Method for predicting transmission power of indoor antenna system based on deep learning

Abstract

This application relates to the field of deep learning technology, and particularly to a method for predicting the transmission power of an indoor antenna system based on deep learning, including establishing communication between a digital beam controller and antennas in a target venue and controlling the overall transmission power of antenna units; obtaining a training data set containing multiple groups of training data, where the training data includes passenger flow density and its corresponding real-time overall transmission power, and calculating the final optimal overall transmission power corresponding to the passenger flow density in each group of training data based on the training data and a preset simulation model; training a preset neural network model to obtain a power prediction model; obtaining the passenger flow density in the target venue and inputting it into the power prediction model to output the optimal overall transmission power, and the digital beam controller generating corresponding digital control signals according to the optimal overall transmission power and sending them to all antenna units. This application can achieve efficient management of the indoor antenna system.

Description

Technical Field

[0001] This application relates to the field of deep learning technology, and particularly to a method for predicting the transmission power of an indoor antenna system based on deep learning. Background Art

[0002] In modern wireless communication systems, indoor antenna systems are widely used in large venues such as shopping malls, underground parking lots, hospitals, etc. to improve signal coverage quality and meet the network needs of users. The passenger flow density in the above-mentioned venues usually has obvious characteristics of changing over time periods. For example, during the peak daytime hours, the flow of people is dense and the network load is high, while at night when the flow of people is scarce, the network demand drops significantly. This tidal change in network demand poses a severe challenge to the management of existing indoor antenna systems.

[0003] Traditional indoor antenna systems usually adopt a design with a fixed power output. The power output of the indoor antenna system is based on pre-set parameters and cannot be dynamically adjusted according to real-time network demands. As a result, during peak passenger flow hours, the indoor antenna system may not be able to effectively handle the high-load network demands, while during low-load hours, the high power output of the indoor antenna system may cause resource waste and increased energy consumption. Therefore, when dealing with changing network demands and complex environments, the indoor antenna system lacks sufficient flexibility and is difficult to achieve the best resource utilization and signal coverage under changing conditions, ultimately unable to achieve efficient management of the indoor antenna system. Summary of the Invention

[0004] This application provides a method for predicting the transmission power of an indoor antenna system based on deep learning, which can solve the problem that the fixed power output of the existing antenna system cannot adapt to the tidal change of network demands. This application provides the following technical solutions:

[0005] In a first aspect, this application provides a method for predicting the transmission power of an indoor antenna system based on deep learning, and the method includes:

[0006] Establish communication between a digital beam controller and all antenna units in the target venue, and the digital beam controller controls the overall transmission power of all antenna units through digital signals;

[0007] Obtain a training data set containing multiple groups of training data, where the training data includes passenger flow density and its corresponding real-time overall transmission power, and based on the training data and a pre-set simulation model, calculate the final optimal overall transmission power corresponding to the passenger flow density in each group of training data;

[0008] Use the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output, and train a pre-set neural network model to obtain a power prediction model;

[0009] Obtain the passenger flow density in the target venue in real time and input it into the power prediction model. The power prediction model outputs the optimal overall transmission power, and the digital beam controller generates corresponding digital control signals according to the optimal overall transmission power output by the power prediction model and sends them to all antenna units.

[0010] In a specific implementable embodiment, establishing the communication between the digital beam controller and all antenna units in the target venue includes:

[0011] Deploy a digital beam controller and several antenna units in the target venue, and the several antenna units are distributed in different areas of the target venue;

[0012] Establish wireless communication between the digital beam controller and the antenna units using the same communication protocol;

[0013] The digital beam controller generates corresponding digital control signals according to the overall transmission power input by the user, combines the distribution characteristics of the antenna units, and sends them to each antenna unit through wireless communication.

[0014] In a specific implementable embodiment, the formula for the transmission power allocation of each antenna unit by the digital beam controller according to the overall transmission power input by the user is as follows:

[0015]

[0016] where P i is the transmission power of the i-th antenna unit, P total is the overall transmission power input by the user, G i is the gain of the i-th antenna unit, d ci is the distance between the digital beam controller and the i-th antenna unit, α is the path loss factor, N is the number of antenna units, and j represents the summation index.

[0017] In a specific implementable embodiment, introduce a calibration coefficient to adjust the calculated transmission power of each antenna unit as follows:

[0018]

[0019] where is the adjusted transmission power of the i-th antenna unit, C cal is the calibration coefficient; the calculation method of the calibration coefficient C cal is as follows:

[0020]

[0021] where It is the total actual transmission power of all antenna units.

[0022] In a specific feasible implementation, the final optimal overall transmission power corresponding to the passenger flow density in each group of training data deduced based on the training data and a preset simulation model includes:

[0023] Each group of training data includes the passenger flow density of a certain venue, the real-time overall transmission power of the antenna corresponding to this passenger flow density, the basic power of this venue, and the maximum passenger flow density set for this venue;

[0024] Introduce a simulation model to deduce the optimal overall transmission power corresponding to the passenger flow density in each group of training data. The simulation model is as follows:

[0025]

[0026] Among them, P best is the optimal overall transmission power, P base is the basic power in this group of training data, D is the passenger flow density in this group of training data, D max is the maximum passenger flow density in this group of training data, and γ is a non-linear adjustment factor.

