Power transmission line audible noise simulation method based on machine learning and PINN
By combining machine learning and the PINN model with acoustic propagation equations, the problem of fast, accurate, and low-cost simulation of audible noise in transmission lines under complex meteorological conditions was solved, achieving efficient audible noise simulation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XICHANG COLLEGE
- Filing Date
- 2025-12-22
- Publication Date
- 2026-05-01
AI Technical Summary
Existing technologies struggle to perform rapid, accurate, and computationally inefficient simulations of audible noise from transmission lines under complex weather conditions.
A simulation method based on machine learning and physical information neural network (PINN) is adopted. By training machine learning models and PINN models, audible noise is predicted using environmental feature data. The loss function is trained by combining the acoustic propagation equation, so as to achieve fast and accurate audible noise simulation.
It enables fast, accurate, and computationally inefficient simulation of audible noise under complex weather conditions, improving the efficiency and accuracy of the simulation process.
Smart Images
Figure CN121960123A_ABST
Abstract
Description
A Simulation Method for Audible Noise in Transmission Lines Based on Machine Learning and PINN Technical Field
[0001] This invention relates to the field of electromagnetic environment simulation of transmission lines, and in particular to a method for simulating audible noise in transmission lines based on machine learning and PINN. Background Technology
[0002] The simulation of audible noise from transmission lines is of great significance in power grid planning and design. Based on the simulation results, appropriate conductor types and branch numbers can be selected, tower heights can be rationally designed, and line corridors can be planned to avoid the impact of audible noise on the lives and production of residents near the line.
[0003] Currently, methods for simulating audible noise from transmission lines mainly include empirical formula-based simulation and physics-based simulation. These two methods have the following drawbacks: empirical formula-based simulations have relatively large prediction errors for audible noise from transmission lines; while physics-based simulations require significant computational resources and time. Furthermore, current methods for simulating audible noise from transmission lines are difficult to apply to complex meteorological conditions.
[0004] Therefore, how to predict the audible noise of transmission lines quickly, accurately, and with low computational overhead under complex weather conditions is a technical bottleneck that needs to be overcome in the field of electromagnetic environment simulation of transmission lines. Summary of the Invention
[0005] This invention provides a method for simulating audible noise of transmission lines based on machine learning and PINN, which addresses the shortcomings of existing technologies in quickly, accurately, and with low computational overhead simulating audible noise of transmission lines under complex meteorological conditions. This method enables audible noise simulation that can respond to complex meteorological environments, while ensuring that the simulation process is fast, accurate, and has low computational overhead.
[0006] This invention provides a method for simulating audible noise in transmission lines based on machine learning and PINN, comprising the following steps: training a machine learning model using a first dataset to obtain a first prediction model; wherein the first dataset is a synchronization dataset matched to the transmission line, the synchronization dataset including audible noise data and environmental feature data; predicting a first prediction result using the first prediction model according to the set environmental features; wherein the first prediction result consists of multiple audible noises, the multiple audible noises corresponding to audible noises on multiple concentric cylindrical surfaces, the multiple concentric cylindrical surfaces all centered on the transmission line; training a PINN model using the multiple audible noises as boundary condition loss and using the acoustic propagation equation as physical loss to obtain a second prediction model; predicting a second prediction result using the second prediction model, and rendering the simulation result of audible noise in the transmission line based on the second prediction result; wherein the second prediction result is the numerical value of audible noise in the spatial coordinates of the transmission line.
[0007] According to the present invention, a method for simulating audible noise in transmission lines based on machine learning and PINN is provided. The method involves training a machine learning model using a first dataset to obtain a first prediction model, comprising: collecting raw audible noise data and raw environmental feature data to obtain a raw synchronous dataset; preprocessing the raw synchronous dataset to obtain a first dataset; the preprocessing of the raw synchronous dataset includes cleaning, correction, and data augmentation; using the environmental feature data in the first dataset as input and the audible noise data in the first dataset as output to train the machine learning model to obtain the first prediction model; wherein the first prediction model maps the relationship between the environmental features of the transmission line and the audible noise.
