a synchronously loaded compensation configuration file.

[0023] As a further aspect of the present invention, the step of obtaining the model correction parameter set specifically includes:

[0024] S401: Collect the actual output spectrum and operating parameters of the microwave component, obtain the ideal target spectrum, perform point-by-point subtraction operation on the power values ​​of the actual output spectrum and the ideal target spectrum at the same set of sampling frequency points, and perform structured pairing of the frequency point deviation value and the corresponding frequency point information to generate a spectrum deviation data sequence;

[0025] S402: Call the frequency point deviation value and the frequency point information in the spectrum deviation data sequence, reconstruct multiple frequency point deviation values ​​into a two-dimensional data matrix based on the frequency point information, wherein the row index and column index of the two-dimensional data matrix correspond to the working time and frequency point respectively, and establish an error spectrum;

[0026] S403: Call the working parameters and the established error map, quantize multiple parameter items of the working parameters into input vectors, quantize the matrix elements of the error map into target vectors, iteratively adjust the internal values ​​of the transformation matrix until the distance between the result vector calculated by the transformation matrix and the target vector is less than a preset convergence threshold, and construct a model correction parameter set.

[0027] As a further aspect of the present invention, the multiphysics coupling solution specifically comprises:

[0028] The target frequency and the power are obtained, the electromagnetic field distribution is calculated based on Maxwell's equations, the ohmic loss caused by the power is solved, and the ohmic loss is used as a heat source input into the heat conduction equation.

[0029] The predicted temperature is then obtained, and the heat conduction equation is solved using the predicted temperature as an initial condition to obtain the steady-state temperature field distribution of the microwave component. The conductivity and dielectric constant of the material in the digital twin model are then updated based on the steady-state temperature field distribution.

[0030] Finally, based on the updated material conductivity and dielectric constant, the frequency-dependent transmission loss of the digital twin model is calculated. The calculation formula is as follows:

[0031] ;

[0032] in, Represents the target frequency and the predicted temperature Total transmission loss Represents the reference temperature The baseline transmission loss is as follows. The temperature loss coefficient represents the material-related properties. This represents the frequency loss coefficient associated with the geometry of the microwave component.

[0033] As a further aspect of the present invention, the steps for constructing the model correction parameter set are as follows:

[0034] Initialize the parameter association weight matrix, set the learning rate and the upper limit of the number of iterations, normalize each parameter item of the working parameters, and construct the input feature vector;

[0035] For each iteration, the weight matrix associated with the input feature vector and the current parameters is obtained, the prediction error vector is calculated by matrix multiplication, and the error map is used as the true error vector.

[0036] The gradient of the loss function is calculated based on the predicted error vector and the true error vector, and the parameter association weight matrix is ​​updated according to the following formula. :

[0037] ;

[0038] in, This represents the parameter association weight matrix after the (k+1)th iteration. This represents the parameter association weight matrix at the k-th iteration. The learning rate represents the step size used to control the update step. Represents the loss function The associated weight matrix of the currently described parameters The gradient;

[0039] Repeat the update operation until the loss function is reached. If the value is lower than the preset convergence threshold or reaches the upper limit of the number of iterations, the parameter association weight matrix is ​​used as the core component of the model correction parameter set, and the model correction parameter set is established.

[0040] As a further aspect of the present invention, the interpolation algorithm is a cubic spline interpolation algorithm, and the specific steps for smoothing the non-uniform digital compensation vector are as follows:

[0041] The discrete data points in the non-uniform digital compensation vector are obtained and used as spline nodes. For each pair of adjacent spline nodes, a cubic polynomial is constructed.

[0042] To ensure the overall smoothness of the curve, constraints are imposed on all connection points, requiring that at each spline node, the cubic polynomial function values ​​corresponding to two adjacent intervals are equal, the first derivative values ​​are equal, and the second derivative values ​​are equal.

[0043] By combining all the aforementioned constraints, a system of linear equations is formed. Solving the system of linear equations yields the coefficients of each of the cubic polynomials.

[0044] Based on the obtained coefficients, a preset number of interpolation points are calculated in each interval, and all interpolation points are merged with the original spline nodes to generate a compensation configuration file.

[0045] A microwave component performance adaptive scheduling system, the microwave component performance adaptive scheduling system being used to implement the above-mentioned microwave component performance adaptive scheduling method, the system comprising:

[0046] The component performance prediction module is used to obtain the target frequency and power of the task instruction set at the next moment, monitor the real-time temperature sequence of the microwave component and calculate the predicted temperature, and calculate the predicted component transmission characteristic curve using a digital twin model based on the target frequency, the power and the predicted temperature, and transmit the predicted component transmission characteristic curve to the distortion compensation calculation module.

[0047] The distortion compensation solution module is used to analyze the transmission characteristic curve of the prediction component, calculate the loss change rate to determine the key distortion frequency point, divide the frequency sub-band with the key distortion frequency point and calculate the average prediction loss, invert the average prediction loss to generate a non-uniform digital compensation vector, and transmit the non-uniform digital compensation vector to the scheduling instruction synthesis module.

[0048] The scheduling instruction synthesis module is used to smooth the discrete points of the non-uniform digital compensation vector transmitted by the distortion compensation solution module using an interpolation algorithm to synthesize a compensation configuration file, obtain the execution timestamp of the task instruction set and append it to the compensation configuration file, generate a synchronous loading compensation configuration file, and transmit the synchronous loading compensation configuration file to the twin model correction module.