[0027] In a specific feasible implementation, the final optimal overall transmission power corresponding to the passenger flow density in each group of training data deduced based on the training data and a preset simulation model further includes:

[0028] Calculate the relative error between the optimal overall transmission power corresponding to the passenger flow density in each group of training data and the real-time overall transmission power as follows:

[0029]

[0030] Among them, ΔP is the relative error, and P real is the real-time overall transmission power in this group of training data;

[0031] Compare the relative error with a preset threshold. If the relative error is less than or equal to the preset threshold, it is determined that the real-time overall transmission power is the final optimal overall transmission power corresponding to the passenger flow density in this group of training data;

[0032] If the relative error is greater than the preset threshold, it is determined that the optimal overall transmission power deduced by using the simulation model is the final optimal overall transmission power corresponding to the passenger flow density in this group of training data.

[0033] In a specific feasible implementation, before training a preset neural network model with the passenger flow density of each group of training data in the training dataset as the input and the corresponding final optimal overall transmission power as the output, the following steps are also included:

[0034] Amplify the training dataset. Assume that the training dataset contains the passenger flow density D of each training data n and the corresponding final optimal overall transmission power P best,n . It is desired to generate the amplified training data D n,new and P best,n,new . The amplification formula is as follows:

[0035] D n,new = D n ×(1 + ∈);

[0036]

[0037] where ∈ is a perturbation term, ∈ ∈ [-0.1, 0.1], and γ is a non-linear adjustment factor;

[0038] Use the passenger flow density of each group of training data in the amplified training dataset as the input and the corresponding final optimal overall transmission power as the output to train a preset neural network model to obtain a power prediction model.

[0039] In a second aspect, the present application provides an indoor antenna system transmission power prediction system based on deep learning, adopting the following technical solution:

[0040] An indoor antenna system transmission power prediction system based on deep learning includes:

[0041] A communication establishment module for establishing communication between a digital beam controller and all antenna units in a target venue, where the digital beam controller controls the overall transmission power of all antenna units through digital signals;

[0042] A data acquisition module for acquiring a training dataset containing multiple groups of training data. The training data includes passenger flow density and its corresponding real-time overall transmission power, and based on the training data and a preset simulation model, calculates the corresponding final optimal overall transmission power for the passenger flow density in each group of training data;

[0043] A model training module for training a preset neural network model with the passenger flow density of each group of training data in the training dataset as the input and the corresponding final optimal overall transmission power as the output to obtain a power prediction model;

[0044] A power control module is used to obtain the passenger flow density in the target venue in real time and input it into the power prediction model. The power prediction model outputs the optimal overall transmission power. The digital beam controller generates corresponding digital control signals according to the optimal overall transmission power output by the power prediction model and sends them to all antenna units.

[0045] In a third aspect, the present application provides an electronic device, which includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement a method for predicting the transmission power of an indoor antenna system based on deep learning as described in the first aspect.

[0046] In a fourth aspect, the present application provides a computer-readable storage medium, in which a program is stored, and when the program is executed by a processor, it is used to implement a method for predicting the transmission power of an indoor antenna system based on deep learning as described in the first aspect.

[0047] In summary, the beneficial effects of the present application at least include:

[0048] 1) By designing the power distribution formula between the digital beam controller and the antenna unit, the transmission power distribution in different scenarios can not only meet the signal coverage requirements but also avoid power waste. The formula not only considers the local characteristics such as the gain and distance of each antenna unit but also combines the limitation of the overall system transmission power to ensure the reasonable distribution of power. This method can dynamically adapt to environmental and scenario changes, optimize power distribution, and further eliminate the small errors caused by environmental factors or configuration imbalance by introducing a calibration coefficient, ensuring that the overall transmission power is close to the target value input by the user, thereby providing stable and efficient communication coverage.

[0049] 2) The adopted simulation model and neural network model enable the system to calculate the optimal overall transmission power based on real-time passenger flow density data. By introducing a non-linear adjustment factor and a path loss factor into the training data, the complex non-linear relationship between passenger flow density and power demand can be accurately reflected. At the same time, an amplification algorithm is introduced to expand the training data through perturbation terms, further improving the robustness and accuracy of the prediction model. This model can be flexibly adjusted in different scenarios to accurately predict the best transmission power, ensuring the adaptive ability and accuracy of the system.

[0050] 3) To improve the generalization ability and robustness of the power prediction model, this application adopts a data augmentation method. Based on the perturbation term and the non-linear adjustment factor, more training data that conforms to the actual application scenario is generated. This augmentation method not only takes into account the non-linear relationship between passenger flow density and transmission power, but also can simulate the random fluctuations in the actual scenario, such as seasonal changes, emergencies, etc. In this way, the augmented training data set can be closer to the real scenario, improving the accuracy of the model, avoiding overfitting, and enhancing the adaptability of the system in different environments.

