Electrical equipment noise annoyance degree prediction model training method, device and equipment

Abstract

The invention relates to an electrical equipment noise annoyance degree prediction model training method, device and equipment. The method comprises the following steps: acquiring simulation noise data and corresponding subjective evaluation data of sample electrical equipment, acquiring objective evaluation data based on the simulation noise data and the subjective evaluation data, and constructing a training data set, a verification data set and a test data set by using the subjective and objective evaluation data; training model parameters of a to-be-trained prediction model by using the training data set to obtain an initial prediction model, training model hyper-parameters of the initial prediction model through an ant colony optimization algorithm based on the initial prediction model and the verification data set to obtain a target prediction model, and finally inputting the test data set into the target prediction model to obtain a target prediction model. And performing a performance test on the target prediction model, and if a result of the performance test represents that the target prediction model passes the performance test, taking the target prediction model as a trained prediction model. By adopting the method, the acquisition timeliness of the annoyance degree can be improved.

Description

Technical Field

[0001] The present application relates to the field of model training technology, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for training a model for predicting noise annoyance of electrical equipment. Background Art

[0002] In recent years, with the continuous improvement of my country's productivity and people's living standards, the national electricity demand has been growing, and the number of power workplaces such as power plants and substations has also increased day by day. However, the equipment noise problem has also increased. The workers in power sites and the surrounding residents are seriously annoyed by these industrial noises, which have an impact on people's physical and mental health.

[0003] At present, the main method for evaluating the noise of power equipment is the A-weighted sound pressure level. However, relevant studies have shown that the A-weighted sound pressure level is not accurate enough in evaluating the noise of power equipment.

[0004] Traditional studies on subjective annoyance are mostly based on questionnaire surveys. However, the current questionnaire survey methods or traditional methods have the problem of untimely noise detection. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for training an electrical equipment noise annoyance prediction model that can detect noise in a timely manner in response to the above-mentioned technical problems.

[0006] In a first aspect, the present application provides a method for training an electrical equipment noise annoyance prediction model, comprising:

[0007] Acquire simulated noise data of sample electrical equipment, and acquire subjective evaluation data corresponding to the simulated noise data;

[0008] Based on the simulated noise data and the subjective evaluation data, objective evaluation data associated with the simulated noise data is obtained, and a training data set, a validation data set, and a test data set are constructed using the subjective evaluation data and the objective evaluation data;

[0009] Using the training data set, the model parameters of the electrical equipment noise annoyance prediction model to be trained are trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model;

[0010] The test data set is input into the target electrical equipment noise annoyance prediction model, and a performance test is performed on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0011] In one embodiment, obtaining objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data includes:

[0012] Performing FFT transformation on the simulated noise data to obtain a simulated noise spectrum, and obtaining original objective evaluation data associated with the simulated noise data based on the simulated noise spectrum;

[0013] The original objective evaluation data were subjected to correlation analysis by SPSS to obtain the Pearson correlation coefficient associated with the original objective evaluation data;

[0014] Based on the Pearson correlation coefficient, redundant objective evaluation data in the original objective evaluation data are removed to obtain initial objective evaluation data;

[0015] Objective evaluation data is obtained based on the initial objective evaluation data and the subjective evaluation data.

[0016] In one embodiment, obtaining objective evaluation data according to initial objective evaluation data and subjective evaluation data includes:

[0017] SPSS was used to conduct correlation analysis on the initial objective evaluation data and subjective evaluation data, and the linear correlation coefficient between the initial objective evaluation data and the subjective evaluation data was obtained;

[0018] Initial objective evaluation data whose linear correlation coefficient is less than a preset linear correlation coefficient threshold are removed to obtain objective evaluation data.

[0019] In one embodiment, obtaining subjective evaluation data corresponding to simulated noise data includes:

[0020] Obtaining an adjective group of annoyance levels of sample users for simulated noise data;

[0021] Based on each annoyance degree adjective group, each simulated noise data is scored for annoyance degree, and a plurality of original subjective evaluation data are obtained;

[0022] The original subjective evaluation data were subjected to correlation analysis by SPSS to obtain the Spearman correlation coefficient associated with the original subjective evaluation data;

[0023] The original subjective evaluation data whose Spearman correlation coefficient is less than a preset Spearman correlation coefficient threshold are removed to obtain the subjective evaluation data.

[0024] In an exemplary embodiment, the simulated noise data is associated with an actual noise annoyance level;

[0025] The model parameters of the electrical equipment noise annoyance prediction model to be trained are trained using the training data set to obtain an initial electrical equipment noise annoyance prediction model, including:

[0026] Using Xavier initialization to initialize the model parameters of the original electrical equipment noise annoyance prediction model, to obtain the electrical equipment noise annoyance prediction model to be trained;

[0027] The subjective evaluation data and the objective evaluation data in the training data set are input into the noise annoyance prediction model of the electrical equipment to be trained to obtain the predicted noise annoyance corresponding to the simulated noise data;

[0028] Based on the difference between the predicted noise annoyance level and the actual noise annoyance level, and the loss function preset for the noise annoyance level prediction model of the electrical equipment to be trained, a corresponding model loss value is obtained;

[0029] The model parameters are updated according to the model loss value until the loss function converges, and the electrical equipment noise annoyance prediction model to be trained is used as the initial electrical equipment noise annoyance prediction model.