[0049] The twin model correction module is used to collect the actual output spectrum and operating parameters of the microwave component after loading the synchronous loading compensation configuration file, obtain the ideal target spectrum, calculate the error spectrum between the actual output spectrum and the ideal target spectrum, associate the operating parameters with the error spectrum to construct a model correction parameter set, and adjust the digital twin model in conjunction with the synchronous loading compensation configuration file.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0051] In this invention, by pre-calculating the component transmission characteristics under future task instructions and taking into account environmental factors such as temperature changes, key distortion frequency points can be identified and frequency sub-bands can be divided. This generates non-uniform digital compensation, achieving precise inverse compensation for losses in different frequency bands. Combined with synchronous loading compensation using task execution timestamps, the timeliness and accuracy of scheduling are ensured. At the same time, the error between the actual output and the ideal target is continuously compared, and a correction set is constructed by correlating working parameters to iteratively optimize the accuracy of predictions. This significantly improves the transmission stability and performance consistency of microwave signals under complex and variable operating conditions. Attached Figure Description

[0052] Figure 1 This is a flowchart of the microwave component performance adaptive scheduling method of the present invention;

[0053] Figure 2 This is a flowchart illustrating the generation process of the transmission characteristic curve of the predictive component in this invention.

[0054] Figure 3 This is a flowchart of the non-uniform digital compensation vector generation process of the present invention;

[0055] Figure 4 This is a flowchart illustrating the process of generating a compensation configuration file for synchronous loading in this invention.

[0056] Figure 5 This is a flowchart illustrating the construction process of the parameter set for the digital twin model correction in this invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the software-based technical solution is described in detail below with reference to system architecture diagrams and embodiments. It should be understood that the specific embodiments described herein are only for explaining the technical solutions of this invention and do not constitute a limitation on the scope of protection.

[0058] In the description of this invention, the system architecture relationships or data processing flows indicated by terms such as "layer," "module," "interface," "data flow," "client," and "server" are all defined based on the architecture diagram or flowchart corresponding to the embodiments. This way of describing is only used to clearly illustrate the logical relationships between the elements in the technical solution, and not to limit the physical deployment form. The term "multiple" includes two or more technical units, including but not limited to multiple data nodes, processing threads, service instances, or functional components and other scalable elements. The specific number is determined according to the actual business scenario and needs to be specifically specified.

[0059] Please see Figure 1 and Figure 2 This invention provides a technical solution: a microwave component performance adaptive scheduling method, comprising the following steps:

[0060] S1: Obtain the target frequency and power of the task instruction set at the next moment, monitor the real-time temperature sequence and calculate the predicted temperature, and use a digital twin model to solve the transmission characteristic curve of the prediction component based on the target frequency, the power and the predicted temperature.

[0061] The specific steps for obtaining the transmission characteristic curve of the prediction component are as follows:

[0062] S101: Obtain the instruction set of the task to be executed at the next moment, and parse the target frequency and power from it through instruction decoding operation. At the same time, call the temperature sensor array to monitor the real-time temperature of several key nodes on the surface of the microwave component with a preset sampling period to form a real-time temperature sequence. Aggregate the target frequency, the power and the real-time temperature sequence into a structured data entity to generate a set of component operating status parameters.

[0063] S102: Based on the set of component operating status parameters, extract the real-time temperature sequence, perform time-series difference operation on the temperature data at multiple time points within the real-time temperature sequence, construct an autoregressive model to fit the temperature change pattern, and extrapolate the temperature at the next prediction time based on the autoregressive model to obtain the predicted temperature.

[0064] S103: For the target frequency, the power and the predicted temperature in the component operating state parameter set, these three parameters are input as boundary conditions into the digital twin model to drive the model to perform multi-physics coupling solution. The electromagnetic response and heat distribution under the combined action of the target frequency, the power and the predicted temperature are solved by the finite element method. The transmission loss values ​​corresponding to the differential frequency points in the solution results are connected to establish the predicted component transmission characteristic curve.

[0065] The transmission characteristic curves of the prediction component include frequency response, phase characteristics, and gain flatness.

[0066] The multiphysics coupling solution specifically involves: obtaining the target frequency and the power, calculating the electromagnetic field distribution based on Maxwell's equations, solving for the ohmic loss caused by the power, and inputting the ohmic loss as a heat source into the heat conduction equation.

[0067] The predicted temperature is then obtained, and the heat conduction equation is solved using the predicted temperature as an initial condition to obtain the steady-state temperature field distribution of the microwave component. The conductivity and dielectric constant of the material in the digital twin model are then updated based on the steady-state temperature field distribution.

[0068] Finally, based on the updated material conductivity and dielectric constant, the frequency-dependent transmission loss of the digital twin model is calculated. The calculation formula is as follows: ;

[0069] in, Represents the target frequency and the predicted temperature Total transmission loss Represents the reference temperature The baseline transmission loss is as follows. The temperature loss coefficient represents the material-related properties. This represents the frequency loss coefficient associated with the geometry of the microwave component.

[0070] S1: The next task instruction set to be executed is received via the bus interface. The instruction format is "CMD_SET_FREQ_14.5_PWR_50_TSTART_20250912T103000Z". Through instruction decoding, the target frequency is determined to be 14.5GHz and the target power to be 50W. Simultaneously, an array of five temperature sensor nodes deployed on the surface of the microwave component housing is invoked. These five nodes are located near the power amplifier chip, at the filter input, the circulator output, the load resistor, and the component edge, respectively. The preset sampling period is 100 milliseconds, continuously monitoring and acquiring temperature data at the five most recent time points (t-400ms, t-300ms, t-200ms, t-100ms, t-0ms). The arithmetic mean of the temperature values ​​from the five nodes is taken to form a real-time temperature sequence. For example, at t-0ms, the acquired real-time temperature sequence is {70.1℃, 70.3℃, 70.4℃, 70.6℃, 70.7℃}. The target frequency of 14.5 GHz, the target power of 50 W, and the real-time temperature sequence are aggregated into a structured data entity to generate a set of component operating status parameters.

[0071] Based on the component's operating status parameter set, the real-time temperature sequence {70.1, 70.3, 70.4, 70.6, 70.7} (unit: °C) was extracted. First-order time-series difference operations were performed on the temperature data at multiple time points within the sequence to obtain the difference sequence: {70.3-70.1, 70.4-70.3, 70.6-70.4, 70.7-70.6}, i.e., {0.2, 0.1, 0.2, 0.1}. A first-order autoregressive model (AR(1)) was constructed to fit the temperature change pattern. The model expression is a first-order autoregressive equation, which expresses that the temperature increment at the next moment is a linear function of the temperature increment at the current moment. The autoregressive coefficients were calculated by fitting the difference sequence using the least squares method. It is 0.85, constant term The value is 0.02. Based on this model, the temperature increment at the next prediction time (t+100ms) is calculated using extrapolation: ℃. Add this increment to the latest temperature value to obtain the predicted temperature: ℃.