[0051] By establishing communication between the digital beam controller and all antenna units in the target venue, the problem that the traditional indoor antenna system cannot adjust the power in real time according to the passenger flow density is first solved. Then, by collecting multiple sets of training data on passenger flow density and the corresponding transmission power, and combining with the simulation model to calculate the optimal overall transmission power corresponding to each data point, it is ensured that the power output can be dynamically adjusted according to the change in the passenger flow. Next, these training data are input into a preset neural network model for training to obtain a power prediction model. This model can predict the corresponding optimal overall transmission power based on the passenger flow density data obtained in real time in the target venue. Finally, the digital beam controller generates corresponding digital control signals according to the prediction result and sends them to all antenna units to achieve precise adjustment of the overall transmission power of the antenna units. Through this process, this application effectively solves the problem that the fixed power output of the existing antenna system cannot adapt to the changing demands of the tidal network, and can adjust the power according to the actual needs at different time periods, thereby improving the utilization efficiency of network resources, avoiding waste, meeting the dynamically changing network demands, and achieving efficient management of the indoor antenna system.

[0052] The above description is only an overview of the technical solution of this application. In order to be able to more clearly understand the technical means of this application and implement it according to the content of the specification, the following takes the preferred embodiments of this application and cooperates with the attached drawings to elaborate in detail as follows. Description of the Drawings

[0053] Figure 1 It is a schematic flowchart of the method for predicting the transmission power of an indoor antenna system based on deep learning in an embodiment of this application.

[0054] Figure 2 It is a schematic overall flowchart of the method for predicting the transmission power of an indoor antenna system based on deep learning in an embodiment of this application.

[0055] Figure 3 It is a structural block diagram of the system for predicting the transmission power of an indoor antenna system based on deep learning in an embodiment of this application.

[0056] Figure 4It is a block diagram of an electronic device for predicting the transmission power of an indoor antenna system based on deep learning in an embodiment of the present application. Detailed implementation manners

[0057] The following further describes in detail the specific implementation manners of the present application in conjunction with the accompanying drawings and embodiments. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application.

[0058] Optionally, the present application takes the method for predicting the transmission power of an indoor antenna system based on deep learning provided in each embodiment and applied to an electronic device as an example for illustration. The electronic device is a terminal or a server. The terminal can be a mobile phone, a computer, a tablet computer, etc. The type of the electronic device is not limited in this embodiment.

[0059] Referring to Figure 1 , it is a schematic flowchart of a method for predicting the transmission power of an indoor antenna system based on deep learning provided in an embodiment of the present application. The method at least includes the following steps:

[0060] Step S101: Establish communication between the digital beam controller and all antenna units in the target venue. The digital beam controller controls the overall transmission power of all antenna units through digital signals.

[0061] In step S101, first, a digital beam controller and several antenna units are deployed in the target venue, where the several antenna units are distributed in different areas of the target venue to ensure comprehensive signal coverage. As the core control device, the digital beam controller can generate digital control signals and send instructions to all antenna units. Specifically, a wireless communication is established between the digital beam controller and the antenna units using a unified communication protocol, such as Wi-Fi, ZigBee, or other low-latency wireless communication protocols suitable for indoor antenna systems.

[0062] In implementation, after the communication is established, the digital beam controller generates corresponding digital control signals according to the overall transmission power input by the user and in combination with the distribution characteristics of the antenna units, and sends them to each antenna unit through wireless communication. After receiving the control signal, each antenna unit adjusts its transmission power according to the instruction so that the overall transmission power of all antenna units is basically equal to the overall transmission power input by the user.

[0063] Among them, the formula for the digital beam controller to allocate the transmission power of each antenna unit according to the overall transmission power input by the user is as follows:

[0064]

[0065] Among them, P i is the transmission power of the i-th antenna unit, P totalis the overall transmission power input by the user, G i is the gain of the i-th antenna element, d ci is the distance between the digital beam controller and the i-th antenna element. α is the path loss factor, which is used to characterize the attenuation degree of the signal in space and is a constant. N is the number of antenna elements, and j represents the summation index.

[0066] In the above formula, an important advantage of the formula design is that when allocating the power of each antenna element, it not only considers the local characteristics of each antenna element (such as gain and distance), but also considers the limitation of the overall transmission power of the entire system. By balancing the overall transmission power and local requirements, it ensures that the system can maintain reasonable coverage and energy distribution in different regions and environments, without causing the power in some regions to be too low to affect the signal quality or too high to waste energy. It not only achieves precise control of power, but also can dynamically adapt to the demand changes in different scenarios. By introducing the distribution relationship between the overall power and each antenna element, the formula can ensure the reasonable allocation of the system resources in space, maximize the satisfaction of the coverage requirements, and avoid resource waste.