[0008] According to the present invention, a method for simulating audible noise in transmission lines based on machine learning and PINN further includes: data augmentation of the original synchronous dataset; wherein, the data augmentation of the original synchronous dataset includes: training a generator and a discriminator of a first generative model using the original synchronous dataset until the discriminator can no longer distinguish the authenticity of a second dataset; wherein, the second dataset is sample data generated by the generator of the first generative model; generating a third dataset using the generator of the first generative model, filtering the third dataset and adding it to the original synchronous dataset to obtain a first dataset; wherein, the filtering of the third dataset refers to the data distribution of the original synchronous dataset.
[0009] According to the present invention, a method for simulating audible noise in transmission lines based on machine learning and PINN is provided. The PINN model includes: an input layer for inputting spatial coordinates; a network layer comprising multiple parallel neural networks, each neural network learning acoustic propagation equations for different center frequencies; an output layer for outputting the A-weighted sound level of audible noise, the A-weighted sound level corresponding to the spatial coordinates of the input layer; and a loss function layer for updating model parameters, the loss function layer consisting of a weighted sum of physical loss and boundary condition loss; the physical loss being the acoustic propagation equation, and the boundary condition loss including multiple audible noises predicted by a first prediction model.
[0010] According to the present invention, a method for simulating audible noise in transmission lines based on machine learning and PINN is provided, wherein the boundary condition loss further includes an audible noise boundary at a distance K from the ground.
[0011] According to the present invention, a method for simulating audible noise of transmission lines based on machine learning and PINN is provided, wherein the learned frequency of the neural network is the center frequency of 1 / 3 octave band.
[0012] According to the present invention, a method for simulating audible noise of transmission lines based on machine learning and PINN is provided. The method renders the simulation result of audible noise of transmission lines based on the second prediction result, including: configuring legend colors and scales according to the numerical extreme values of the audible noise; wherein the numerical extreme values include maximum and minimum values; calculating the RGB values of the gradient color band based on the numerical value of the audible noise; and rendering the simulation result of audible noise of transmission lines based on the legend colors and scales, and the RGB values of the gradient color band.
[0013] This invention also provides a transmission line audible noise simulation device based on machine learning and PINN, comprising the following modules: a first training module, used to train a machine learning model using a first dataset to obtain a first prediction model; wherein the first dataset is a synchronization dataset matched to the transmission line, the synchronization dataset including audible noise data and environmental feature data; a prediction module, used to predict a first prediction result using the first prediction model according to set environmental features; wherein the first prediction result is multiple audible noises, the multiple audible noises corresponding to audible noises on multiple concentric cylindrical surfaces, the multiple concentric cylindrical surfaces all centered on the transmission line; a second training module, used to train a PINN model using multiple audible noises as boundary condition loss and using the acoustic propagation equation as physical loss to obtain a second prediction model; and a simulation module, used to predict a second prediction result using the second prediction model and render a transmission line audible noise simulation result based on the second prediction result; wherein the second prediction result is the numerical value of the audible noise in the spatial coordinates of the transmission line.
[0014] According to the present invention, a transmission line audible noise simulation device based on machine learning and PINN is provided. The first training module includes: a data acquisition submodule for acquiring raw audible noise data and raw environmental feature data to obtain a raw synchronous dataset; a preprocessing submodule for preprocessing the raw synchronous dataset to obtain a first dataset; wherein the preprocessing of the raw synchronous dataset includes cleaning, correction, and data augmentation; and a training submodule for using environmental feature data from the first dataset as input and audible noise data from the first dataset as output to train a machine learning model to obtain a first prediction model; wherein the first prediction model maps the relationship between transmission line environmental features and audible noise.