[0030] In one embodiment, based on the initial electrical equipment noise annoyance prediction model and the validation data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model, including:

[0031] Initialize candidate model hyperparameters of the initial electrical equipment noise annoyance prediction model, the initial ant colony pheromone of each path, the maximum number of iterations and the global error;

[0032] According to the pheromone concentration of the initial pheromone of the ant colony of each path, the node selection probability influence function, the influence factor and the expected influence factor, the transition probability of the current candidate model hyperparameter to the next candidate model hyperparameter is obtained, and the model hyperparameter combination of each path is obtained based on the transition probability;

[0033] Input the subjective evaluation data and objective evaluation data of the verification data set into the initial electrical equipment noise annoyance prediction model to obtain the predicted noise annoyance of the simulated noise data;

[0034] When the difference between the predicted noise annoyance and the actual noise annoyance is not greater than the global error, the initial pheromone of the ant colony of each path is updated, and the current model hyperparameter combination of each path is used as the current optimal solution, and the current model hyperparameter combination of each path is used as the new model hyperparameter combination of each path. The above process is repeated until the number of iterations reaches the maximum number of iterations, and the initial electrical equipment noise annoyance prediction model at this time is determined as the target electrical equipment noise annoyance prediction model.

[0035] In one embodiment, obtaining simulated noise data of a sample electrical device includes:

[0036] Acquire the current electric field force of the sample electrical device, and set the current electric field force in a pre-built finite element model of the sample electrical device to obtain the surface vibration acceleration of the sample electrical device;

[0037] The surface vibration acceleration is set as a simulation boundary condition in a pre-built simulation model of a sample electrical device, and an acoustic field simulation is performed in the simulation model to obtain simulated noise data of the sample electrical device.

[0038] In an exemplary embodiment, constructing a training data set, a validation data set, and a test data set using subjective evaluation data and objective evaluation data includes:

[0039] Remove outliers in the subjective evaluation data and the objective evaluation data, and supplement missing values ​​in the subjective evaluation data and the objective evaluation data to obtain cleaned subjective evaluation data and objective evaluation data;

[0040] Standardizing the cleaned subjective evaluation data and objective evaluation data to obtain standardized subjective evaluation data and objective evaluation data;

[0041] The standardized subjective evaluation data and objective evaluation data are used to construct training data sets, validation data sets, and test data sets.

[0042] In a second aspect, the present application also provides a device for training a noise annoyance prediction model for electrical equipment, comprising:

[0043] A data acquisition module, used to acquire simulated noise data of sample electrical equipment, and to acquire subjective evaluation data corresponding to the simulated noise data;

[0044] A data set construction module, used to obtain objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data, and to construct a training data set, a verification data set, and a test data set using the subjective evaluation data and the objective evaluation data;

[0045] A parameter training module is used to train the model parameters of the electrical equipment noise annoyance prediction model to be trained using the training data set to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, train the model hyperparameters of the initial electrical equipment noise annoyance prediction model using an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model;

[0046] The performance testing module is used to input the test data set into the target electrical equipment noise annoyance prediction model, perform a performance test on the target electrical equipment noise annoyance prediction model, and if the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0047] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0048] Acquire simulated noise data of sample electrical equipment, and acquire subjective evaluation data corresponding to the simulated noise data;

[0049] Based on the simulated noise data and the subjective evaluation data, objective evaluation data associated with the simulated noise data is obtained, and a training data set, a validation data set, and a test data set are constructed using the subjective evaluation data and the objective evaluation data;

[0050] Using the training data set, the model parameters of the electrical equipment noise annoyance prediction model to be trained are trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model;

[0051] The test data set is input into the target electrical equipment noise annoyance prediction model, and a performance test is performed on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0052] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

[0053] Acquire simulated noise data of sample electrical equipment, and acquire subjective evaluation data corresponding to the simulated noise data;

[0054] Based on the simulated noise data and the subjective evaluation data, objective evaluation data associated with the simulated noise data is obtained, and a training data set, a validation data set, and a test data set are constructed using the subjective evaluation data and the objective evaluation data;

[0055] Using the training data set, the model parameters of the electrical equipment noise annoyance prediction model to be trained are trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model;

[0056] The test data set is input into the target electrical equipment noise annoyance prediction model, and a performance test is performed on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0057] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0058] Acquire simulated noise data of sample electrical equipment, and acquire subjective evaluation data corresponding to the simulated noise data;

[0059] Based on the simulated noise data and the subjective evaluation data, objective evaluation data associated with the simulated noise data is obtained, and a training data set, a validation data set, and a test data set are constructed using the subjective evaluation data and the objective evaluation data;

[0060] Using the training data set, the model parameters of the electrical equipment noise annoyance prediction model to be trained are trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model;