[0067] In the prior art, although there are some methods to allocate antenna power through weight factors, most of them are only based on simple weighted averages and ignore the interactive effects of multiple complex factors. For example, many existing power allocation algorithms simply perform power allocation through linear weighting of distance and gain, and fail to fully consider the non-linear attenuation characteristics of the signal in space propagation and the complex changes that may occur in the environment. Therefore, there may be certain errors in the effect of this method in practical applications and it cannot flexibly adapt to the requirements of different scenarios. However, the formula designed above combines the path loss factor and gain, making the power allocation not only depend on the distance between antennas, but comprehensively consider the performance characteristics of each antenna element and the actual propagation environment. This way can better adapt to the changes in the environment through accurate mathematical modeling and avoid the over-simplification problems existing in the prior art.

[0068] In addition, preferably, after calculating the transmission power of each antenna element according to the above formula, the overall transmission power should be close to the overall transmission power input by the user, but not necessarily exactly equal. Because the method adopted in this application is a distribution method, there may be a slight difference in the overall power due to site limitations, environmental factors or uneven antenna element configurations. Therefore, in order to ensure that the overall transmission power is as close as possible to the power input by the user and make up for the slight difference caused by site limitations, environmental factors or uneven antenna element configurations, a calibration coefficient is introduced to adjust the transmission power of each antenna element calculated to achieve precise matching of the overall transmission power.

[0069] Specifically, a calibration coefficient is set as a constant to adjust the ratio of the transmission powers of all antenna elements. Its purpose is to perform proportional correction based on the gap between the sum of the actual transmission powers and the target transmission power input by the user. The introduction of the calibration coefficient adjusts the calculated transmission power of each antenna element as follows:

[0070]

[0071] Among them, is the transmission power of the i-th antenna element after adjustment, and C cal is the calibration coefficient. The calculation method of the calibration coefficient C cal is as follows:

[0072]

[0073] Among them, is the sum of the actual transmission powers of all antenna elements.

[0074] The calibration coefficient ensures that even if there are slight power differences caused by environmental factors or uneven configurations, the overall transmission power can ultimately match the target transmission power input by the user. Through this method, the overall transmission power of the system can be precisely adjusted, and errors caused by uneven configurations or external environmental factors can be avoided as much as possible, ultimately making the system output stable and meeting expectations. In addition, the calculation of the calibration coefficient in this application is based on the overall transmission power distribution, and a unified digital control signal is generated through an optimized formula for adjustment, rather than responding to signal changes in real time for each antenna. Therefore, the high computational burden caused by frequent adjustments is avoided. In addition, the number of antennas is limited in the typical application scenario of this application, which further reduces the complexity and real-time requirements of the system.

[0075] Step S102: Obtain a training data set containing multiple groups of training data. The training data includes the passenger flow density and its corresponding real-time overall transmission power, and based on the training data and a preset simulation model, calculate the final optimal overall transmission power corresponding to the passenger flow density in each group of training data.

[0076] In step S102, first collect the training data set from a public database. The training data set includes multiple groups of training data, where each group of training data includes the passenger flow density of a certain venue, the real-time overall transmission power of the antenna corresponding to the passenger flow density, the basic power of the venue, and the maximum passenger flow density set for the venue. The basic power of the venue refers to the lowest power level of the system when there is no influence from the passenger flow density.

[0077] It should be noted that the acquisition channel of the training dataset in this application is an open database. Specifically, it is a public database formed by combining the passenger flow statistics data of certain areas that may be made public by some websites during the construction of a smart city with the relevant power data of the already public wireless communication base stations.

[0078] Subsequently, based on the training dataset and the preset simulation model, the optimal overall transmission power corresponding to each group of passenger flow densities is calculated. Specifically, since the overall transmission power of the antenna corresponding to the passenger flow density of a certain place in each group of training data only represents the overall transmission power of the antenna applied in the actual implementation process and does not represent the most optimal overall transmission power of the antenna, it is necessary to introduce a simulation model to calculate the optimal overall transmission power corresponding to the passenger flow density in each group of training data. The simulation model is as follows:

[0079]

[0080] Among them, P best is the optimal overall transmission power, P base is the base power in this group of training data, D is the passenger flow density in this group of training data, D max is the maximum passenger flow density in this group of training data, and γ is a non-linear adjustment factor used to control the non-linear growth degree of the influence of passenger flow density on the transmission power. In this application, it is a fixed value, set to 1.2. It should be noted that the setting of its value is obtained through a full analysis of typical scenarios and experimental verification, and has wide applicability. Although the non-linear characteristics of different places may vary, the goal of this application is to provide a general, stable and easy-to-implement technical solution, rather than being tailored for each scenario. The setting of a fixed value can achieve balanced performance in most scenarios and meet the actual needs.

