Power line carrier module sensitivity test system based on model fusion

Abstract

The invention relates to the technical field of model fusion, in particular to a power line carrier module sensitivity test system based on model fusion. The system comprises an analysis establishment unit, a transformation fusion unit and a test sensitivity unit. The sensitivity module takes the converted signal total power as the reference of sensitivity optimization, tests the loss function of the sensitivity of the power line carrier module by using the converted signal total power and the converted output, outputs a standard modulation waveform through a signal generator, and extracts a key reference signal; calculating a loss function error term of the power line carrier module through the loss function of the power line carrier module and the key reference signal, improving the test reliability by comprehensively considering the loss function and the converted signal total power, and discovering the potential problem of the sensitivity of the power line carrier module through the error term. And the reliability of the power line carrier module in a complex and changeable environment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of model fusion, and in particular to a power carrier module sensitivity testing system based on model fusion. Background Art

[0002] In real-world scenarios, there are many challenges, including complex interference environments, test equipment limitations, model coordination issues, changing environmental conditions, human factors, new modulation waveforms, miniaturized equipment testing, extreme working conditions, test efficiency, and data quality. Advanced signal processing technology, optimized model fusion algorithms, improved test equipment performance, developed dynamic adjustment functions, supported new waveforms, solved miniaturized equipment testing, and coped with extreme working conditions can ensure that the sensitivity test results of power carrier modules are true and reliable.

[0003] Since the power carrier module is the core device in the power communication system, its sensitivity is an important indicator to measure the module's performance. However, traditional sensitivity testing has the following problems in complex power environments:

[0004] Difficulty in multi-model coordination: Existing testing methods usually use a single model or a combination of simple models, which are difficult to cope with dynamic changes in the environment and complex interference. This causes the power carrier module to lose a low degree of quantitative sensitivity under different working conditions, reducing the reliability of the test.

[0005] Poor environmental adaptability: The power carrier module's environment may contain nonlinear interference, dynamic impedance changes, and multipath effects, which can affect the accuracy of test results. This can lead to delayed detection of potential sensitivity issues with the power carrier module, and a lack of support for improving the power carrier module's anti-interference capabilities and adaptability. This reduces the reliability of the power carrier module in complex and changing environments.

[0006] Therefore, we provide a power carrier module sensitivity testing system based on model fusion. Summary of the Invention

[0007] The purpose of the present invention is to provide a power carrier module sensitivity test system based on model fusion to solve the problems raised in the above background technology, including:

[0008] Because the power carrier module has difficulty coping with dynamic changes and complex interference in the environment, the power carrier module loses a lot of quantitative sensitivity under different working conditions, which reduces the reliability of the test. Therefore, this case comprehensively considers the loss function and the total power of the transformed signal to accurately quantify the sensitivity loss of the power carrier module under different working conditions and improve the reliability of the test.

[0009] To achieve the above object, the present invention provides a power carrier module sensitivity test system based on model fusion, comprising an analysis and establishment unit, a transformation and fusion unit, and a test sensitivity unit;

[0010] As a further improvement to the present technical solution, the sensitivity module uses the total power of the transformed signal as a benchmark for sensitivity optimization. The sensitivity loss function of the power carrier module is tested by the total power of the transformed signal and the transformed output. In actual operation, the sensitivity of the power carrier module will change due to factors such as noise, interference, and signal attenuation. By constructing a loss function based on the total power of the transformed signal and the transformed output, the impact of these complex factors on the sensitivity can be converted into a specific value, providing an objective and accurate measurement of the loss of the power carrier module's sensitivity.

[0011] A standard modulation waveform is output through a signal generator and a key reference signal is extracted. The error term of the power carrier module loss function is calculated using the power carrier module's loss function and the key reference signal. The key reference signal represents the performance of the power carrier module under ideal conditions. By comparing the loss function with the key reference signal and calculating the error term, the deviation between the actual performance and the ideal performance of the power carrier module can be quantified. For example, in signal transmission, the characteristics of the reference signal are known and ideal. The error term can intuitively reflect the degree to which the power carrier module deviates from the ideal state when processing the signal, thereby accurately evaluating the actual performance of the power carrier module.