[0015] According to the present invention, a transmission line audible noise simulation device based on machine learning and PINN is provided. The preprocessing submodule includes: a generative model training unit, which is used to train the generator and discriminator of a first generative model using an original synchronous dataset until the discriminator can no longer distinguish the authenticity of a second dataset; wherein the second dataset is sample data generated by the generator of the first generative model; and a data generation unit, which is used to generate a third dataset using the generator of the first generative model, filter the third dataset and supplement it to the original synchronous dataset to obtain a first dataset; wherein the filtering of the third dataset refers to the data distribution of the original synchronous dataset.
[0016] According to the present invention, a transmission line audible noise simulation device based on machine learning and PINN is provided. The second training module includes a PINN model submodule, which includes: an input unit for inputting spatial coordinates; a network unit comprising multiple parallel neural networks, each neural network for learning acoustic propagation equations for different center frequencies; an output unit for outputting the A-weighted sound level of audible noise, the A-weighted sound level of which corresponds to the spatial coordinates of the input layer; and a loss function unit for updating model parameters, the loss function layer being composed of a weighted sum of physical loss and boundary condition loss; the physical loss being the acoustic propagation equation, and the boundary condition loss including multiple audible noises; the multiple audible noises being predicted by a first prediction model.
[0017] According to the present invention, a transmission line audible noise simulation device based on machine learning and PINN is provided, wherein the boundary condition loss further includes an audible noise boundary at a distance K from the ground.
[0018] According to the present invention, a transmission line audible noise simulation device based on machine learning and PINN is provided, wherein the learned frequency of the neural network is the center frequency of 1 / 3 octave band.
[0019] According to the present invention, a transmission line audible noise simulation device based on machine learning and PINN is provided. The simulation module includes: a configuration submodule, which is used to configure the legend color and scale according to the numerical extreme values of the audible noise; wherein the numerical extreme values include a maximum value and a minimum value; a calculation submodule, which is used to calculate the RGB values of the gradient color band according to the numerical value of the audible noise; and a rendering submodule, which is used to render the transmission line audible noise simulation result according to the legend color and scale, and the RGB values of the gradient color band.
[0020] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for simulating audible noise in transmission lines based on machine learning and PINN.
[0021] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for simulating audible noise in transmission lines based on machine learning and PINN.
[0022] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described methods for simulating audible noise in transmission lines based on machine learning and PINN.
[0023] The technical effects of this invention are as follows: This invention provides a method for simulating audible noise in transmission lines based on machine learning and PINN (Physics-Informed Neural Networks). A first prediction model is obtained by training a machine learning model using a first dataset. The first dataset is a synchronization dataset matched to the transmission line, and the synchronization dataset includes audible noise data and environmental feature data. Based on the set environmental features, a first prediction result is obtained using the first prediction model. The first prediction result consists of multiple audible noises, each corresponding to audible noise on multiple concentric cylindrical surfaces, all centered on the transmission line. The PINN model is trained using these multiple audible noises as boundary condition losses and the acoustic propagation equation as the physical loss, resulting in a second prediction model. A second prediction result is obtained using the second prediction model, and the simulation result of audible noise in the transmission line is rendered based on the second prediction result. The second prediction result is the numerical value of the audible noise in the spatial coordinates of the transmission line. Compared with existing technologies that struggle to quickly, accurately, and with low computational overhead achieve audible noise simulation of transmission lines under complex meteorological conditions, this method achieves audible noise simulation that can respond to complex meteorological environments while ensuring a fast, accurate, and low computational overhead simulation process. Attached Figure Description
[0024] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0025] Figure 1 is a flowchart illustrating a method for simulating audible noise in transmission lines based on machine learning and PINN provided by the present invention.
[0026] Figure 2 is a schematic diagram of the structure of the first generative model provided by the present invention for use in an embodiment of the present invention.
[0027] Figure 3 is a flowchart illustrating the machine learning model determination method provided by the present invention for embodiments of the present invention.
[0028] Figure 4 is a schematic diagram of the structure of the PINN model used in an embodiment of the present invention.
[0029] Figure 5 is a schematic diagram of the structure of a transmission line audible noise simulation device based on machine learning and PINN provided by the present invention.