[0081] In the design process of the above simulation model, by introducing a non-linear adjustment factor to describe the influence of passenger flow density on the transmission power in the form of a power, it has higher flexibility and adaptability. The core advantage of this design is to avoid the possible limitation of the conventional linear weighting model being too simple to reflect the complex non-linear relationship between the change of passenger flow density and the power demand in the actual scenario. The combination of the passenger flow density and the maximum passenger flow density in the model can not only dynamically adapt to the differences in different scales and places, but also reflect the proportion of the passenger flow density in the venue capacity, and has upper and lower limit constraints on the calculation of the power demand. P baseAs the introduction of the base power, it ensures that the system can still provide stable basic communication coverage in the case of low or even zero passenger flow density, avoiding the signal blind spot problem caused by excessive power reduction. Compared with the conventional linear model, this design simulates the significant growth of communication power demand under high passenger flow density through a non-linear growth method. The setting of the non-linear adjustment factor provides additional flexibility, enabling the model to adjust the curve slope of power growth according to the actual characteristics of the venue. In addition, this design innovation is reflected in the gradual enhancement of the influence weight of passenger flow density, especially in the case of high density, the power distribution is more reasonable, solving the defect that the conventional linear model cannot accurately reflect the surging power demand. Generally speaking, this model not only ensures the practical feasibility of power distribution, but also fully considers the complexity of passenger flow changes in the wireless communication environment.

[0082] Finally, the relative errors between the optimal overall transmission power and the real-time overall transmission power corresponding to the passenger flow density in each group of training data are calculated as follows:

[0083]

[0084] where ΔP is the relative error, and P real is the real-time overall transmission power in this group of training data. Compare the relative error with the preset threshold. If the relative error is less than or equal to the preset threshold, it is determined that the real-time overall transmission power is the final optimal overall transmission power corresponding to the passenger flow density in this group of training data. If the relative error is greater than the preset threshold, it is determined that the optimal overall transmission power calculated by the simulation model is the final optimal overall transmission power corresponding to the passenger flow density in this group of training data.

[0085] In implementation, by performing relative error analysis on the optimal overall transmission power calculated by the simulation model and the real-time overall transmission power in the training data through the above method, the advantages of theoretical calculation and actual data can be effectively combined, ensuring that the final optimal overall transmission power not only has the theoretical optimization effect but also meets the actual scenario requirements. If the relative error is small, it means that the real-time overall transmission power is already close enough to the theoretical optimal value, and directly adopting it can avoid the deviation that may be caused by over-relying on the simulation model; if the relative error is large, it means that the real-time overall transmission power may not reach the optimum, and at this time, using the simulation result can make up for the potential unreasonable configuration in the actual scenario, thereby enhancing the adaptability and robustness of the system to different scenarios while ensuring accuracy.

[0086] Step S103: Use the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output to train the preset neural network model to obtain a power prediction model.

[0087] In implementation, before training a preset neural network model, since the training dataset usually comes from the data collected in the actual environment, and these data may be insufficient in some cases, especially for complex non-linear problems, it is necessary to augment the training dataset. Using the passenger flow density of each group of training data in the augmented training dataset as the input and the corresponding final optimal overall transmission power as the output, the preset neural network model is trained to obtain a power prediction model.

[0088] In implementation, augmenting the dataset enables the power prediction model to be exposed to more different data points, preventing the model from being trained only in the local area of the training dataset, thereby improving its prediction accuracy on new data. Through data augmentation, small fluctuations that may occur in the passenger flow density and transmission power in the real world can be considered, which can make the training data closer to the actual application scenario.

[0089] It should be noted that without data augmentation, the model may only be trained within the range of the original data, which may lead to a decline in its prediction ability when facing new and unseen combinations of passenger flow density and transmission power. However, in essence, it is also feasible not to augment the training dataset in the technical solution of this application.

[0090] Specifically, when augmenting the training dataset, assume that the training dataset contains the passenger flow density D of each training data n and the corresponding final optimal overall transmission power P best,n , and it is desired to generate the augmented training data D n,new and P best,n,new . The augmentation formula is as follows:

[0091] D n,new = D n ×(1 + ∈);

[0092]

[0093] where ∈ is a perturbation term representing the random change in the passenger flow density, ∈ ∈ [-0.1, 0.1]. γ is a non-linear adjustment factor, which has the same meaning as in step S102 and is set to 1.2.

[0094] In the design of the above amplification formula, the amplification method utilizes the non - linear relationship between passenger flow density and transmission power, and adjusts the existing data by introducing a perturbation term. The perturbation term is used to simulate the fluctuations of passenger flow density in the actual scenario (such as seasonal changes, emergencies, etc.), making the data set closer to the real scenario. The range of the perturbation term is set between - 0.1 and 0.1, that is, a small - amplitude random fluctuation is added based on the passenger flow density and the optimal transmission power. The perturbation with a smaller range can generate new data points near the original data distribution, ensuring that the amplified data does not deviate from the actual scenario and retaining the authenticity and reliability of the original data. The small - range perturbation increases the diversity of the data but does not introduce extreme values, which can improve the generalization ability of the prediction model, avoid over - fitting caused by data deviation. In addition, the small perturbation can help the model learn the data distribution characteristics in a smoother way during the training process, avoiding an increase in training difficulty or unstable convergence due to excessive perturbation. In this way, the model can learn more changing situations, improve the robustness of the model, and avoid over - fitting.