[0072] The target frequency of 14.5 GHz, the target power of 50 W, and the calculated predicted temperature of 70.805 °C were used as boundary conditions input into the digital twin model. This model incorporates the microwave component's three-dimensional geometry based on design drawings, as well as material properties (such as the dielectric constant of the substrate and the conductivity of the copper foil). First, based on Maxwell's equations, using a frequency of 14.5 GHz and a power of 50 W as the excitation source, the electromagnetic field distribution inside the component was calculated using the finite element method, particularly the current density in transmission lines and passive devices. Based on the current density and the resistivity of the materials, the ohmic losses caused by the power were solved, which are represented in the model as spatially distributed heat sources. For example, the calculated heat flux density at the power amplifier output matching network is... The ohmic loss is then used as a heat source input into the heat conduction equation. The predicted temperature of 70.805℃ is then used as the initial temperature condition for the entire component, and the heat conduction equation is solved to obtain the transient temperature field distribution of the component at the next moment. This distribution shows that the junction temperature of the power amplifier chip reaches 95.2℃, while the edge temperature of the component is 71.5℃. Based on this steady-state temperature field distribution, the conductivity and dielectric constant of the materials at various locations in the digital twin model are updated by calling the material temperature property database. For example, the dielectric constant of the substrate material near the power amplifier chip (assumed to be Rogers4350B) is adjusted from 3.48 at room temperature to 3.51 at 95.2℃. Finally, based on the updated material conductivity and dielectric constant, the electromagnetic field is recalculated to calculate the frequency-dependent transmission loss at a frequency of 14.5GHz and an equivalent operating temperature of 70.805℃. .

[0073] The frequency-related transmission loss The calculation formula is: .

[0074] In the above formula for calculating frequency-dependent transmission loss, the logic of the parameters and their interactions lies in: total transmission loss It consists of three linearly superimposed parts, the first part being at the reference frequency. and reference temperature The reference transmission loss The first part reflects the inherent loss baseline of the component; the second part is the loss increment introduced by temperature change, which is determined by the material-related temperature loss coefficient. Operating temperature and reference temperature The difference determines the third part, which is the increase in dispersion loss introduced by frequency variation, determined by the frequency loss coefficient related to the microwave component geometry. With target operating frequency and reference frequency The squared difference is determined by this formula. This formula linearly superimposes the reference loss, temperature effect loss, and frequency effect loss through addition, thereby obtaining the result at a specific target frequency. and predicted temperature The predicted total transmission loss is as follows. Represents the total transmission loss at the target frequency f and the predicted temperature T, expressed in decibels (dB). Represents the reference frequency and reference temperature The reference transmission loss is expressed in decibels (dB). The temperature loss coefficient represents the material-related loss coefficient, used to quantify the loss change caused by a unit temperature change, and the unit is decibels per degree Celsius (dB / ℃). This represents the predicted operating temperature of the component, in degrees Celsius (°C). This represents a reference temperature, expressed in degrees Celsius (°C). The frequency loss coefficient represents the frequency loss associated with the geometry of the microwave component. It is used to quantify the degree to which the loss changes with the square of the frequency, and is measured in decibels per square hertz (dB / Hz²). Represents the target operating frequency, measured in Hertz (Hz); This represents the reference frequency, measured in Hertz (Hz). To determine the parameters in the formula, perform the following steps:

[0075] 1. and Obtaining: Setting the industry standard reference temperature Set the center frequency of the component's operating frequency band to 25℃. It is 14.0GHz.

[0076] 2. Data acquisition: The transmission loss of the microwave component at a center frequency of 14.0 GHz was measured using a vector network analyzer (VNA) at 25°C, and the measured value was 1.20 dB. Therefore, dB.

[0077] 3. Setup process: To determine the material-related temperature loss coefficient, a specific experimental verification was conducted. The microwave component was placed in a temperature-controlled chamber, and the test port of the VNA was connected to it via a high-temperature resistant cable. At a frequency of 14.0 GHz, the transmission loss was measured at 25℃, 45℃, 65℃, and 85℃, and the data are shown in Table 1.

[0078] Table 1. Experimental data on temperature loss characteristics

[0079]

[0080] As shown in Table 1, linear regression analysis was performed on the experimental data, and the loss value was correlated with the temperature change value. There is a linear relationship between them. The loss changes are 0.04dB, 0.08dB, and 0.12dB, corresponding to temperature changes of 20℃, 40℃, and 60℃. Calculate the loss change rate per unit temperature, i.e., the temperature loss coefficient. : dB / ℃

[0081] 4. Setup process: To determine the frequency loss coefficient related to the geometry, the transmission loss of the microwave assembly in the 12GHz to 16GHz band was scanned using a VNA at a reference temperature of 25°C. Using the loss at 14GHz as a baseline, the loss due to frequency variations was calculated. The total losses at the three frequency points of 12GHz, 14GHz, and 16GHz were 1.12dB, 1.20dB, and 1.31dB, respectively. The corresponding frequency-dependent losses are... The values ​​are -0.08dB, 0dB, and 0.11dB, respectively. This is achieved by solving the system of equations. For example, using data points from 12GHz and 16GHz to perform calculations and take the average:

[0082] Depend on ,get dB / Hz²;

[0083] Depend on ,get dB / Hz²;

[0084] Take the average value to get dB / Hz².

[0085] 5. Parameter assignment and calculation:

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092]

[0093] Substitute into the formula to calculate:

[0094]

[0095] The results show that the predicted total transmission loss is 1.3156 dB at the target frequency of 14.5 GHz and the predicted temperature of 70.805 °C. By repeating this calculation in 10 MHz steps across the entire operating frequency band, a complete predicted component transmission characteristic curve can be obtained, which includes information on frequency response, phase characteristics, and gain flatness.