[0030] Figure 6 is a schematic diagram of the structure of the electronic device provided by the present invention.
[0031] Figure 7 shows the software framework diagram of the audible noise simulation system for transmission lines based on machine learning and PINN provided in this embodiment of the invention. Detailed Implementation
[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0033] The present invention will now be described with reference to Figures 1-7.
[0034] Figure 1 is a flowchart of a method for simulating audible noise in transmission lines based on machine learning and PINN provided by the present invention. As shown in Figure 1, the method includes the following steps: Step 101: Train a machine learning model using a first dataset to obtain a first prediction model; wherein, the first dataset is a synchronization dataset matched to the transmission line, and the synchronization dataset includes audible noise data and environmental feature data.
[0035] In step 101, the machine learning models include decision trees, backpropagation neural networks, random forests, etc., and the machine learning model with the highest prediction accuracy is selected.
[0036] Optionally, in step 101, the machine learning model with the highest prediction accuracy is selected and implemented through automated machine learning program design; the parameter selection of the machine learning model is optimized through grid optimization or Bayesian optimization methods.
[0037] Optionally, step 101 includes steps A1 to A3: Step A1: Collect raw audible noise data and raw environmental feature data to obtain raw synchronous dataset.
[0038] Step A2: Preprocess the original synchronous dataset to obtain the first dataset; wherein, the preprocessing of the original synchronous dataset includes cleaning, correction and data augmentation; the purpose of preprocessing is to fill in missing values in the original monitoring data, correct outliers in the original data, convert the original data into the format and type required for model training, and expand the dataset using data augmentation.
[0039] Step A3: Using the environmental feature data in the first dataset as input and the audible noise data in the first dataset as output, train the machine learning model to obtain the first prediction model. The first prediction model maps the relationship between the environmental features of the transmission line and the audible noise. In the above step A1, the sources of the original audible noise data and the original environmental feature data include the transmission lines of the corona cage, the test section and the actual line.
[0040] Optionally, in step A1, the environmental characteristic data includes temperature (°C), relative humidity (%), precipitation (mm), wind speed (m / s), air pressure (KPa), and light intensity.
[0041] Optionally, in step A1, the raw environmental characteristic data is obtained through meteorological monitoring equipment installed on the ground or on a tower.
[0042] Optionally, in step A1, the raw audible noise data is acquired by an audible noise monitoring device installed on the ground.
[0043] It is understandable that by obtaining the raw environmental characteristic data from the synchronous monitoring equipment and the raw audible noise data from the audible noise monitoring equipment, the synchronous raw data of the audible noise and meteorological characteristics of the target line can be obtained. These data can reflect the correlation between audible noise and environmental characteristics.
[0044] In step A2 above, missing values are filled in and outliers are corrected through statistical methods, interpolation methods, machine learning methods, etc.
[0045] In step A2 above, data augmentation is performed on the original synchronous dataset, including steps A21 and A22: Step A21: The generator and discriminator of the first generative model are trained using the original synchronous dataset until the discriminator can no longer distinguish the authenticity of the second dataset; the second dataset is sample data generated by the generator of the first generative model.
[0046] Step A22: Use the generator of the first generative model to generate a third dataset, filter the third dataset and add it to the original synchronous dataset to obtain the first dataset, thereby achieving a balanced distribution of sample environmental features; the filtering of the third dataset refers to the data distribution of the original synchronous dataset, and the third dataset is a fake sample similar to real data.
[0047] Optionally, in step A21, the first generative model for data augmentation of the original synchronous dataset includes generative adversarial networks (GANs), diffusion models (DDPM, DDIM, Stable Diffusion, DALL-E 2, Imagen), and variational autoencoders (VAE, β-VAE, VQ-VA).
[0048] Optionally, referring to Figure 2, step A21 provides a schematic diagram of the structure of the first generative model used in the embodiment of the present invention, including: an environment parameter generator 211, a fake sample generator 212 and a non-coupled data discriminator 220, wherein the combination of the environment parameter generator and the fake sample generator is called the non-coupled data generator 210.