[0095] In summary, by introducing the perturbation term and the non - linear adjustment factor, the amplification formula enables the generated new data points to not only consider the non - linear relationship between passenger flow density and the optimal transmission power but also simulate the random fluctuations of data in the real environment. This method is more in line with the uncertainty and complexity of data in practical applications than traditional amplification methods, and can effectively improve the generalization ability and robustness of the model. Compared with conventional methods, this amplification method can better reflect complex non - linear relationships, thereby improving the quality and diversity of training data and ensuring that the power prediction model performs more accurately and stably in the actual scenario.

[0096] Optionally, the preset neural network model in this application is a regression neural network model. A regression neural network model is a neural network suitable for handling regression problems. Especially when there is a certain non - linear relationship between the input and output, the regression neural network model can capture complex mapping relationships through its multi - layer non - linear structure. Other types of neural network models can also be selected, and this application does not limit the specific type of the preset neural network model.

[0097] Step S104: Real - time obtain the passenger flow density in the target venue and input it into the power prediction model. The power prediction model outputs the optimal overall transmission power, and the digital beam controller generates corresponding digital control signals according to the optimal overall transmission power output by the power prediction model and sends them to all antenna units.

[0098] In step S104, first, the passenger flow density in the target venue is obtained in real time by sensors pre-deployed in the target venue. For example, infrared sensors are arranged in the main channels or areas in the target venue to detect the number of people passing through; or the existing Wi-Fi probe technology is used to estimate the passenger flow density by counting the number of mobile devices. In this application, the passenger flow density is usually expressed as the number of people per unit area or per square meter (e.g., people / square meter). Subsequently, the obtained passenger flow density in the target venue is input into the power prediction model. Based on the input passenger flow density, the power prediction model outputs the optimal overall transmission power at this passenger flow density through the learned rules and experiences, that is, the most suitable antenna transmission power level at this passenger flow density. Finally, the digital beam controller generates corresponding digital control signals according to the optimal overall transmission power output by the power prediction model and the distribution characteristics of the antenna units, and sends them to each antenna unit through wireless communication.

[0099] It should be noted that in implementation, the passenger flow density in the target venue is not obtained and the transmission power is adjusted in real time and continuously. Instead, the passenger flow density in the target venue is obtained at regular intervals, and the overall transmission power of the antenna is controlled and adjusted based on the latest passenger flow density.

[0100] In addition, preferably, a loss function is designed, and the parameters of the power prediction model are updated using the gradient descent method based on the loss function until the loss function is minimized. The loss function is as follows:

[0101]

[0102] where M is the number of data in the augmented training dataset, that is, the number of groups of training data. P best,n is the true optimal overall transmission power of the nth training data calculated by the simulation model, is the predicted optimal overall transmission power of the nth training data output by the power prediction model, D max represents the maximum value of the passenger flow density in the training data. D n represents the passenger flow density of the nth training data. ∈ is a perturbation term, indicating the random change of the passenger flow density, and ∈∈[-0.1,0.1].

[0103] In the design process of the above loss function, introducing relative error into the loss function helps reduce the over-punishment of some samples caused by excessive absolute value differences, and avoids the model unfairly biasing towards certain data points due to excessive errors in the case of high transmission power. It ensures that the model focuses on relative prediction errors rather than just numerical differences. Data points with higher passenger flow density contribute more to the total loss, which makes the model pay more attention to the performance in high-density situations, meeting the requirements in the actual situation. For example, in high-density situations, the system requires more precise transmission power control, so high-density data points should have a greater impact on model training. The perturbation term can simulate the random fluctuations of the data, enabling the model to better adapt to different environmental changes that may occur in reality and enhancing its generalization ability. By adding these perturbations to the training data, the model can be trained in an environment that is not completely consistent or ideal, enhancing the model's adaptability to future changes. Through the above loss function, the power prediction model not only depends on the influence of passenger flow density on transmission power but also considers the relative error of the actual power. This approach can guide the power prediction model to make more accurate and practical transmission power predictions, especially when the system faces a dynamically changing environment.

[0104] In summary, combined with Figure 2 , by establishing communication between the digital beam controller and all antenna units in the target venue, the problem that the traditional indoor antenna system cannot adjust power in real time according to passenger flow density is first solved. Then, by collecting multiple sets of training data of passenger flow density and corresponding transmission power, and combining with the simulation model to calculate the optimal overall transmission power corresponding to each data point, it is ensured that the power output can be dynamically adjusted according to the change in passenger flow. Next, these training data are input into a preset neural network model for training to obtain a power prediction model. This model can predict the corresponding optimal overall transmission power based on the real-time obtained passenger flow density data in the target venue. Finally, the digital beam controller generates corresponding digital control signals according to the prediction result and sends them to all antenna units to achieve precise adjustment of the overall transmission power of the antenna units. Through this process, this application effectively solves the problem that the fixed power output of the existing antenna system cannot adapt to the changing demands of the tidal network, and can adjust the power according to actual needs at different time periods, thereby improving the utilization efficiency of network resources, avoiding waste, meeting the dynamically changing network demands, and achieving efficient management of the indoor antenna system.