[0012] As a further improvement of the present technical solution, the constraint update module formulates an error constraint gradient value by setting a loss function error term threshold, a gradient value of the original loss function, and a calculated loss function error term. As training progresses, the loss function error term will continue to change, and the error constraint gradient value can adaptively adjust the gradient update method and amplitude according to this change, so that the model can achieve optimal parameter adjustment at different training stages, thereby improving the performance and adaptability of the power carrier module.

[0013] The signal power constraint gradient value is formulated by setting the maximum signal power value and the total signal power of the transformation. During the signal transmission process, appropriate signal power is crucial to ensuring signal integrity and accuracy. Excessive power may cause signal distortion, while too little power may cause the signal to attenuate too quickly during transmission and cannot be effectively received. The signal power constraint gradient value can stabilize the total signal power of the transformation within a range that can both ensure signal strength and avoid distortion, thereby improving the signal transmission quality.

[0014] As a further improvement of the present technical solution, the sensitivity gain module obtains the signal that has not been optimized in the power carrier module, and then performs sensitivity gain of the power carrier module through the fusion of signal characteristics, transformation output, total power of the transformed signal and the signal that has not been optimized. The power carrier signal will be affected by various factors during the transmission process, such as noise interference, signal attenuation, and multipath effect. The signal characteristics contain multiple information such as frequency, phase, and amplitude of the signal. The output of the transformation reflects the result of the module's signal processing. The total power of the transformed signal reflects the energy situation of the signal, and the signal that has not been optimized is used as a reference benchmark. These factors are combined to calculate the sensitivity gain, which can comprehensively consider the characteristics of the signal in all aspects, thereby more accurately evaluating the actual performance of the power carrier module.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. In the power carrier module sensitivity test system based on model fusion, the sensitivity module uses the total power of the transformed signal as the benchmark for sensitivity optimization, and uses the total power of the transformed signal and the transformed output to test the loss function of the power carrier module sensitivity. The standard modulation waveform is output by the signal generator and the key reference signal is extracted. The error term of the power carrier module loss function is then calculated by the loss function of the power carrier module and the key reference signal. By comprehensively considering the factors of the loss function and the total power of the transformed signal, the degree of sensitivity loss of the power carrier module under different working conditions can be accurately quantified, thereby improving the reliability of the test. At the same time, potential problems with the sensitivity of the power carrier module can be discovered through the error term, providing support for improving the anti-interference ability and adaptability of the power carrier module, and improving the reliability of the power carrier module in complex and changing environments.

[0017] 2. In the power carrier module sensitivity test system based on model fusion, the sensitivity gain module obtains the signal that has not been optimized in the power carrier module through the calculation and analysis module, and then gains the sensitivity of the power carrier module through the fused signal characteristics, the output of the transformation, the total power of the transformed signal and the signal that has not been optimized. The power carrier module is gained by multiple factors, and the sensitivity of the power carrier module can be accurately tested in different signal environments, avoiding the influence of errors caused by testing the sensitivity with a single signal, reducing the influence of errors, and improving the accuracy of the sensitivity test of the power carrier module. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a block diagram of the overall system structure of the present invention;

[0019] Figure 2 It is a module block diagram of the present invention.

[0020] The meaning of each number in the figure is:

[0021] 10. Analysis and establishment unit; 110. Calculation and analysis module; 120. Model establishment module;

[0022] 20. Transformation and fusion unit; 210. Signal transformation module; 220. Weight distribution module; 230. Model fusion module;

[0023] 30. Test sensitivity unit; 310. Sensitivity module; 320. Constraint update module;

[0024] 330. Sensitivity gain module. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] Example 1

[0027] The present invention provides a power carrier module sensitivity test system based on model fusion, please refer to Figure 1-Figure 2 , including an analysis and establishment unit 10, a transformation and fusion unit 20 and a test sensitivity unit 30;

[0028] The analysis and establishment unit 10 includes a calculation and analysis module 110 and a model establishment module 120;

[0029] The calculation and analysis module 110 monitors the standard parameters of the power carrier module and the power line environment data in real time, obtains historical data, and calculates the power factor using the voltage and current values in the power line environment data. Then, the inductive load characteristics are analyzed through power factor analysis. The analyzed inductive load is used as load characteristic data. The line topology is identified using the monitored power carrier module standard parameters and power line environment data. Then, a signal generation model is established using protocol specifications, frequency range, load characteristics, power line length, topology, and historical test signal data. The formula for the signal generation model is:

[0030] F g =σ(W g s g +b g ); where s g is the input of the signal generation model, W g is the weight matrix, which linearly transforms the input vector, b g is the bias vector, which adds a constant offset to the result of the linear transformation, Fg is the output of the signal generation model, σ is the activation function, which performs nonlinear mapping on the result after linear transformation;

[0031] Power carrier standard parameters include protocol specification data (different power carrier communication protocols (such as IEC61334 and G3-PLC) have clear provisions for signal format, encoding method, and modulation type. For example, the G3-PLC protocol uses OFDM (orthogonal frequency division multiplexing) modulation, which specifies the number of subcarriers and subcarrier spacing parameters.) and frequency range data (power carrier communication operates within a specific frequency range, such as 3kHz-500kHz).

[0032] The power line environment data includes the voltage and current values on the power line;

[0033] Historical data includes historical test signal data (actual test signal data collected under different power conditions (such as peak power consumption and off-peak power consumption)), power line length data, power line physical characteristics data, environmental interference data, interference source characteristic data, mixed data of signals and interference, historical interference suppression effect data, and the gradient value of the original loss function.

[0034] Power line physical characteristic data includes cable parameters (cable material (such as copper, aluminum), cross-sectional area, insulation material) and line impedance data (the impedance of the power line changes with frequency, and the line impedance data at different frequencies is measured);

[0035] Environmental interference data includes noise characteristic data (various noises exist in the power line environment, such as Gaussian white noise and impulse noise) and electromagnetic interference data (the surrounding electromagnetic environment can also interfere with power line communications, such as radio signals and electromagnetic radiation from industrial equipment).

[0036] Interference source characteristic data includes interference type data (identifying interference types in the power line environment, such as harmonic interference, switching noise interference, and radio frequency interference) and interference intensity and frequency data (measuring the intensity and frequency distribution of interference sources);

[0037] The mixed data of signals and interference include mixed signal data from actual tests (sample data of mixed signals and interference collected in an actual power line environment) and mixed data under different signal-to-noise ratios (mixed data collected under different signal-to-noise ratios (the ratio of signal to noise power) to simulate interference conditions of varying degrees).

[0038] Historical interference suppression effect data refers to the effects of various interference suppression measures (such as filtering and encoding) used in previous tests;

[0039] Calculation process of power factor:

[0040] Use a power meter to measure the active power P of the circuit, calculate the apparent power S = U*I based on the voltage value U and the current value I, and divide the active power P by the apparent power S to get the power factor For example, if the measured voltage U = 220V and the current I = 5A, then the apparent power S = 220×5 = 1100 (V / A). If the active power P = 880W, then the power factor

[0041] Analyze the implementation process of inductive load characteristics:

[0042] The current of an inductive load lags behind the voltage, and the power factor is less than 1. Common inductive loads include motors and transformers. The presence and characteristics of inductive loads can be determined based on the phase difference between current and voltage and the power factor.

[0043] The model building module 120 builds a channel simulation model by acquiring the power line physical characteristic data and environmental interference data from the historical data of the calculation and analysis module 110. The formula of the channel simulation model is: c =σ(W c s c +b c ); where s c is the input of the channel simulation model, W c is the weight matrix, which linearly transforms the input vector, b c is the bias vector, which adds a constant offset to the result of the linear transformation, F c is the output of the channel simulation model, σ is the activation function, which performs nonlinear mapping on the result after linear transformation;

[0044] Then, an interference suppression model is established through the interference source characteristic data in the historical data, the mixed data of the signal and interference, and the historical interference suppression effect data. The formula of the interference suppression model is: F i =σ(W i s i +b i ); where s i is the input of the interference rejection model, W i is the weight matrix, which linearly transforms the input vector, b i is the bias vector, which adds a constant offset to the result of the linear transformation, F i is the output of the interference suppression model, σ is the activation function, which performs nonlinear mapping on the result after linear transformation;

[0045] The transformation and fusion unit 20 includes a signal transformation module 210 and a weight distribution module 220;