[0049] Referring to Figure 2, we define an input noise variable PZ(Z) and use an environment parameter generator to generate an environment parameter variable C. The fake sample generator includes a machine learning model that uses a trained environment to predict audible noise and combines environment and audible noise data to obtain a decoupled data variable Ci.
[0050] Optionally, the audible noise dataset can be categorized by having fewer samples, and more samples can be removed to keep the amount of data in each segment of the audible noise similar. These data can then be used to train the machine learning model in the GAN.
[0051] The aforementioned "few samples" refers to numerical ranges where audible noise occurs less frequently.
[0052] Optionally, the uncoupled data discriminator uses the preprocessed audible noise dataset as real samples to train a discriminator that judges the quality of generated data, with an output of 1 for true or 0 for false.
[0053] Optionally, in step A3, the first dataset is randomly divided into a training set and a test set, with the ratio of training set to test set being greater than 8:2.
[0054] Referring to Figure 3, a machine learning model determination method for an embodiment of the present invention is provided, including steps 310 to 320: Step 310: Design an automated machine learning framework.
[0055] Step 320: Based on the input and output requirements of different types of machine learning models, convert the data into the format required by the model.
[0056] Step 330: Design multiple different machine learning model structures according to requirements, train the machine learning models, and achieve automatic parameter optimization.
[0057] Step 102: Based on the set environmental characteristics, use the first prediction model to predict and obtain the first prediction result; wherein, the first prediction result is multiple audible noises, the multiple audible noises correspond to audible noises on multiple concentric cylindrical surfaces, and the multiple concentric cylindrical surfaces are all centered on the transmission line.
[0058] Step 103: Use multiple audible noises as boundary condition losses and the acoustic propagation equation as physical loss to train the PINN model and obtain the second prediction model; the PINN model is trained on a coarse-grained coordinate set.
[0059] In step 103 above, the acoustic propagation equations include the linear Navier-Stokes equations and empirical formulas.
[0060] Referring to Figure 4, in step 103 above, a schematic diagram of the PINN model used in this embodiment of the invention is provided. The PINN model includes: an input layer 410 for inputting spatial coordinates (X, Y, Z); a network layer 420 including multiple parallel neural networks, each neural network for learning acoustic propagation equations for different center frequencies; an output layer 430 for outputting audible noise A-weighted sound level values; the audible noise A-weighted sound level values correspond to the spatial coordinates of the input layer 410; and a loss function layer 440 for updating model parameters; the loss function layer 440 is composed of a weighted sum of physical loss and boundary condition loss; the physical loss is the acoustic propagation equation, and the boundary condition loss includes multiple audible noises; the multiple audible noises are predicted by a first prediction model.
[0061] Optionally, the output layer 410 outputs audible noise A-weighted sound level values, which are obtained by predicting the sound intensity at different center frequencies by each neural network and weighting the results.
[0062] Optionally, the boundary condition loss also includes an audible noise boundary at a distance K from the ground, generating a high-altitude sound field that matches the terrain. The distance K is typically 1.5m.
[0063] Optionally, the learned frequency of the neural network is the center frequency of 1 / 3 octave band.
[0064] In step 103 above, both boundary condition loss and physical loss are part of the loss function in the PINN model. The loss function includes the difference between the sound intensity output by different neural networks and the monitored value, and the difference between the A-weighted sound level output at the boundary and the monitored value.
[0065] Step 104: Use the second prediction model to predict the second prediction result, and render the simulation result of the audible noise of the transmission line based on the second prediction result; wherein, the second prediction result is the numerical value of the audible noise under the spatial coordinates of the transmission line; the simulation result of the audible noise of the transmission line is a heat map of the spatial distribution of audible noise.
[0066] In step 104 above, the audible noise in the spatial coordinates of the transmission line is predicted using the second prediction model, and the prediction is performed on a finer-grained coordinate set. The finer-grainedness in step 104 is compared with the coarse-grainedness in step 103. The spatial coordinate granularity is selected according to the actual situation of balancing simulation speed and simulation accuracy.