[0105] Figure 3 FIG. 9 is a structural block diagram of a transmission power prediction system for an indoor antenna system based on deep learning provided by an embodiment of the present application. The system at least includes the following modules:

[0106] A communication establishment module, configured to establish communication between the digital beam controller and all antenna units in the target site. The digital beam controller controls the overall transmission power of all antenna units through digital signals.

[0107] A data acquisition module, configured to acquire a training data set containing multiple groups of training data. The training data includes the passenger flow density and its corresponding real-time overall transmission power, and based on the training data and a preset simulation model, calculates the final optimal overall transmission power corresponding to the passenger flow density in each group of training data.

[0108] A model training module, configured to use the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output, to train a preset neural network model to obtain a power prediction model.

[0109] A power control module, configured to acquire the passenger flow density in the target site in real time and input it into the power prediction model. The power prediction model outputs the optimal overall transmission power, and the digital beam controller generates a corresponding digital control signal according to the optimal overall transmission power output by the power prediction model and sends it to all antenna units.

[0110] For related details, refer to the above method embodiment.

[0111] Figure 4 It is a block diagram of an electronic device provided by an embodiment of the present application. The device at least includes a processor 401 and a memory 402.

[0112] The processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 401 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor 401 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 401 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 401 may further include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0113] The memory 402 may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory 402 may further include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 402 is used to store at least one instruction, and the at least one instruction is used to be executed by the processor 401 to implement the method for predicting the transmission power of an indoor antenna system based on deep learning provided in the method embodiments of the present application.

[0114] In some embodiments, the electronic device may further optionally include: a peripheral device interface and at least one peripheral device. The processor 401, the memory 402, and the peripheral device interface may be connected through a bus or signal lines. Each peripheral device may be connected to the peripheral device interface through a bus, signal lines, or a circuit board. Schematically, the peripheral devices include, but are not limited to: a radio frequency circuit, a touch display screen, an audio circuit, and a power supply, etc.

[0115] Of course, the electronic device may also include fewer or more components, and this embodiment does not limit this.

[0116] Optionally, the present application further provides a computer-readable storage medium, and a program is stored in the computer-readable storage medium. The program is loaded and executed by the processor to implement the method for predicting the transmission power of an indoor antenna system based on deep learning in the above method embodiments.

[0117] Optionally, the present application also provides a computer product, which includes a computer-readable storage medium. A program is stored in the computer-readable storage medium and is loaded and executed by a processor to implement the method for predicting the transmission power of an indoor antenna system based on deep learning in the above method embodiments.

[0118] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0119] The above embodiments only express several implementation manners of the present application, and the description thereof is relatively specific and detailed. However, it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for predicting the transmission power of an indoor antenna system based on deep learning, characterized in that, The method includes: Establish communication between the digital beam controller and all antenna units in the target venue, and the digital beam controller controls the overall transmission power of all antenna units through digital signals; Obtain a training data set containing multiple groups of training data. The training data includes the passenger flow density and its corresponding real-time overall transmission power. Based on the training data and a preset simulation model, calculate the final optimal overall transmission power corresponding to the passenger flow density in each group of training data; Use the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output to train a preset neural network model to obtain a power prediction model; Before using the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output to train a preset neural network model to obtain a power prediction model, it further includes: Augment the training dataset. Assume that the passenger flow density D of each training data is included in the training dataset n and the corresponding final optimal overall transmission power P best,n , and it is desired to generate the augmented training data D n,new and P best,n,new . The augmentation formula is as follows: D n,new = D n × (1 + ∈); Where ∈ is a disturbance term, ∈∈[-0.1, 0.1], and γ is a non-linear adjustment factor; use the passenger flow density of each group of training data in the augmented training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output to train a preset neural network model to obtain a power prediction model; Design a loss function, and update the parameters of the power prediction model using the gradient descent method based on the loss function until the loss function is minimized. The loss function is as follows: Where M is the number of data in the augmented training dataset, and P best,n is the true optimal overall transmit power of the nth training data calculated by the simulation model, is the predicted optimal overall transmit power of the nth training data output by the power prediction model, D max represents the maximum value of passenger flow density in the training data, D n represents the passenger flow density of the nth training data, and ∈ is a perturbation term representing the random variation of passenger flow density; Real-time obtain the passenger flow density in the target venue and input it into the power prediction model. The power prediction model outputs the optimal overall transmission power. The digital beam controller generates corresponding digital control signals according to the optimal overall transmission power output by the power prediction model and sends them to all antenna units.

2. The method for predicting the transmission power of an indoor antenna system based on deep learning according to claim 1, wherein The establishment of communication between the digital beam controller and all antenna units in the target venue includes: Deploy a digital beam controller and several antenna units in the target venue, and several of the antenna units are distributed in different areas of the target venue; Establish wireless communication between the digital beam controller and the antenna units using the same communication protocol; The digital beam controller generates corresponding digital control signals according to the overall transmission power input by the user, combines the distribution characteristics of the antenna units, and sends them to each antenna unit through wireless communication.