[0061] The test data set is input into the target electrical equipment noise annoyance prediction model, and a performance test is performed on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0062] The above-mentioned electrical equipment noise annoyance prediction model training method, device, computer device, computer-readable storage medium and computer program product obtain simulated noise data of sample electrical equipment and subjective evaluation data corresponding to the simulated noise data, obtain objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data, and use the subjective evaluation data and the objective evaluation data to construct a training data set, a verification data set and a test data set, use the training data set to train the model parameters of the electrical equipment noise annoyance prediction model to be trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, train the model hyperparameters of the initial electrical equipment noise annoyance prediction model through an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model, finally input the test data set into the target electrical equipment noise annoyance prediction model, perform a performance test on the target electrical equipment noise annoyance prediction model, and if the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, then the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model. By training a model with the ability to predict user annoyance, relevant data is input into this model to obtain the corresponding predicted noise annoyance, ensuring that the user annoyance in various regions is obtained in a timely and advance manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0064] FIG1 is a diagram of an application environment of a method for training a model for predicting noise annoyance of electrical equipment in one embodiment;

[0065] FIG2 is a flow chart of a method for training a model for predicting the annoyance level of noise of electrical equipment in one embodiment;

[0066] FIG3 is a flow chart of a method for training a model for predicting the annoyance level of noise of electrical equipment in another embodiment;

[0067] FIG4 is a schematic diagram of measuring point settings of a simulation model of a sample electrical device in one embodiment;

[0068] FIG5 is a schematic diagram of measuring point settings of a simulation model of a sample electrical device in another embodiment;

[0069] FIG6 is a schematic diagram of a flow chart of an ant colony optimization algorithm in one embodiment;

[0070] FIG7 is a structural block diagram of an electrical equipment noise annoyance prediction model training device in one embodiment;

[0071] FIG. 8 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0073] The electrical equipment noise annoyance prediction model training method provided in the embodiment of the present application can be applied in the application environment shown in FIG1. ​​Among them, the sample electrical equipment 102 can perform various simulations on the terminal 104 to obtain corresponding simulated noise data, and the terminal 104 communicates with the server 106 through the network. The data storage system can store the data that the server 106 needs to process. The data storage system can be integrated on the server 106, or it can be placed on the cloud or other network servers. The server 106 obtains the simulated noise data of the sample electrical device 102 and the subjective evaluation data corresponding to the simulated noise data, obtains the objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data, and constructs a training data set, a verification data set and a test data set using the subjective evaluation data and the objective evaluation data, trains the model parameters of the electrical device noise annoyance prediction model to be trained using the training data set to obtain an initial electrical device noise annoyance prediction model, and trains the model hyperparameters of the initial electrical device noise annoyance prediction model using an ant colony optimization algorithm based on the initial electrical device noise annoyance prediction model and the verification data set to obtain a target electrical device noise annoyance prediction model, and finally inputs the test data set into the target electrical device noise annoyance prediction model, performs a performance test on the target electrical device noise annoyance prediction model, and if the result of the performance test indicates that the target electrical device noise annoyance prediction model passes the performance test, the target electrical device noise annoyance prediction model is used as the trained electrical device noise annoyance prediction model. The terminal 104 may be, but is not limited to, various personal computers, laptop computers, tablet computers, etc. The server 106 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0074] In an exemplary embodiment, as shown in FIG2 , a method for training a prediction model for noise annoyance of electrical equipment is provided, and the method is applied to the server 106 in FIG1 as an example for explanation, including the following steps S201 to S204. Among them:

[0075] Step S201 : obtaining simulated noise data of a sample electrical device, and obtaining subjective evaluation data corresponding to the simulated noise data.

[0076] Step S202 , based on the simulated noise data and the subjective evaluation data, objective evaluation data associated with the simulated noise data is acquired, and a training data set, a verification data set, and a test data set are constructed using the subjective evaluation data and the objective evaluation data.

[0077] Among them, the simulated noise data can be understood as the model after simulating the relevant electrical capacitor, and the noise generated by the model operation collected at the relevant measuring points after applying the relevant electric field force. The subjective evaluation data can be understood as the user's evaluation level and evaluation score for noise of different decibel levels. The objective evaluation data can be understood as the sound evaluation data used to describe the simulated noise data, such as sound pressure, loudness, pitch, jitter, sharpness, etc.

[0078] Optionally, the server 106 obtains the simulated noise data of the sample electrical device 102 from the terminal 104, and obtains the subjective evaluation data corresponding to the simulated noise data, and then obtains the objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data, and uses the subjective evaluation data and the objective data to construct a training data set, a verification data set, and a test data set. By collecting a large amount of data to construct the corresponding data set, a data foundation is laid for the subsequent model training of the electrical equipment noise annoyance prediction model, thereby improving the output reliability of the trained electrical equipment noise annoyance prediction model.

[0079] Step S203, using the training data set, training the model parameters of the electrical equipment noise annoyance prediction model to be trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, training the model hyperparameters of the initial electrical equipment noise annoyance prediction model through an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model.