[0067] In step 104 above, the step of rendering the audible noise simulation result of the transmission line based on the numerical values of the audible noise in the spatial coordinates of the transmission line includes steps B1 to B3: Step B1: Configure the legend color and scale according to the numerical extreme values of the audible noise; wherein, the numerical extreme values include the maximum value and the minimum value; Step B2: Calculate the RGB values of the gradient color band according to the numerical values of the audible noise; Step B3: Render the audible noise simulation result of the transmission line based on the legend color and scale, and the RGB values of the gradient color band.
[0068] Optionally, the audible noise simulation method for transmission lines based on machine learning and PINN provided in the embodiment further includes step 105: Step 105: Generate an annual audible noise animation and audible noise characteristics of key spatial locations based on the interannual environmental monitoring time series data of the transmission line.
[0069] In step 105, an annual audible noise animation and audible noise characteristics of key spatial locations are generated based on the interannual environmental monitoring time-series data of transmission lines, including steps C1 to C4: Step C1: Collect the original interannual environmental monitoring time-series data of transmission lines, with sampling time not limited to 1 hour per sampling; Step C2: Perform outlier processing and missing value completion on the original interannual environmental monitoring time-series data of transmission lines to obtain the interannual environmental monitoring time-series data of transmission lines; Step C3: Generate an audible noise animation based on the interannual environmental monitoring time-series data of transmission lines; Step C4: Obtain the annual audible noise frequency distribution of key landmarks based on the audible noise animation and specific coordinates; the key landmarks include residential buildings and schools.
[0070] Optionally, step 105 further includes step C5: Step C5: The annual audible noise frequency distribution and audible noise spatial distribution heat map of key landmarks are superimposed on the geographic information layer as separate layers.
[0071] Optionally, step 105 also includes step C6: Step C6: Based on the annual audible noise frequency distribution of key landmarks, determine whether the audible noise at the key landmarks exceeds the standard; if so, issue an audible noise exceeding the standard alarm.
[0072] In step C6, according to GB 3096—2008 "Environmental Quality Standard for Noise", if the audible noise exceeds the standard requirements, a warning of excessive noise pollution will be displayed, and the warning color will be adjusted according to the proportion of exceeding the limit (green for exceeding the standard, yellow for exceeding the standard by 5% to 10%, orange for exceeding the standard by 10% to 15%, and red for exceeding the standard by more than 15%).
[0073] Optionally, the audible noise simulation method for transmission lines based on machine learning and PINN provided in the embodiment further includes steps 106 and 107: Step 106: Perform audible noise reflection sound field simulation.
[0074] Step 107: Overlay the simulation results of the audible noise reflection sound field with the simulation results of the audible noise of the transmission line to obtain the sound field simulation results considering the effect of the ground on the reflection of audible noise.
[0075] Optionally, in step 106, the simulation of the audible noise reflection sound field is performed using the PINN method or an empirical model method.
[0076] In step 106, the PINN method is used to simulate the audible noise reflection sound field, including: combining the reflectivity of the ground reflector to the audible noise and the audible noise intensity at the ground to obtain the sound intensity of the ground reflected sound source, and using this as the boundary condition to simulate the audible noise reflection sound field.
[0077] This invention provides a method for simulating audible noise in transmission lines based on machine learning and PINN. The method involves training a machine learning model using a first dataset to obtain a first prediction model; predicting multiple audible noises based on defined environmental characteristics using the first prediction model; training a PINN model using these multiple audible noises as boundary condition losses and the acoustic propagation equation as the physical loss to obtain a second prediction model; predicting the audible noise in the spatial coordinates of the transmission line using the second prediction model; and rendering the simulation results based on the numerical values of the audible noise in the spatial coordinates of the transmission line. Compared to existing technologies that struggle to quickly, accurately, and with low computational overhead achieve audible noise simulation of transmission lines under complex meteorological conditions, this method achieves audible noise simulation responsive to complex meteorological environments while ensuring a fast, accurate, and computationally inefficient simulation process.