[0046] The signal conversion module 210 obtains the signal data of the power carrier module through the calculation and analysis module 110, obtains the obtained signal data x, and inputs the obtained signal data as input data into the signal generation model, the channel simulation model, and the interference suppression model respectively. The signal generation model, the channel simulation model, and the interference suppression model process the input data and output the signal F under the noise environment. g (x), signal F under channel fading environment c (x), signal F under interference i (x), and then integrate the signal in the noise environment, the signal in the channel fading environment, and the signal in the interference environment into the signal set {F g (x), F c (x), F i (x)}, the number of recorded signal samples N;

[0047] Output signal algorithm formula in noisy environment: Among them, F g (x) is a feature mapping function that maps the input signal to a new feature space, T represents the transposition operation, is the weight matrix W g The transpose of φ g It is a combination of a series of convolutional layers, pooling layers and activation functions to extract high-level features of the input signal. g is the bias vector, which adds a constant offset to the result of the linear transformation;

[0048] Output signal algorithm formula in noisy channel fading environment: F c (x) is a feature mapping function that maps the input signal to a new feature space, T represents the transposition operation, is the weight matrix W c The transpose of φ c It is a combination of a series of convolutional layers, pooling layers and activation functions to extract high-level features of the input signal. c is the bias vector, which adds a constant offset to the result of the linear transformation;

[0049] Signal algorithm formula under output interference: F i (x) is a feature mapping function that maps the input signal to a new feature space, T represents the transposition operation, is the weight matrix W i The transpose of φ i It is a combination of a series of convolutional layers, pooling layers and activation functions to extract high-level features of the input signal. i is the bias vector, which adds a constant offset to the result of the linear transformation;

[0050] Since the acquired signal data includes the clean signal x clean (the original signal without noise interference) and the noise signal x noise (refers to the interference component superimposed on the clean signal), the acquired signal data is input into the neural network model for processing and the transformation function ψ is output k (x)=σ(W k x+b k ); where ψ k (x) is the kth signal of the output transformation, ψ k (·) is the transformation function used to transform the input signal, k refers to the kth signal, σ is the activation function, W k is the weight matrix of the kth signal, b k is the bias vector of k signals, and then the clean signal is transformed through the transformation function ψ k (·) After the transformation, the transformed clean signal is squared to obtain the transformed clean signal power

[0051] The transformed clean signal power is similar to the transformed noise signal power. The noise signal is transformed through the transformation function ψ k After the (·) transformation, the transformed noise signal is squared to obtain the transformed noise signal power

[0052] The clean signal power through the transformation and the transformed noise signal power Evaluate the signal-to-noise ratio (SNR) of the kth signal k , its specific algorithm formula is: Among them, le-6 is a very small constant (i.e. 1×10 -6 ), its role is to avoid the situation where the denominator is zero, the signal-to-noise ratio SNR of the kth signal k The unit is decibel (dB);

[0053] The weight distribution module 220 uses the signal conversion module 210 to set the signal {F g (x), F c (x), F i Calculate the signal-to-noise ratio (SNR) of the jth signal in (x)} j , according to the signal-to-noise ratio SNR of the jth signal j Calculate the average signal-to-noise ratio using the number of signal samples N Then according to the signal-to-noise ratio SNR of the jth signal j , the number of signal samples N and the average signal-to-noise ratio μ to calculate the standard deviation of the signal-to-noise ratio std(SNR j ), its specific algorithm formula is:

[0054] According to the standard deviation of the signal-to-noise ratio (SNR j ) Calculate the dynamic adjustment temperature parameter η temp =α·std(SNR j )+β, where the temperature parameter η is dynamically adjusted temp It is an adjustable constant that controls the "smoothness" or "sharpness" of the softmax function. α refers to the scaling factor that controls the influence of the standard deviation on the dynamic adjustment of the temperature parameter. β refers to the offset that ensures the baseline value of the dynamic adjustment of the temperature parameter.