[0080] Among them, the ant colony optimization algorithm can be understood as a heuristic optimization algorithm based on the foraging behavior of ants in nature.

[0081] For example, the server 106 trains the model parameters of the electrical equipment noise annoyance prediction model to be trained using the training data set to obtain an initial electrical equipment noise annoyance prediction model, and trains the model hyperparameters of the initial electrical equipment noise annoyance prediction model through an ant colony optimization algorithm based on the initial electrical equipment noise annoyance prediction model and the verification data set to obtain a target electrical equipment noise annoyance prediction model. The model parameters of the model are trained using the training data set, and the model hyperparameters of the model are trained using the verification data set, thereby enhancing the generalization ability of the model.

[0082] Step S204, inputting the test data set into the target electrical equipment noise annoyance prediction model, performing a performance test on the target electrical equipment noise annoyance prediction model, and if the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0083] Optionally, the server 106 inputs the test data set into the target electrical equipment noise annoyance prediction model, and performs a performance test on the target electrical equipment noise annoyance prediction model. For example, a correlation analysis is performed between the predicted noise annoyance output by the target electrical equipment noise annoyance prediction model and the actual noise annoyance associated with the simulated noise data. The two are linearly positively correlated, and the correlation coefficient is greater than a preset correlation coefficient threshold. At this time, the result of the performance test indicates that the target electrical equipment noise annoyance prediction model has passed the test. The target electrical equipment noise annoyance prediction model is then used as the trained electrical equipment noise annoyance prediction model. By further performing a performance test on the target electrical equipment noise annoyance prediction model, the generalization ability and learning ability of the trained electrical equipment noise annoyance prediction model are further enhanced.

[0084] In the above-mentioned electrical equipment noise annoyance prediction model training method, by obtaining simulated noise data of sample electrical equipment and obtaining subjective evaluation data corresponding to the simulated noise data, objective evaluation data associated with the simulated noise data are obtained based on the simulated noise data and the subjective evaluation data, and a training data set, a verification data set and a test data set are constructed using the subjective evaluation data and the objective evaluation data. The training data set is used to train the model parameters of the electrical equipment noise annoyance prediction model to be trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model. Finally, the test data set is input into the target electrical equipment noise annoyance prediction model, and a performance test is performed on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model. By training a model with the ability to predict user annoyance, relevant data is input into this model to obtain the corresponding predicted noise annoyance, ensuring that the user annoyance in various regions is obtained in a timely and advance manner.

[0085] In one embodiment, objective evaluation data associated with the simulated noise data is obtained based on the simulated noise data and the subjective evaluation data, including: performing FFT transformation on the simulated noise data to obtain a simulated noise spectrum, and obtaining original objective evaluation data associated with the simulated noise data based on the simulated noise spectrum; performing correlation analysis on the original objective evaluation data through SPSS to obtain a Pearson correlation coefficient associated with the original objective evaluation data; removing redundant objective evaluation data in the original objective evaluation data based on the Pearson correlation coefficient to obtain initial objective evaluation data; and obtaining objective evaluation data based on the initial objective evaluation data and the subjective evaluation data.

[0086] Among them, FFT (Fast Fourier Transform) can be understood as converting the time domain signal to the frequency domain to reveal the components of the signal at different frequencies; SPSS (Statistical Package for the Social Sciences) can be understood as a statistical analysis software, which is mainly used for data analysis in the field of social sciences and provides a wealth of statistical methods, such as descriptive statistics, inferential statistics, regression analysis, variance analysis, factor analysis, etc.; Pearson correlation coefficient can be understood as a statistical indicator used to measure the strength and direction of the linear relationship between two variables, with a value range of -1 to 1: 1 indicates a perfect positive correlation: when one variable increases, the other variable also increases at the same proportion; -1 indicates a perfect negative correlation: when one variable increases, the other variable decreases at the same proportion; 0 indicates no linear relationship: there is no linear correlation between the two variables.

[0087] For example, the server performs FFT transformation on the simulated noise data to obtain the simulated noise spectrum, obtains the main frequency distribution range based on the simulated noise spectrum, extracts the original objective evaluation data at the integer frequency point, and performs correlation analysis on the original objective evaluation data through SPSS to obtain the Pearson correlation coefficient associated with the original objective evaluation data, obtains the initial objective evaluation data, and finally obtains the objective evaluation data based on the initial objective evaluation data and the subjective evaluation data. Through spectrum analysis and corresponding data extraction, the accuracy of understanding and research on the characteristics of simulated noise is improved.

[0088] In one of the embodiments, objective evaluation data is obtained based on initial objective evaluation data and subjective evaluation data, including: using SPSS to perform correlation analysis on the initial objective evaluation data and the subjective evaluation data to obtain a linear correlation coefficient between the initial objective evaluation data and the subjective evaluation data; removing the initial objective evaluation data whose linear correlation coefficient is less than a preset linear correlation coefficient threshold to obtain the objective evaluation data.