[0096] The specific method for obtaining the phase characteristics and gain flatness information in the transmission characteristic curve of the aforementioned predictive component is as follows: When performing multi-physics coupling calculations and obtaining the electromagnetic response of the updated material parameters, the finite element method solver outputs the transmission phase corresponding to each frequency point while calculating the transmission loss at each frequency point. Connecting the transmission phase values ​​corresponding to all frequency points forms the predicted phase characteristic curve. Gain flatness is then calculated from the loss data in the obtained predicted component transmission characteristic curves. Specifically, within a specified operating frequency band, such as 14.0 GHz to 15.0 GHz, the predicted transmission loss for all frequency points within that band is determined. The difference between the maximum and minimum values ​​of the loss is the gain flatness of that frequency band. For example, if the maximum loss in this frequency band is 1.42dB and the minimum loss is 1.35dB, then the gain flatness is... dB.

[0097] Please see Figure 1 and Figure 3 S2: Analyze the transmission characteristic curve of the prediction component, calculate the loss change rate to determine the key distortion frequency point, divide the frequency sub-band with the key distortion frequency point and calculate the average prediction loss, and invert the average prediction loss to generate a non-uniform digital compensation vector.

[0098] The specific steps for obtaining the non-uniform digital compensation vector are as follows:

[0099] S201: Perform a first-order difference operation along the frequency axis of the transmission characteristic curve of the prediction component to calculate the loss change rate, set a reference value for the loss change rate, traverse all frequency points, compare the loss change rate corresponding to each point with the reference value for the loss change rate, filter points whose change rate values ​​exceed the reference value for the loss change rate, and generate key distortion frequency points.

[0100] S202: Call the key distortion frequency points and sort them in ascending order. Using the adjacent key distortion frequency points as interval boundaries, divide the frequency axis of the transmission characteristic curve of the prediction component into multiple frequency sub-bands, establish a frequency sub-band set, and for each frequency sub-band in the frequency sub-band set, extract all the predicted loss values ​​and calculate the arithmetic mean to obtain the average predicted loss.

[0101] S203: For the value corresponding to each frequency sub-band in the average predicted loss, perform an inversion operation one by one, and combine all the new values ​​obtained after the operation according to the original frequency order of the frequency sub-band set to establish a non-uniform digital compensation vector.

[0102] The non-uniform digital compensation vector specifically refers to the frequency sub-band division criteria, sub-band compensation values, and vector dimensions.

[0103] S2: Analyze the predicted component transmission characteristic curve generated by step S1. This curve is a set of discrete loss data points covering the 12GHz to 16GHz frequency band in 10MHz increments. Along the frequency axis of this curve, starting from 12GHz, perform a first-order difference operation on the transmission loss values ​​of each pair of adjacent frequency points to calculate the loss change rate. For example, if the loss at 14.20GHz is 1.38dB and the loss at 14.21GHz is 1.45dB, then the loss change rate within this 10MHz interval is... To establish a baseline value for the loss change rate, extensive simulations and field measurements were conducted on the microwave component within its normal operating range (temperature 20℃-80℃, power 10W-60W). Loss change rate data under different operating conditions were recorded and statistically analyzed. The 90th percentile value among all change rate data was selected as the baseline value. This value was calculated to be 0.04 dB / 10 MHz, which is the baseline value for the loss change rate. All frequency points were iterated, and the absolute value of the calculated loss change rate at each point was compared with the baseline value of 0.04. For example, the calculated change rate of 0.07 at 14.20 GHz is greater than 0.04, so the frequency points of 14.20 GHz and 14.21 GHz were marked as candidate points. When the change rate at multiple consecutive points exceeds the baseline value, the start and end points of that interval were selected. After traversing the entire frequency band, all the marked frequency points together constitute the set of key distortion frequency points. For example, the selected key distortion frequency points are {13.52GHz, 14.18GHz, 14.85GHz}.

[0104] The selected set of critical distortion frequencies {13.52GHz, 14.18GHz, 14.85GHz} is retrieved and sorted in ascending order, with the result remaining unchanged. The entire operating frequency band (12GHz-16GHz) is divided into multiple frequency sub-bands using adjacent critical distortion frequencies as interval boundaries. The specific sub-bands are divided as follows: Sub-band 1 is [12.00GHz, 13.52GHz], Sub-band 2 is (13.52GHz, 14.18GHz], Sub-band 3 is (14.18GHz, 14.85GHz], and Sub-band 4 is (14.85GHz, 16.00GHz). These four frequency sub-bands constitute a frequency sub-band set. For each sub-band in this set, the predicted loss values ​​for all frequency points within that sub-band are extracted from the transmission characteristic curve of the predictive component. Taking sub-band 2 as an example, the predicted loss values ​​corresponding to all frequency points with a step size of 10MHz between 13.53GHz and 14.18GHz are extracted. Assuming the obtained data sequence is {1.28dB, 1.29dB, ..., 1.35dB}, a total of 66 data points are obtained. The arithmetic mean of these 66 loss values ​​is calculated as follows: dB. This is the average predicted loss of subband 2. The average predicted loss of the other subbands is calculated using the same method, resulting in a set of average predicted loss values ​​corresponding to all subbands, for example: {subband 1: 1.15dB, subband 2: 1.31dB, subband 3: 1.42dB, subband 4: 1.25dB}.

[0105] For the calculated average predicted loss sequence {1.15, 1.31, 1.42, 1.25} (in dB), each value is inverted (multiplied by -1). The resulting sequence is {-1.15, -1.31, -1.42, -1.25}. These new values ​​are then combined according to the original frequency order of their corresponding frequency sub-band sets. This combination establishes a data structure that defines the frequency range to which each compensation value applies. The final non-uniform digital compensation vector is a data structure containing the frequency sub-band division criteria, sub-band compensation values, and vector dimension information. The specific format is as follows: {Dimension: 4, Sub-band 1: {Range: [12.00GHz, 13.52GHz], Compensation value: -1.15dB}, Sub-band 2: {Range: (13.52GHz, 14.18GHz], Compensation value: -1.31dB}, Sub-band 3: {Range: (14.18GHz, 14.85GHz], Compensation value: -1.42dB}, Sub-band 4: {Range: (14.85GHz, 16.00GHz], Compensation value: -1.25dB}}.