[0055] Use the softmax function to calculate the signal-to-noise ratio (SNR) of the kth signal k , the signal-to-noise ratio (SNR) of the jth signal j and dynamically adjust the temperature parameter η temp Perform soft weight assignment to obtain the weight w for assigning the kth signal k , its specific algorithm formula is:

[0056] Among them, exp(·) is the exponential function, that is, e x , e≈2.71828 is a natural constant, w k The value range is [0,1]. When the temperature parameter η is adjusted dynamically temp When it is larger, the difference between the exponential terms will be reduced, the output of the softmax function will be smoother, and the individual w k The values of will be closer, which means that the difference between different signal-to-noise ratios has relatively little effect on the weight;

[0057] When the temperature parameter η is dynamically adjusted temp When it is smaller, the difference between the exponential terms will be amplified, the output of the softmax function will be sharper, and the weight w corresponding to the element with a higher signal-to-noise ratio will be k will increase significantly, while the weights of other elements will decrease accordingly, that is, the importance of elements with high signal-to-noise ratio will be emphasized more;

[0058] exp(η temp SNR k ) represents the kth signal-to-noise ratio SNR k Scaling (multiplying by the dynamic adjustment temperature parameter η temp ) and then take the index; the larger the value, the higher the signal-to-noise ratio of the k-th signal is, and it will make a greater contribution in the weight calculation;

[0059] The signal conversion module 210 receives the command of the weight distribution module 220 to distribute the signal weight, and converts the signal F g(x), signal F under channel fading environment c (x), signal F under interference i Input transformation function ψ k (·) and after transformation, the corresponding transformed noise environment signal ψ is output g (x), transformed channel fading signal ψ c (x), the transformed interference signal ψ i (x) The signal power is transformed by the transformed noise environment signal, the transformed channel fading signal, and the transformed interference situation signal to obtain the transformed noise environment signal, the channel fading signal, and the interference situation signal power to evaluate the signal-to-noise ratio of the noise environment signal, the channel fading signal, and the interference situation signal. The transformed noise environment signal, the channel fading signal, the interference situation signal power and the transformed clean signal power are weighted and summed to obtain the transformed signal total power.

[0060] The weight allocation module 220 uses the signal-to-noise ratios of the noise environment signal, channel fading signal, and interference signal in the signal conversion module 210 to calculate the weights of the noise environment signal, channel fading signal, and interference signal, and obtains the allocated noise environment signal weight w g , the assigned channel fading signal weight w c , the assigned noise interference signal weight w i ;

[0061] The transformation fusion unit 20 further includes a model fusion module 230;

[0062] The model fusion module 230 transforms the noise environment signal ψ g (x), transformed channel fading signal ψ c (x), the transformed interference signal ψ i (x), the transformed k-th signal ψ k (x) and the noise environment signal weight w assigned in the weight assignment module 220 g , the assigned channel fading signal weight w c , the assigned noise interference signal weight w i , assign the weight w to the kth signal k Perform multi-model signal feature fusion to obtain the fused signal feature F. The specific algorithm formula is:

[0063] F=∑(ψ g w g +ψ c w c +ψ i w i +ψ k w k)(x); Among them, this formula combines the advantages of different models and is usually more robust than the characteristics of a single model;

[0064] By performing a linear transformation on the fused signal features, the transformed output y is obtained. The specific algorithm formula is:

[0065] Among them, W o is the weight matrix used to map the fused signal features to the target space, b o is the bias term, used to adjust the output offset;

[0066] The test sensitivity unit 30 includes a sensitivity module 310 and a constraint update module 320;

[0067] The sensitivity module 310 receives the total signal power transformed in the signal transformation module 210 and the output transformed in the model fusion module 230, takes the transformed total signal power as the benchmark for sensitivity optimization, and tests the loss function of the sensitivity of the power carrier module through the transformed total signal power and the transformed output, and obtains the sensitivity loss function L of the test. sens , its specific algorithm formula is:

[0068] This formula is specifically used to improve the sensitivity and test consistency of power carrier communication modules. It directly maximizes the ratio of signal power to signal-to-noise ratio. The negative sign indicates gradient ascent optimization, forcing the model to prioritize improving weak signal recognition capabilities.