[0089] Optionally, the server 106 uses SPSS to perform correlation analysis on the initial objective data and the subjective evaluation data to obtain a linear correlation coefficient between the initial objective evaluation data and the subjective evaluation data, removes the initial objective evaluation data whose linear correlation coefficient is less than a preset linear correlation coefficient threshold, and obtains the objective evaluation data. Combining the subjective evaluation data with the objective evaluation data provides a multi-dimensional evaluation result to better support decision-making and analysis.

[0090] In an exemplary embodiment, obtaining subjective evaluation data corresponding to simulated noise data includes: obtaining sample users' annoyance degree adjective groups for simulated noise data; scoring the annoyance degree of each simulated noise data based on each annoyance degree adjective group to obtain multiple original subjective evaluation data; performing correlation analysis on the original subjective evaluation data through SPSS to obtain the Spearman correlation coefficient associated with the original subjective evaluation data; and removing the original subjective evaluation data whose Spearman correlation coefficient is less than a preset Spearman correlation coefficient threshold to obtain the subjective evaluation data.

[0091] Among them, the Spearman's Rank Correlation Coefficient can be understood as a non-parametric statistical method for measuring the monotonic relationship between two variables; the annoyance degree adjective group can be understood as a pair of adjective antonyms used for a certain noise data user, such as quiet-noisy, etc.

[0092] For example, the server 106 obtains the sample user's annoyance degree adjective group for the simulated noise data, and then scores the annoyance degree of each simulated noise data based on each annoyance degree adjective group to obtain multiple original subjective evaluation data, and performs correlation analysis on the original subjective evaluation data through SPSS to obtain the Spearman correlation coefficient associated with the original subjective evaluation data, and removes the original subjective evaluation data whose Spearman correlation coefficient is less than a preset Spearman correlation coefficient threshold to obtain the subjective evaluation data. Through the scoring and screening process, irrelevant data is removed, the quality of the subjective evaluation data is improved, and it is ensured that the data can better reflect the user's true feelings.

[0093] In one embodiment, the simulated noise data is associated with an actual noise annoyance level;

[0094] The method uses a training data set to train the model parameters of the electrical equipment noise annoyance prediction model to be trained to obtain an initial electrical equipment noise annoyance prediction model, including: using Xavier initialization to initialize the model parameters of the original electrical equipment noise annoyance prediction model to obtain the electrical equipment noise annoyance prediction model to be trained; inputting the subjective evaluation data and the objective evaluation data in the training data set into the electrical equipment noise annoyance prediction model to be trained to obtain a predicted noise annoyance corresponding to the simulated noise data; based on the difference between the predicted noise annoyance and the actual noise annoyance, and the loss function pre-set for the electrical equipment noise annoyance prediction model to be trained, obtaining a corresponding model loss value; updating the model parameters according to the model loss value until the loss function converges, and using the electrical equipment noise annoyance prediction model to be trained as the initial electrical equipment noise annoyance prediction model.

[0095] Among them, Xavier Initialization (also known as Glorot Initialization) can be understood as a method for initializing the weights of a neural network.

[0096] Optionally, the server 106 uses Xavier initialization to initialize the model parameters of the original electrical equipment noise annoyance prediction model to obtain the electrical equipment noise annoyance prediction model to be trained, and then inputs the subjective evaluation data and the objective evaluation data in the training data set into the electrical equipment noise annoyance prediction model to be trained to obtain the predicted noise annoyance corresponding to the simulated noise data, and based on the difference between the predicted noise annoyance and the actual noise annoyance associated with the simulated noise data, and the loss function pre-set for the electrical equipment noise annoyance prediction model to be trained, the corresponding model loss value is obtained, and finally the model parameters are updated according to the model loss value until the loss function converges, and the electrical equipment noise annoyance prediction model to be trained is used as the initial electrical equipment noise annoyance prediction model. Through Xavier initialization, the model parameters have a reasonable initial distribution at the beginning of training, thereby improving the convergence speed of the model and the final prediction performance. By calculating and optimizing the model loss value, the model can continuously adjust its parameters during the training process to improve the prediction ability of the electrical equipment noise annoyance.

[0097] In one embodiment, based on the initial electrical equipment noise annoyance prediction model and the validation data set, the model hyperparameters of the initial electrical equipment noise annoyance prediction model are trained by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model, including:

[0098] Initialize the candidate model hyperparameters of the initial electrical equipment noise annoyance prediction model, the ant colony initial pheromones of each path, the maximum number of iterations and the global error; obtain the transition probability of the current candidate model hyperparameter to the next candidate model hyperparameter according to the pheromone concentration of the ant colony initial pheromones of each path, the node selection probability influence function, the influence factor and the expected influence factor, and obtain the model hyperparameter combination of each path based on the transition probability; input the subjective evaluation data and the objective evaluation data of the verification data set into the initial electrical equipment noise annoyance prediction model to obtain the predicted noise annoyance of the simulated noise data; when the difference between the predicted noise annoyance and the actual noise annoyance is not greater than the global error, update the ant colony initial pheromones of each path, and use the current model hyperparameter combination of each path as the current optimal solution, and use the current model hyperparameter combination of each path as the new model hyperparameter combination of each path, repeat the above process until the number of iterations reaches the maximum number of iterations, and determine the initial electrical equipment noise annoyance prediction model at this time as the target electrical equipment noise annoyance prediction model.