3. The method for predicting the transmission power of an indoor antenna system based on deep learning according to claim 2, wherein The formula for the transmission power allocation of each antenna unit by the digital beam controller according to the overall transmission power input by the user is as follows: where, P i is the transmission power of the i-th antenna element, P total is the overall transmission power input by the user, G i is the gain of the i-th antenna element, d ci is the distance between the digital beam controller and the i-th antenna element, α is the path loss factor, N is the number of antenna elements, and j represents the summation index.

4. The method for predicting the transmission power of an indoor antenna system based on deep learning according to claim 3, characterized in that The calibration coefficient adjusts the calculated transmission power of each antenna unit as follows: Among them, is the transmit power of the i-th antenna unit after adjustment, C cal is the calibration coefficient; the calculation method of the calibration coefficient C cal is as follows: Among them, is the total actual transmission power of all antenna elements.

5. The method for predicting the transmission power of an indoor antenna system based on deep learning according to claim 1, characterized in that, The calculation of the final optimal overall transmission power corresponding to the passenger flow density in each group of training data based on the training data and a preset simulation model includes: Each group of training data includes the passenger flow density of a certain venue, the real-time overall transmission power of the antenna corresponding to the passenger flow density, the base power of the venue, and the set maximum passenger flow density of the venue; Introduce a simulation model to calculate the optimal overall transmission power corresponding to the passenger flow density in each group of training data. The simulation model is as follows: Among them, P best is the optimal overall transmission power, P base is the base power in this set of training data, D is the passenger flow density in this set of training data, D max is the maximum passenger flow density in this set of training data, and γ is the non-linear adjustment factor.

6. The method for predicting the transmission power of an indoor antenna system based on deep learning according to claim 5, wherein, The calculation of the final optimal overall transmission power corresponding to the passenger flow density in each group of training data based on the training data and a preset simulation model further includes: Calculate the relative errors between the optimal overall transmission power and the real-time overall transmission power corresponding to the passenger flow density in each group of training data as follows: where ΔP is the relative error, and P real is the real-time overall transmission power in this set of training data; Compare the relative error with a preset threshold. If the relative error is less than or equal to the preset threshold, it is determined that the real-time overall transmission power is the final optimal overall transmission power corresponding to the passenger flow density in this group of training data; If the relative error is greater than the preset threshold, it is determined that the optimal overall transmission power calculated by using the simulation model is the final optimal overall transmission power corresponding to the passenger flow density in this group of training data.

7. An indoor antenna system transmit power prediction system based on deep learning, characterized in that, The system includes: A communication establishment module, configured to establish communication between the digital beam controller and all antenna units in the target venue. The digital beam controller controls the overall transmission power of all antenna units through digital signals; A data acquisition module, configured to acquire a training data set including multiple groups of training data. The training data includes the passenger flow density and its corresponding real-time overall transmission power, and based on the training data and a preset simulation model, calculate the final optimal overall transmission power corresponding to the passenger flow density in each group of training data; A model training module, configured to use the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output, and train a preset neural network model to obtain a power prediction model; before using the passenger flow density of each group of training data in the training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output, and training a preset neural network model to obtain a power prediction model, it further includes: Augment the training dataset. Assume that the passenger flow density D of each training data is included in the training dataset n and the corresponding final optimal overall transmission power P best,n , and it is desired to generate the augmented training data Dx ,new and P best,n,new . The augmentation formula is as follows: D n,new = D n × (1 + ∈); where ∈ is a perturbation term, ∈∈[-0.1,0.1], and γ is a non-linear adjustment factor; use the passenger flow density of each group of training data in the augmented training data set as the input, and the final optimal overall transmission power corresponding to the passenger flow density as the output, and train a preset neural network model to obtain a power prediction model; Design a loss function, and update the parameters of the power prediction model using the gradient descent method based on the loss function until the loss function is minimized. The loss function is as follows: Among them, M is the number of data in the augmented training dataset, and P best,n is the true optimal overall transmission power of the nth training data calculated by the simulation model, is the predicted optimal overall transmission power of the nth training data output by the power prediction model, D max represents the maximum value of passenger flow density in the training data, D n represents the passenger flow density of the nth training data, ∈ is a disturbance term representing the random change of passenger flow density; A power control module, configured to acquire the passenger flow density in the target venue in real time and input it into the power prediction model. The power prediction model outputs the optimal overall transmission power, and the digital beam controller generates corresponding digital control signals according to the optimal overall transmission power output by the power prediction model and sends them to all antenna units.

8. An electronic device, characterized in that, The device includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement a method for predicting the transmission power of an indoor antenna system based on deep learning as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, A program is stored in the storage medium, and when the program is executed by the processor, it is used to implement a method for predicting the transmission power of an indoor antenna system based on deep learning as described in any one of claims 1 to 6.

Images