[0069] Output the standard modulation waveform through the signal generator and extract the key reference signal y from the standard modulation waveform ref , through the loss function L of the power carrier module sens With the key reference signal y ref Calculate the error term of the power carrier module loss function and obtain the calculated loss function error term L err , its specific algorithm formula is:

[0070] The constraint update module 320 extracts the gradient value of the original loss function from the historical data in the calculation and analysis module 110 By setting the error term threshold ∈ of the loss function and the gradient value of the original loss function and the loss function error term L calculated in the sensitivity module 310 err Formulate error constraint gradient value The specific algorithm formula is: Where I(·) is the indicator function, which is defined as: Used to control the reverse transmission strength of the error term of the power carrier module loss function;

[0071] By setting the maximum signal power value P max The total power of the signal converted in the signal conversion module 210 Formulate signal power constraint gradient value The specific algorithm formula is: Here, ∞ refers to infinity and I(·) is the indicator function, which is defined as:

[0072] The sensitivity loss function L tested in the sensitivity module 310 sens , calculate the loss function error term L err and error constraint gradient value Sum signal power constraint gradient value Input the neural network model, the neural network model processes the input data and outputs the corresponding weight coefficients γ1, γ2, γ3, γ4, using the test sensitivity loss function L sens , calculate the loss function error term L err , error constraint gradient value Sum signal power constraint gradient value Multiply weighted by weight coefficients γ1, γ2, γ3, and γ4

[0073] Weight the product Input into the deep learning model for learning, and output the model parameters θ and learning rate τ, weighted by product The parameters are updated in real time with the model parameters θ and learning rate τ. The update formula is:

[0074]

[0075] The test sensitivity unit 30 further includes a sensitivity gain module 330;

[0076] The sensitivity gain module 330 receives the parameter update command from the constraint update module 320, and the sensitivity gain module 330 obtains the signal x that is not optimized in the power carrier module through the calculation and analysis module 110. base By combining the signal features F fused in the model fusion module 230, the transformed output y, and the total power of the signal transformed in the signal transformation module 210 With the unoptimized signal x base Normalize and perform the gain of the power carrier module sensitivity to obtain the gained power carrier module sensitivity ΔS. The specific algorithm formula is: This formula can more accurately evaluate the sensitivity of the power carrier module in these key frequency bands, avoiding the information ambiguity caused by full-band analysis.

[0077] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. The power carrier module sensitivity test system based on model fusion is characterized by: It includes an analysis and establishment unit (10), a transformation and fusion unit (20) and a test sensitivity unit (30); The analysis and establishment unit (10) acquires historical data and establishes a signal generation model, a channel simulation model and an interference suppression model; The transformation and fusion unit (20) obtains the signal data of the power carrier module through the analysis and establishment unit (10) and inputs the data into the signal generation model, the channel simulation model, and the interference suppression model for processing, outputs the signal in the noise environment, the signal in the channel fading environment, and the signal in the interference situation, performs signal power transformation, obtains the transformed noise environment signal, the channel fading signal, and the interference situation signal power to evaluate the signal-to-noise ratio of the noise environment signal, the channel fading signal, and the interference situation signal, then performs weighted summation to obtain the transformed signal total power, then performs multi-model signal feature fusion on the transformed noise environment signal, the transformed channel fading signal, and the transformed interference situation signal, and then performs linear transformation on the fused signal feature to obtain the transformed output; The test sensitivity unit (30) performs sensitivity gain of the power carrier module through the signal characteristics fused in the transformation and fusion unit (20), the transformed output, and the transformed signal total power.

2. The power carrier module sensitivity test system based on model fusion according to claim 1 is characterized in that: The analysis and establishment unit (10) includes a calculation and analysis module (110) and a model establishment module (120); The calculation and analysis module (110) monitors the standard parameters of the power carrier module and the power line environment data, then obtains historical data, calculates the power factor using the power line environment data, and then analyzes the inductive load characteristics through the power factor, uses the analyzed inductive load as load characteristic data, and establishes a signal generation model.

3. The power carrier module sensitivity test system based on model fusion according to claim 2 is characterized in that: The model building module (120) builds a channel simulation model by acquiring the power line physical characteristics and environmental interference data in the historical data of the calculation and analysis module (110), and then builds an interference suppression model based on the interference source characteristics, the mixture of signals and interference, and the historical interference suppression effect data in the historical data.