[0106] Please see Figure 1 and Figure 4S3: Smooth the non-uniform digital compensation vector using an interpolation algorithm to synthesize a compensation configuration file, obtain the execution timestamp of the task instruction set at the next moment and append it to the compensation configuration file to generate a synchronously loaded compensation configuration file;

[0107] The specific steps for obtaining the synchronous loading compensation configuration file are as follows:

[0108] S301: For the discrete data points arranged in time sequence in the non-uniform digital compensation vector, a predetermined number of new compensation values ​​are calculated between each pair of adjacent original data points according to their sequence index positions using a preset interpolation algorithm. All the calculated new compensation values ​​are then inserted between the original data points of the non-uniform digital compensation vector in index order to reconstruct and expand the data sequence and generate a compensation configuration file.

[0109] S302: Obtain the next moment task instruction set, retrieve the timestamp field in the instruction set data structure, extract the future execution time information recorded in Coordinated Universal Time format, without performing any format conversion or time zone adjustment operation, and directly establish the extracted raw time data as the execution timestamp;

[0110] S303: Call the compensation configuration file and the execution timestamp, take the execution timestamp as the current data element, append it to the end of the compensation configuration file data structure according to the preset binary encapsulation rules, establish a combined data structure, and generate a synchronously loaded compensation configuration file.

[0111] The interpolation algorithm is a cubic spline interpolation algorithm. The specific steps for smoothing the non-uniform digital compensation vector are as follows: obtain discrete data points in the non-uniform digital compensation vector and use them as spline nodes. For each pair of adjacent spline nodes, construct a cubic polynomial.

[0112] To ensure the overall smoothness of the curve, constraints are imposed on all connection points, requiring that at each spline node, the cubic polynomial function values, first derivative values, and second derivative values ​​of the corresponding two adjacent intervals are equal.

[0113] By combining all the aforementioned constraints, a system of linear equations is formed. Solving the system of linear equations yields the coefficients of each of the cubic polynomials.

[0114] Based on the obtained coefficients, a preset number of interpolation points are calculated in each interval, and all interpolation points are merged with the original spline nodes to generate a compensation configuration file.

[0115] The synchronous loading compensation configuration file includes smoothing compensation curve data, task synchronization timecode, and file loading instructions.

[0116] S3: For the non-uniform digital compensation vector generated in step S2, extract the discrete data points as spline nodes for cubic spline interpolation. These nodes are the boundary points of each frequency sub-band and their corresponding compensation values, specifically: (12.00GHz, -1.15dB), (13.52GHz, -1.15dB), (14.18GHz, -1.31dB), (14.85GHz, -1.42dB), (16.00GHz, -1.25dB). Note that at the sub-band boundary, the compensation value undergoes a step change. A very small frequency offset (e.g., 0.01GHz) is introduced on both sides of the boundary point to define two different nodes, or the compensation value of the sub-band is directly used as the function value of the sub-band's end point. In this embodiment, the average compensation value of the sub-band is assigned to the end point of the sub-band. For each interval formed by adjacent spline nodes, such as the first interval [12.00GHz, 13.52GHz], a cubic polynomial is constructed. Constraints are imposed: At the internal nodes (13.52GHz, 14.18GHz, 14.85GHz), the cubic polynomial function values ​​corresponding to the left and right intervals must be equal, the first derivative values ​​must be equal, and the second derivative values ​​must also be equal. For example, at the 13.52GHz node, the following constraints are required: , ,as well as By combining all constraints at all nodes, a large system of linear equations is formed. Solving this system of equations uniquely determines all coefficients of all cubic polynomials. Based on the obtained coefficients, a predetermined number of interpolation points are calculated within each interval at a preset frequency resolution (e.g., 1 MHz). For example, within the interval [12.00 GHz, 13.52 GHz], new compensation values ​​are calculated at f = 12.01 GHz, 12.02 GHz, ..., 13.51 GHz. All calculated interpolation points are merged with the original spline nodes and arranged in ascending frequency order to form a high-density data sequence, which is the compensation profile.

[0117] Obtain the next time step task instruction set, namely "CMD_SET_FREQ_14.5_PWR_50_TSTART_20250912T103000Z". Retrieve the timestamp field "TSTART_20250912T103000Z" within the instruction set data structure, and extract the future execution time information "2025-09-12T10:30:00.000Z" recorded in Coordinated Universal Time (UTC) format. Without performing any format conversion or time zone adjustment, directly construct the execution timestamp from this raw time data string "2025-09-12T10:30:00.000Z".

[0118] The compensation configuration file (i.e., the smooth, high-density compensation data sequence) generated in the previous step is called along with the newly established execution timestamp "2025-09-12T10:30:00.000Z". This execution timestamp string is treated as an independent data element and appended to the end of the compensation configuration file data structure according to the preset binary encapsulation rules to create a combined data structure and generate a synchronously loaded compensation configuration file.

[0119] Please see Figure 1 and Figure 5 S4: Collect the actual output spectrum and operating parameters of the microwave component, obtain the ideal target spectrum, calculate the error spectrum between the actual output spectrum and the ideal target spectrum, associate the operating parameters with the error spectrum to construct a model correction parameter set, and adjust the digital twin model in conjunction with the synchronous loading compensation configuration file.