[0078] The following describes a transmission line audible noise simulation device based on machine learning and PINN provided by the present invention. The transmission line audible noise simulation device based on machine learning and PINN described below can be referred to in correspondence with the transmission line audible noise simulation method based on machine learning and PINN described above.
[0079] Figure 5 is a schematic diagram of a transmission line audible noise simulation device based on machine learning and PINN provided by the present invention. As shown in Figure 5, the device includes the following: a first training module 201, which is used to train a machine learning model using a first dataset to obtain a first prediction model; wherein, the first dataset is a synchronization dataset matched to the transmission line, and the synchronization dataset includes audible noise data and environmental feature data; a prediction module 202, which is used to predict a first prediction result using the first prediction model according to the set environmental features; wherein, the first prediction result is multiple audible noises, the multiple audible noises correspond to audible noises on multiple concentric cylindrical surfaces, and the multiple concentric cylindrical surfaces are all centered on the transmission line; a second training module 203, which is used to train a PINN model using multiple audible noises as boundary condition losses and using the acoustic propagation equation as physical losses to obtain a second prediction model; and a simulation module 204, which is used to predict a second prediction result using the second prediction model and render a simulation result of the transmission line audible noise based on the second prediction result; wherein, the second prediction result is the numerical value of the audible noise in the spatial coordinates of the transmission line.
[0080] This invention provides a transmission line audible noise simulation device based on machine learning and PINN. It trains a machine learning model using a first dataset to obtain a first prediction model. Based on defined environmental characteristics, the first prediction model predicts multiple audible noises. Using these multiple audible noises as boundary condition losses and the acoustic propagation equation as the physical loss, a PINN model is trained to obtain a second prediction model. The second prediction model is used to predict the audible noise in the spatial coordinates of the transmission line. The simulation results of the transmission line audible noise are then rendered based on the numerical values of the audible noise in the spatial coordinates of the transmission line. Compared with existing technologies that struggle to quickly, accurately, and with low computational overhead achieve audible noise simulation of transmission lines under complex meteorological conditions, this invention achieves audible noise simulation that can respond to complex meteorological environments while ensuring a fast, accurate, and computationally inefficient simulation process.
[0081] Figure 6 illustrates a schematic diagram of the physical structure of an electronic device. As shown in Figure 6, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logic instructions in the memory 830 to execute a transmission line audible noise simulation method based on machine learning and PINN.
[0082] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0083] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute a transmission line audible noise simulation method based on machine learning and PINN provided by the above methods.
[0084] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a method for simulating audible noise in transmission lines based on machine learning and PINN provided by the methods described above.
[0085] Figure 7 illustrates the software framework diagram of the transmission line audible noise simulation system based on machine learning and PINN provided in an embodiment of the present invention. As shown in Figure 7, the software includes a front-end, a server-side, and a database. The front-end includes a view layer, and the development tools for the view layer include CSS, Vue, Element UI, and Echarts, etc. The view layer includes interface design, user login page, function selection page, simulation result display page, etc. The server-side includes a business logic layer and a data persistence layer. The business logic layer includes modules such as user management, permission management, function management, and line management. The data persistence layer includes user information, permission information, function information, line information, etc. The database includes data tables such as user information, line information, audible noise information, environmental characteristic information, and equipment information.
[0086] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0087] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for simulating audible noise in transmission lines based on machine learning and PINN, characterized in that, include: A machine learning model is trained using a first dataset to obtain a first prediction model. The first dataset is a synchronization dataset matched to the transmission line, comprising audible noise data and environmental feature data. Based on the defined environmental features, the first prediction model is used to predict a first prediction result. This first prediction result consists of multiple audible noises, each corresponding to audible noise on multiple concentric cylindrical surfaces, all centered on the transmission line. Using these multiple audible noises as boundary condition losses and the acoustic propagation equation as the physical loss, a PINN model is trained to obtain a second prediction model. The second prediction model is used to predict a second prediction result, and the simulation result of the audible noise of the transmission line is rendered based on this second prediction result. The second prediction result is the numerical value of the audible noise in the spatial coordinates of the transmission line.