[0099] Among them, the candidate model hyperparameters can be understood as dividing all weights and thresholds into n equal parts within a certain range of values; pheromone can be understood as a key component used to guide the search process, which is a virtual two used to simulate the chemical substances released by ants in the process of searching for food, and is used to transmit information and influence the path selection of other ants. The initial pheromone content in each path is equal; the global error can be understood as the gap or deviation between the current optimal solution found by the algorithm and the optimal solution of the problem in the process of finding the optimal solution.

[0100] For example, the server 106 initializes the candidate model hyperparameters of the initial electrical equipment noise annoyance prediction model, the ant colony initial pheromones of each path, the maximum number of iterations and the global error, and then obtains the transition probability of the current candidate model hyperparameter to the next candidate model hyperparameter based on the pheromone concentration of the ant colony initial pheromones of each path, the node selection probability influence function, the influence factor and the expected influence factor, and obtains the model hyperparameter combination of each path based on the transition probability, and then inputs the subjective evaluation data and the objective evaluation data of the verification data set into the initial electrical equipment noise annoyance prediction model to obtain the predicted noise annoyance of the simulated noise data, and when the difference between the predicted noise annoyance and the actual noise annoyance is not greater than the global error, the ant colony initial pheromones of each path are updated, and the current model hyperparameter combination of each path is used as the current optimal solution, and the current model hypercombination of each path is used as the new model hyperparameter combination of each path, and the above process is repeated until the number of iterations reaches the maximum number of iterations, and the initial electrical equipment noise annoyance prediction model at this time is confirmed as the target electrical equipment noise annoyance prediction model. This process optimizes the hyperparameters of the electrical equipment noise annoyance prediction model based on the ant colony algorithm, ensures the continuous improvement of the model performance through dynamic pheromone update and error control mechanism, and finally determines the optimal target model. This method can effectively improve the prediction accuracy of the model.

[0101] In an exemplary embodiment, obtaining simulated noise data of a sample electrical device includes:

[0102] The current electric field force of the sample electrical equipment is obtained, and the current electric field force is set in a pre-built finite element model of the sample electrical equipment to obtain the surface vibration acceleration of the sample electrical equipment; the surface vibration acceleration is set as a simulation boundary condition in the pre-built simulation model of the sample electrical equipment, and an acoustic field simulation is performed in the simulation model to obtain the simulated noise data of the sample electrical equipment.

[0103] Among them, surface vibration acceleration can be understood as the rate of change of velocity per unit time. When an object vibrates, its surface will undergo periodic displacements, and its acceleration can be used to describe the speed and changing pattern of these displacements.

[0104] Optionally, the server 106 obtains the current electric field force of the sample electrical device 102, and sets the current electric field force in the pre-built finite element model of the sample electrical device 102 to obtain the surface vibration acceleration of the sample electrical device 102. Finally, the surface vibration acceleration is set as a simulation boundary condition in the pre-built simulation model of the sample electrical device, and the sound field simulation is performed in the simulation model to obtain the simulated noise data of the sample electrical device. By combining the electric field force with the surface vibration acceleration, the dynamic behavior of the electrical device in actual work can be simulated more accurately, providing more reliable analysis results, and effectively improving the accuracy of the sound field simulation.

[0105] In one embodiment, the subjective evaluation data and the objective evaluation data are used to construct a training data set, a validation data set, and a test data set, including:

[0106] Outliers in the subjective evaluation data and the objective evaluation data are removed, and missing values ​​in the subjective evaluation data and the objective evaluation data are supplemented to obtain cleaned subjective evaluation data and objective evaluation data; the cleaned subjective evaluation data and the objective evaluation data are standardized to obtain standardized subjective evaluation data and objective evaluation data; and the standardized subjective evaluation data and the objective evaluation data are used to construct a training data set, a verification data set, and a test data set.

[0107] For example, the server 106 removes abnormal values ​​from the subjective evaluation data and the objective evaluation data, and supplements the missing values ​​in the subjective evaluation data and the objective evaluation data to obtain cleaned subjective evaluation data and objective evaluation data, and then performs standardization processing on the cleaned subjective evaluation data and the objective evaluation data - scaling the cleaned subjective evaluation data and the objective evaluation data to the same range, and finally uses the standardized subjective evaluation data and the objective evaluation data to construct a training data set, a verification data set, and a test data set. By preprocessing the subjective evaluation data and the objective evaluation data, and scaling the numerical range, the usability of the divided data set is enhanced, thereby accelerating the training process of the electrical equipment noise annoyance prediction model.