[0120] The specific steps for obtaining the model correction parameter set are as follows:

[0121] S401: Collect the actual output spectrum and operating parameters of the microwave component, obtain the ideal target spectrum, perform point-by-point subtraction operation on the power values ​​of the actual output spectrum and the ideal target spectrum at the same set of sampling frequency points, and perform structured pairing of the frequency point deviation value and the corresponding frequency point information to generate a spectrum deviation data sequence;

[0122] S402: Call the frequency point deviation value and the frequency point information in the spectrum deviation data sequence, reconstruct multiple frequency point deviation values ​​into a two-dimensional data matrix based on the frequency point information, wherein the row index and column index of the two-dimensional data matrix correspond to the working time and frequency point respectively, and establish an error spectrum;

[0123] S403: Call the working parameters and the established error map, quantize multiple parameter items of the working parameters into input vectors, quantize the matrix elements of the error map into target vectors, iteratively adjust the internal values ​​of the transformation matrix until the distance between the result vector calculated by the transformation matrix and the target vector is less than a preset convergence threshold, and construct a model correction parameter set.

[0124] The specific steps for constructing the model correction parameter set are as follows: initializing the parameter association weight matrix, setting the upper limit of the learning rate and the number of iterations, normalizing each parameter item of the working parameters, and constructing the input feature vector;

[0125] For each iteration, the weight matrix associated with the input feature vector and the current parameters is obtained, the prediction error vector is calculated by matrix multiplication, and the error map is used as the true error vector.

[0126] The gradient of the loss function is calculated based on the predicted error vector and the true error vector, and the parameter association weight matrix is ​​updated according to the following formula. : ;

[0127] in, This represents the parameter association weight matrix after the (k+1)th iteration. This represents the parameter association weight matrix at the k-th iteration. The learning rate represents the step size used to control the update step. Represents the loss function The associated weight matrix of the currently described parameters The gradient;

[0128] Repeat the update operation until the loss function is reached. If the value is lower than the preset convergence threshold or reaches the upper limit of the number of iterations, the parameter association weight matrix is ​​used as the core component of the model correction parameter set, and the model correction parameter set is established.

[0129] The model correction parameter set specifically refers to the spectral error matrix, parameter correlation weights, and model correction coefficients.

[0130] S4: After the microwave component is actually running according to the compensation configuration, the actual output spectrum of the component is acquired in the 12GHz to 16GHz frequency band with a resolution of 1MHz using a coupler and a high-precision spectrum analyzer. Simultaneously, the operating parameters are recorded by the monitoring system; for example, the actual measured average component temperature is 71.2℃, the output power is 49.8W, and the DC bias voltage is 28.05V. The ideal target spectrum corresponding to the task command is obtained. In this example, the ideal target spectrum is a reference spectrum with constant gain within the operating frequency band. For the actual output spectrum and the ideal target spectrum, at the same set of sampling frequency points (12.00GHz, 12.01GHz, ..., 16.00GHz), the power values ​​(usually in dBm) are subtracted point by point. For example, at the 14.50GHz frequency point, the ideal output power is 47.00dBm (corresponding to 50W), and the actual acquired output power is 46.95dBm. The deviation value at this frequency point is... dB. The calculated frequency offset value (-0.05dB) and the corresponding frequency information (14.50GHz) are then structurally paired. This operation is repeated for all frequency points to generate a spectral offset data sequence.

[0131] The previously generated spectral deviation data sequence, containing frequency points and their corresponding deviation values, is retrieved. This one-dimensional sequence data is then reconstructed into a two-dimensional data matrix based on the acquisition time and frequency information. The matrix's row indices correspond to the working time (e.g., acquisition once per second for 60 seconds), and the column indices correspond to frequency points (e.g., 4001 frequency points from 12.00 GHz to 16.00 GHz). Each element (i, j) in the matrix represents the spectral deviation value acquired at the i-th second and the j-th frequency point. In this way, an error spectrum reflecting the time-frequency variation of the error is established.

[0132] Table 2 Error Map Data Fragment Table

[0133]

[0134] As shown in Table 2, this table displays a small segment of the error spectrum. The recorded operating parameters (average temperature 71.2℃, output power 49.8W, bias voltage 28.05V) and the established error spectrum are retrieved. After normalizing multiple parameters (temperature, power, voltage), they are quantized into an input vector, for example... The row vectors of the error map matrix at a certain moment (e.g., the 1st second) are quantized into the target vector. Subsequent iterations adjust the transformation matrix. The internal values ​​of make the input vector Transformed matrix The calculated result vector With the target vector The distance between them is less than the preset convergence threshold.

[0135] To construct the model and refine the parameter set, a parameter-associated weight matrix is ​​first randomly initialized. Its dimensions are (number of frequency points in the error plot × number of operating parameters), for example (4001 × 3). Set the learning rate. The value is 0.01, determined through previous experiments. The maximum number of iterations is set to 1000, and the convergence threshold is... The collected operating parameters (temperature 71.2℃, power 49.8W, voltage 28.05V) were normalized based on their normal operating range (e.g., temperature ℃, power W, voltage V) to obtain the input feature vector. For each iteration, obtain the input feature vector. Weight matrix associated with the current parameters Through matrix multiplication Calculate the prediction error vector. Use the row vector at the corresponding time step from the error map as the true error vector. The loss function is calculated based on the predicted error vector and the true error vector. The gradient is calculated, and the parameter-associated weight matrix is ​​updated according to the gradient descent formula. : .in, Repeat this update operation until the loss function is reached. The value is lower than It may reach the upper limit of 1000 iterations. The final parameter-associated weight matrix is ​​obtained. Together with the error plot and the calculated model correction coefficients (e.g., from...), The correction ratios for the original model parameters are extracted, and together they form a model correction parameter set. This parameter set will be used to adjust the material parameters or loss model in the digital twin model.

[0136] A microwave component performance adaptive scheduling system is provided for executing the aforementioned microwave component performance adaptive scheduling method. The system includes:

[0137] The component performance prediction module is used to obtain the target frequency and power of the task instruction set at the next moment, monitor the real-time temperature sequence of the microwave component and calculate the predicted temperature, and calculate the predicted component transmission characteristic curve using a digital twin model based on the target frequency, the power and the predicted temperature, and transmit the predicted component transmission characteristic curve to the distortion compensation calculation module.