2. The method for simulating audible noise in transmission lines based on machine learning and PINN according to claim 1, characterized in that, The step of training a machine learning model using a first dataset to obtain a first prediction model includes: collecting raw audible noise data and raw environmental feature data to obtain a raw synchronous dataset; preprocessing the raw synchronous dataset to obtain a first dataset; wherein the preprocessing of the raw synchronous dataset includes cleaning, correction, and data augmentation; using the environmental feature data in the first dataset as input and the audible noise data in the first dataset as output to train a machine learning model to obtain a first prediction model; wherein the first prediction model maps the relationship between the environmental features of the transmission line and audible noise.
3. The method for simulating audible noise in transmission lines based on machine learning and PINN according to claim 2, characterized in that, Also includes: Data augmentation is performed on the original synchronous dataset; wherein, the data augmentation of the original synchronous dataset includes: training the generator and discriminator of a first generative model using the original synchronous dataset until the discriminator can no longer distinguish the authenticity of a second dataset; wherein, the second dataset is sample data generated by the generator of the first generative model; generating a third dataset using the generator of the first generative model, filtering the third dataset and adding it to the original synchronous dataset to obtain the first dataset; wherein, the filtering of the third dataset refers to the data distribution of the original synchronous dataset.
4. The method for simulating audible noise in transmission lines based on machine learning and PINN according to claim 1, characterized in that, The PINN model includes: an input layer for inputting spatial coordinates; a network layer comprising multiple parallel neural networks, each learning the acoustic propagation equation for different center frequencies; an output layer for outputting the A-weighted sound level of audible noise, the A-weighted sound level corresponding to the spatial coordinates of the input layer; and a loss function layer for updating model parameters, the loss function layer consisting of a weighted sum of physical loss and boundary condition loss; the physical loss being the acoustic propagation equation, and the boundary condition loss including multiple audible noises predicted by a first prediction model.
5. The method for simulating audible noise in transmission lines based on machine learning and PINN according to claim 4, characterized in that, The boundary condition loss also includes the audible noise boundary at a distance K from the ground.
6. The method for simulating audible noise in transmission lines based on machine learning and PINN according to claim 4, characterized in that, The learned frequency of the neural network is the center frequency of 1 / 3 octave band.
7. The method for simulating audible noise in transmission lines based on machine learning and PINN according to claim 1, characterized in that, The simulation results of audible noise of the transmission line are rendered based on the second prediction result, including: configuring legend colors and scales according to the numerical extreme values of the audible noise; wherein the numerical extreme values include the maximum value and the minimum value; calculating the RGB values of the gradient color band according to the numerical value of the audible noise; and rendering the simulation results of audible noise of the transmission line based on the legend colors and scales and the RGB values of the gradient color band.
8. A transmission line audible noise simulation device based on machine learning and PINN, characterized in that, include: A first training module is used to train a machine learning model using a first dataset to obtain a first prediction model; wherein the first dataset is a synchronization dataset matched to the transmission line, and the synchronization dataset includes audible noise data and environmental feature data; a prediction module is used to predict a first prediction result using the first prediction model based on the set environmental features; wherein the first prediction result consists of multiple audible noises, each corresponding to audible noises on multiple concentric cylindrical surfaces, all centered on the transmission line; a second training module is used to train a PINN model using multiple audible noises as boundary condition losses and the acoustic propagation equation as the physical loss to obtain a second prediction model; a simulation module is used to predict a second prediction result using the second prediction model and render a simulation result of the audible noise of the transmission line based on the second prediction result; wherein the second prediction result is the numerical value of the audible noise in the spatial coordinates of the transmission line.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements a method for simulating audible noise in transmission lines based on machine learning and PINN as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a method for simulating audible noise in transmission lines based on machine learning and PINN as described in any one of claims 1 to 7.