4. The power carrier module sensitivity test system based on model fusion according to claim 3 is characterized in that: The transformation and fusion unit (20) includes a signal transformation module (210) and a weight distribution module (220); The signal conversion module (210) obtains signal data of the power carrier module through the calculation and analysis module (110), inputs the data into the signal generation model, the channel simulation model, and the interference suppression model, processes the data, and outputs a signal in a noise environment, a signal in a channel fading environment, and a signal in an interference environment to integrate a signal set; Since the acquired signal data includes a clean signal and a noise signal, the clean signal and the noise signal are transformed by a transformation function to obtain the transformed clean signal power and the transformed noise signal power, and the signal-to-noise ratio of the kth signal is evaluated based on the transformed clean signal power and the transformed noise signal power.

5. The power carrier module sensitivity testing system based on model fusion according to claim 4 is characterized in that: The weight distribution module (220) calculates the signal-to-noise ratio of the jth signal in the signal set of the signal conversion module (210), calculates the average signal-to-noise ratio and the standard deviation of the signal-to-noise ratio based on the signal-to-noise ratio of the jth signal, and then calculates the dynamic adjustment temperature parameter based on the standard deviation of the signal-to-noise ratio; The softmax function is used to perform soft weight allocation according to the signal-to-noise ratio of the k-th signal, the signal-to-noise ratio of the j-th signal and the dynamically adjusted temperature parameter to obtain the weight allocated to the k-th signal.

6. The power carrier module sensitivity testing system based on model fusion according to claim 5 is characterized in that: The signal conversion module (210) converts the signal in the noise environment, the signal in the channel fading environment, and the signal in the interference environment, and outputs the converted noise environment signal, the converted channel fading signal, and the converted interference situation signal to perform signal power conversion, obtain the converted noise environment signal, the channel fading signal, and the interference situation signal power to evaluate the signal-to-noise ratio of the noise environment signal, the channel fading signal, and the interference situation signal, and then perform weighted summation to obtain the converted signal total power; The weight distribution module (220) uses the signal-to-noise ratios of the noise environment signal, the channel fading signal, and the interference situation signal in the signal conversion module (210) to calculate the weights for distributing the noise environment signal, the channel fading signal, and the interference situation signal.

7. The power carrier module sensitivity testing system based on model fusion according to claim 6 is characterized in that: The transformation fusion unit (20) further includes a model fusion module (230); The model fusion module (230) performs multi-model signal feature fusion on the noise environment signal, the channel fading signal, and the interference situation signal transformed in the signal transformation module (210) with the noise environment signal weight, the channel fading signal weight, and the noise interference situation signal weight allocated in the weight allocation module (220) to obtain a fused signal feature, and then performs a linear transformation on the fused signal feature to obtain a transformed output.

8. The power carrier module sensitivity testing system based on model fusion according to claim 7 is characterized in that: The test sensitivity unit (30) includes a sensitivity module (310) and a constraint update module (320); The sensitivity module (310) uses the total power of the signal transformed in the signal transformation module (210) as a benchmark for sensitivity optimization, and tests the loss function of the sensitivity of the power carrier module through the total power of the transformed signal and the output transformed in the model fusion module (230); The signal generator is used to output the standard modulation waveform and extract the key reference signal. Then, the error term of the power carrier module loss function is calculated by the loss function of the power carrier module and the key reference signal.

9. The power carrier module sensitivity testing system based on model fusion according to claim 8 is characterized in that: The constraint updating module (320) formulates an error constraint gradient value by setting a loss function error term threshold, a gradient value of the original loss function in the calculation and analysis module (110), and a loss function error term calculated in the sensitivity module (310); Formulating a signal power constraint gradient value by using a set maximum signal power value and the total signal power converted in the signal conversion module (210); The sensitivity loss function tested in the sensitivity module (310), the calculated loss function error term, the error constraint gradient value, and the signal power constraint gradient value are input into a neural network model and multiplied and weighted, the multiplied and weighted product is input into a deep learning model for learning, and the model parameters and learning rate are output, and the parameters are updated in real time through the multiplication and weighting of the product, the model parameters, and the learning rate.

10. The power carrier module sensitivity testing system based on model fusion according to claim 9 is characterized in that: The test sensitivity unit (30) further includes a sensitivity gain module (330); When the parameter update is known, the sensitivity gain module (330) obtains the signal that has not been optimized in the power carrier module through the calculation and analysis module (110), and performs sensitivity gain of the power carrier module through the signal characteristics fused in the model fusion module (230), the output of the transformation, the total power of the signal transformed in the signal transformation module (210), and the signal that has not been optimized.

Images