[0108] In an exemplary embodiment, as shown in FIG3 , a specific implementation of a method for training a prediction model for noise annoyance of electric equipment is provided. The data mentioned below are all examples, and the present embodiment can be implemented in this case, including:

[0109] 1. Simulate and obtain sound samples of electrical equipment:

[0110] Simulate and obtain sound samples of electrical equipment, including:

[0111] (1) Taking the capacitor, a typical equipment in the converter station, as an example, ANSYS is selected as the finite element analysis software. A series of simplified assumptions are made on the capacitor model, and a finite element model of the capacitor is established based on the actual geometric shape and structure. The finite element model of the capacitor consists of three parts: the shell, the capacitor element, and the insulating porcelain column. Each part is assigned corresponding material properties.

[0112] (2) According to the typical working conditions of the capacitor, the magnitude and direction of the electric field force are calculated, the corresponding force is applied to it in ANSYS, and a fixed constraint is applied to the capacitor's fixed mounting ears. Finally, a transient response analysis is performed on the capacitor model to obtain the surface vibration acceleration.

[0113] (3) Perform acoustic field simulation in COMSOL Multiphysics, establish a capacitor simulation model, select and arrange measurement points, and use the surface vibration acceleration obtained above as the simulation boundary condition to obtain noise samples at each measurement point. The measurement points are selected as shown in Figures 4 and 5, and a total of 20 measurement points are arranged at a distance of 20 cm from the capacitor surface.

[0114] 2. Conduct subjective and objective evaluation tests on the samples obtained, including:

[0115] (1) When conducting a subjective evaluation test, the degree of annoyance is selected as a subjective evaluation indicator, which can reflect the degree of annoyance caused by the noise generated by electrical equipment. Further, the number and type of evaluation subjects are determined, and the degree of annoyance of the test subjects is statistically analyzed based on the questionnaire survey and the semantic segmentation method. For test subjects who have no listening experience, they need to receive appropriate listening training before conducting a subjective evaluation test. Among them, the semantic segmentation method means that each evaluation subject uses semantically opposite adjectives to evaluate a certain attribute of the sound, and each pair of adjectives is divided into multiple levels, and each level is assigned a certain score for subsequent data statistical analysis.

[0116] (2) When conducting an objective evaluation test, first select basic objective parameters based on the frequency range and properties of the noise generated by the electrical equipment under study, and consider whether to add objective parameters such as loudness, pitch, jitter, roughness, and sharpness on the basis of sound pressure level and A-level. Before extracting objective sound parameters, perform FFT transformation on the obtained noise samples, and extract relevant objective parameters by taking integer multiple frequency points within the main frequency distribution range.

[0117] 3. Conduct correlation analysis on subjective evaluation results and objective factors, and retain factors with strong correlation, including:

[0118] Correlation analysis was performed on the subjective evaluation results, objective evaluation results, and subjective and objective evaluation results, and factors with strong correlation were retained, including: using SPSS (Statistical Package for the Social Sciences) analysis software to perform correlation analysis on the subjective evaluation test results and the selected objective factors, where the correlation analysis of the subjective evaluation results is a correlation analysis between the selected subjective evaluation subjects, and the Spearman correlation coefficient is selected to express the correlation between subjects, so as to assess the consistency and rationality of the subjective evaluation results; when performing correlation analysis on objective parameters, the Pearson correlation coefficient is selected for analysis, and whether there is redundancy in the selection of objective parameters is assessed based on the correlation analysis results, and unnecessary relevant objective parameters are eliminated; correlation analysis is performed between subjective and objective evaluation results, and linear correlation coefficients are selected, and objective parameters with correlation coefficients less than 0.5 are further eliminated, indicating that these objective parameters are not closely related to the subjective annoyance and are insignificant. Including the following formulas:

[0119]

[0120]

[0121] 4. Establish a neural network annoyance prediction model based on ant colony optimization algorithm, including:

[0122] Establishing a neural network annoyance prediction model based on the ant colony optimization algorithm includes: firstly, determining the structure of the neural network model, and processing missing values ​​and removing outliers for the subjective and objective evaluation samples obtained above, further standardizing the samples, and scaling the data to the same range to accelerate training; dividing the samples into training set, test set and validation set; determining the number of variables and neurons in each layer, the input layer is all the subjective and objective parameters selected in the above steps, the number of hidden layers is determined, and the result error and operation time corresponding to different numbers of hidden layers are adjusted and finally determined in combination with experience; selecting Xavier initialization (Xavier Initialization, also known as Glorot initialization, is a method for initializing neural network weights) to set parameters for initialization; selecting binary cross entropy as the loss function, selecting Adam (Adaptive Moment Estimation) optimizer, further training the model and outputting the results, and the output layer is the evaluation result of the subjective annoyance.

[0123] In order to obtain more applicable network initial weights and thresholds, thereby improving the performance of the BP neural network model and optimizing the prediction results, the present application proposes a flow chart based on an ant colony optimization algorithm, as shown in FIG6 , which may specifically include the following steps:

[0124] Parameter initialization, all weights and thresholds are divided into n equal parts, the initial pheromone of the ant colony, the maximum number of iterations, and the global error are all set to constants based on experience. Among them, pheromone is a key component used to guide the search process, used to simulate the virtual amount of chemicals released by ants in the process of searching for food, used to transmit information and influence the path selection of other ants, and the content of pheromone in each path is equal; the maximum number of iterations refers to the maximum number of iterations that can be achieved in the process of finding the optimal parameters; the global error refers to the gap or deviation between the current optimal solution found by the algorithm and the optimal solution of the problem in the process of finding the optimal solution.