[0138] The distortion compensation solution module is used to analyze the transmission characteristic curve of the prediction component, calculate the loss change rate to determine the key distortion frequency point, divide the frequency sub-band with the key distortion frequency point and calculate the average prediction loss, invert the average prediction loss to generate a non-uniform digital compensation vector, and transmit the non-uniform digital compensation vector to the scheduling instruction synthesis module.

[0139] The scheduling instruction synthesis module is used to smooth the discrete points of the non-uniform digital compensation vector transmitted by the distortion compensation solution module using an interpolation algorithm to synthesize a compensation configuration file, obtain the execution timestamp of the task instruction set and append it to the compensation configuration file, generate a synchronous loading compensation configuration file, and transmit the synchronous loading compensation configuration file to the twin model correction module.

[0140] The twin model correction module is used to collect the actual output spectrum and operating parameters of the microwave component after loading the synchronous loading compensation configuration file, obtain the ideal target spectrum, calculate the error spectrum between the actual output spectrum and the ideal target spectrum, associate the operating parameters with the error spectrum to construct a model correction parameter set, and adjust the digital twin model in conjunction with the synchronous loading compensation configuration file.

[0141] The above embodiments illustrate preferred embodiments of the present invention. Any equivalent adjustments to the technical solution based on software engineering methods are within the scope of protection, including but not limited to: implementing algorithm logic using different programming languages, refactoring functional modules into services, adjusting data interaction protocols, and optimizing resource scheduling strategies. Any implementation scheme derived from reasonable modifications to the data processing flow, service call chain, or system architecture layer without departing from the core technology of the present invention should be considered within the scope of protection defined by the claims of the present invention.

Claims

1. A microwave component performance adaptive scheduling method, characterized in that, Includes the following steps: S1: Obtain the target frequency and power of the task instruction set at the next moment, monitor the real-time temperature sequence and calculate the predicted temperature, and use a digital twin model to solve the transmission characteristic curve of the prediction component based on the target frequency, the power and the predicted temperature. S101: Obtain the instruction set of the task to be executed at the next moment, and parse the target frequency and power from it through instruction decoding operation. At the same time, call the temperature sensor array to monitor the real-time temperature of several key nodes on the surface of the microwave component with a preset sampling period to form a real-time temperature sequence. Aggregate the target frequency, the power and the real-time temperature sequence into a structured data entity to generate a set of component operating status parameters. S102: Based on the set of component operating status parameters, extract the real-time temperature sequence, perform time-series difference operation on the temperature data at multiple time points within the real-time temperature sequence, construct an autoregressive model to fit the temperature change pattern, and extrapolate the temperature at the next prediction time based on the autoregressive model to obtain the predicted temperature. S103: For the target frequency, the power and the predicted temperature in the component operating state parameter set, these three parameters are input as boundary conditions into the digital twin model to drive the model to perform multi-physics coupling calculation. The electromagnetic response and heat distribution under the combined action of the target frequency, the power and the predicted temperature are solved by the finite element method. The transmission loss values ​​corresponding to the differential frequency points in the solution results are connected to establish the predicted component transmission characteristic curve. S2: Analyze the transmission characteristic curve of the prediction component, calculate the loss change rate to determine the key distortion frequency point, divide the frequency sub-band with the key distortion frequency point and calculate the average prediction loss, and invert the average prediction loss to generate a non-uniform digital compensation vector. S3: Smooth the non-uniform digital compensation vector using an interpolation algorithm to synthesize a compensation configuration file, obtain the execution timestamp of the task instruction set at the next moment and append it to the compensation configuration file to generate a synchronously loaded compensation configuration file; S4: Collect the actual output spectrum and operating parameters of the microwave component, obtain the ideal target spectrum, calculate the error spectrum between the actual output spectrum and the ideal target spectrum, associate the operating parameters with the error spectrum to construct a model correction parameter set, and adjust the digital twin model in conjunction with the synchronous loading compensation configuration file.

2. The microwave component performance adaptive scheduling method according to claim 1, characterized in that, The transmission characteristic curve of the prediction component includes frequency response, phase characteristics and gain flatness. The non-uniform digital compensation vector specifically includes the frequency sub-band division basis, sub-band compensation value and vector dimension. The synchronous loading compensation configuration file includes smoothing compensation curve data, task synchronization time code and file loading instructions. The model correction parameter set specifically refers to the spectrum error matrix, parameter correlation weights and model correction coefficients.

3. The microwave component performance adaptive scheduling method according to claim 1, characterized in that, The specific steps for obtaining the non-uniform digital compensation vector are as follows: S201: Perform a first-order difference operation along the frequency axis of the transmission characteristic curve of the prediction component to calculate the loss change rate, set a reference value for the loss change rate, traverse all frequency points, compare the loss change rate corresponding to each point with the reference value for the loss change rate, filter points whose change rate values ​​exceed the reference value for the loss change rate, and generate key distortion frequency points. S202: Call the key distortion frequency points and sort them in ascending order. Using the adjacent key distortion frequency points as interval boundaries, divide the frequency axis of the transmission characteristic curve of the prediction component into multiple frequency sub-bands, establish a frequency sub-band set, and for each frequency sub-band in the frequency sub-band set, extract all the predicted loss values ​​and calculate the arithmetic mean to obtain the average predicted loss. S203: For the value corresponding to each frequency sub-band in the average predicted loss, perform an inversion operation one by one, and combine all the new values ​​obtained after the operation according to the original frequency order of the frequency sub-band set to establish a non-uniform digital compensation vector.

4. The microwave component performance adaptive scheduling method according to claim 3, characterized in that, The specific steps for obtaining the synchronous loading compensation configuration file are as follows: S301: For the discrete data points arranged in time sequence in the non-uniform digital compensation vector, a predetermined number of new compensation values ​​are calculated between each pair of adjacent original data points according to their sequence index positions using a preset interpolation algorithm. All the calculated new compensation values ​​are then inserted between the original data points of the non-uniform digital compensation vector in index order to reconstruct and expand the data sequence and generate a compensation configuration file. S302: Obtain the next moment task instruction set, retrieve the timestamp field in the instruction set data structure, extract the future execution time information recorded in Coordinated Universal Time format, without performing any format conversion or time zone adjustment operation, and directly establish the extracted raw time data as the execution timestamp; S303: Call the compensation configuration file and the execution timestamp, take the execution timestamp as the current data element, append it to the end of the compensation configuration file data structure according to the preset binary encapsulation rules, establish a combined data structure, and generate a synchronously loaded compensation configuration file.

5. The microwave component performance adaptive scheduling method according to claim 4, characterized in that, The specific steps for obtaining the model correction parameter set are as follows: S401: Collect the actual output spectrum and operating parameters of the microwave component, obtain the ideal target spectrum, perform point-by-point subtraction operation on the power values ​​of the actual output spectrum and the ideal target spectrum at the same set of sampling frequency points, and perform structured pairing of the frequency point deviation value and the corresponding frequency point information to generate a spectrum deviation data sequence; S402: Call the frequency point deviation value and the frequency point information in the spectrum deviation data sequence, reconstruct multiple frequency point deviation values ​​into a two-dimensional data matrix based on the frequency point information, wherein the row index and column index of the two-dimensional data matrix correspond to the working time and frequency point respectively, and establish an error spectrum; S403: Call the working parameters and the established error map, quantize multiple parameter items of the working parameters into input vectors, quantize the matrix elements of the error map into target vectors, iteratively adjust the internal values ​​of the transformation matrix until the distance between the result vector calculated by the transformation matrix and the target vector is less than a preset convergence threshold, and construct a model correction parameter set.

6. The microwave component performance adaptive scheduling method according to claim 5, characterized in that, The multiphysics coupling solution is specifically as follows: The target frequency and the power are obtained, the electromagnetic field distribution is calculated based on Maxwell's equations, the ohmic loss caused by the power is solved, and the ohmic loss is used as a heat source input into the heat conduction equation. The predicted temperature is then obtained, and the heat conduction equation is solved using the predicted temperature as an initial condition to obtain the steady-state temperature field distribution of the microwave component. The conductivity and dielectric constant of the material in the digital twin model are then updated based on the steady-state temperature field distribution. Finally, based on the updated material conductivity and dielectric constant, the frequency-dependent transmission loss of the digital twin model is calculated. The calculation formula is as follows: ； in, Represents the target frequency and the predicted temperature Total transmission loss Represents the reference temperature The baseline transmission loss is as follows. The temperature loss coefficient represents the material-related properties. This represents the frequency loss coefficient associated with the geometry of the microwave component.

7. The microwave component performance adaptive scheduling method according to claim 6, characterized in that, The specific steps for constructing the model correction parameter set are as follows: Initialize the parameter association weight matrix, set the learning rate and the upper limit of the number of iterations, normalize each parameter item of the working parameters, and construct the input feature vector; For each iteration, the weight matrix associated with the input feature vector and the current parameters is obtained, the prediction error vector is calculated by matrix multiplication, and the error map is used as the true error vector. The gradient of the loss function is calculated based on the predicted error vector and the true error vector, and the parameter association weight matrix is ​​updated according to the following formula. : ； in, This represents the parameter association weight matrix after the (k+1)th iteration. This represents the parameter association weight matrix at the k-th iteration. The learning rate represents the step size used to control the update step. Represents the loss function The associated weight matrix of the currently described parameters The gradient; Repeat the update operation until the loss function is reached. If the value is lower than the preset convergence threshold or reaches the upper limit of the number of iterations, the parameter association weight matrix is ​​used as the core component of the model correction parameter set, and the model correction parameter set is established.

8. The microwave component performance adaptive scheduling method according to claim 7, characterized in that, The interpolation algorithm is a cubic spline interpolation algorithm, and its specific steps for smoothing the non-uniform digital compensation vector are as follows: The discrete data points in the non-uniform digital compensation vector are obtained and used as spline nodes. For each pair of adjacent spline nodes, a cubic polynomial is constructed. To ensure the overall smoothness of the curve, constraints are imposed on all connection points, requiring that at each spline node, the cubic polynomial function values ​​corresponding to two adjacent intervals are equal, the first derivative values ​​are equal, and the second derivative values ​​are equal. By combining all the aforementioned constraints, a system of linear equations is formed. Solving the system of linear equations yields the coefficients of each of the cubic polynomials. Based on the obtained coefficients, a preset number of interpolation points are calculated in each interval, and all interpolation points are merged with the original spline nodes to generate a compensation configuration file.

9. A microwave component performance adaptive scheduling system, characterized in that, The system is used to implement the microwave component performance adaptive scheduling method according to any one of claims 1-8, the system comprising: The component performance prediction module is used to obtain the target frequency and power of the task instruction set at the next moment, monitor the real-time temperature sequence of the microwave component and calculate the predicted temperature, and calculate the predicted component transmission characteristic curve using a digital twin model based on the target frequency, the power and the predicted temperature, and transmit the predicted component transmission characteristic curve to the distortion compensation calculation module. The distortion compensation solution module is used to analyze the transmission characteristic curve of the prediction component, calculate the loss change rate to determine the key distortion frequency point, divide the frequency sub-band with the key distortion frequency point and calculate the average prediction loss, invert the average prediction loss to generate a non-uniform digital compensation vector, and transmit the non-uniform digital compensation vector to the scheduling instruction synthesis module. The scheduling instruction synthesis module is used to smooth the discrete points of the non-uniform digital compensation vector transmitted by the distortion compensation solution module using an interpolation algorithm to synthesize a compensation configuration file, obtain the execution timestamp of the task instruction set and append it to the compensation configuration file, generate a synchronous loading compensation configuration file, and transmit the synchronous loading compensation configuration file to the twin model correction module. The twin model correction module is used to collect the actual output spectrum and operating parameters of the microwave component after loading the synchronous loading compensation configuration file, obtain the ideal target spectrum, calculate the error spectrum between the actual output spectrum and the ideal target spectrum, associate the operating parameters with the error spectrum to construct a model correction parameter set, and adjust the digital twin model in conjunction with the synchronous loading compensation configuration file.

Images