[0125] The transfer probability in path optimization is calculated, and the global optimal solution is further calculated based on this. If the optimal solution obtained by the ant colony after continuous iteration meets the set error limit, the pheromone is updated, and the movement direction of the ant colony is selected according to the amount of pheromone; after continuous iteration, when the maximum number of iterations is reached, the iteration process ends and the final solution is output. The optimal solution is used as the initial hyperparameter value of the BP neural network, including the following formula:

[0126]

[0127]

[0128]

[0129] in, Indicates pheromone concentration; Represents the influence function of pheromone on node selection probability; and are the impact factor and expected impact factor of pheromone respectively; For the The ant is in the path-finding cycle at the node The amount of pheromones left on the body; is the value of the pheromone's effectiveness coefficient.

[0130] 5. Conduct correlation analysis on the prediction and test results of the annoyance model to verify the rationality of the model, including:

[0131] A correlation analysis was performed on the prediction and test results of the annoyance model to verify the rationality of the model, including: inputting the test set divided in the previous steps into the established improved neural network model for prediction, and performing a correlation analysis on the predicted results and the actual subjective annoyance of the samples. The two were linearly positively correlated with a large correlation coefficient. The results showed that the model has good generalization and learning capabilities.

[0132] Compared with the prior art, this application has the following advantages:

[0133] 1. Through spectrum analysis and corresponding data extraction, the understanding and research of simulated noise characteristics are improved. Secondly, the subjective evaluation data and objective evaluation data are combined to provide multi-dimensional evaluation results to better support decision-making and analysis.

[0134] 2. Based on the ant colony algorithm, the hyperparameters of the electrical equipment noise annoyance prediction model are optimized, and the dynamic pheromone update and error control mechanism are used to ensure the continuous improvement of the model performance and finally determine the optimal target model. This method can effectively improve the prediction accuracy of the model.

[0135] 3. By preprocessing the subjective and objective evaluation data and scaling the numerical range, the usability of the divided data set is enhanced, thereby accelerating the training process of the electrical equipment noise annoyance prediction model.

[0136] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0137] Based on the same inventive concept, the embodiment of the present application also provides an electrical equipment noise annoyance prediction model training device for implementing the electrical equipment noise annoyance prediction model training method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more electrical equipment noise annoyance prediction model training device embodiments provided below can refer to the limitations of the electrical equipment noise annoyance prediction model training method above, and will not be repeated here.

[0138] In an exemplary embodiment, as shown in FIG7 , a device for training a prediction model of noise annoyance of electrical equipment is provided, comprising: a data acquisition module 701, a data set construction module 702, a parameter training module 703 and a performance testing module 704, wherein:

[0139] The data acquisition module 701 is used to acquire simulated noise data of sample electrical equipment and acquire subjective evaluation data corresponding to the simulated noise data;

[0140] A data set construction module 702 is used to obtain objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data, and to construct a training data set, a verification data set, and a test data set using the subjective evaluation data and the objective evaluation data;

[0141] The parameter training module 703 is used to train the model parameters of the electrical equipment noise annoyance prediction model to be trained using the training data set to obtain the initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, train the model hyperparameters of the initial electrical equipment noise annoyance prediction model using the ant colony optimization algorithm to obtain the target electrical equipment noise annoyance prediction model;

[0142] The performance testing module 704 is used to input the test data set into the target electrical equipment noise annoyance prediction model, and perform a performance test on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

[0143] In one embodiment, the data set construction module 702 is also used to perform FFT transformation on the simulated noise data to obtain a simulated noise spectrum, and obtain original objective evaluation data associated with the simulated noise data based on the simulated noise spectrum; perform correlation analysis on the original objective evaluation data through SPSS to obtain the Pearson correlation coefficient associated with the original objective evaluation data; remove redundant objective evaluation data in the original objective evaluation data based on the Pearson correlation coefficient to obtain initial objective evaluation data; and obtain objective evaluation data based on the initial objective evaluation data and the subjective evaluation data.

[0144] In one embodiment, the data set construction 702 is further used to perform correlation analysis on the initial objective evaluation data and the subjective evaluation data using SPSS to obtain the linear correlation coefficient between the initial objective evaluation data and the subjective evaluation data; remove the initial objective evaluation data whose linear correlation coefficient is less than a preset linear correlation coefficient threshold to obtain the objective evaluation data.

[0145] In one embodiment, the data acquisition module 701 is also used to obtain adjective groups of the sample users' annoyance levels for the simulated noise data; score the annoyance level of each simulated noise data based on each annoyance adjective group to obtain multiple original subjective evaluation data; perform correlation analysis on the original subjective evaluation data through SPSS to obtain the Spearman correlation coefficient associated with the original subjective evaluation data; remove the original subjective evaluation data whose Spearman correlation coefficient is less than a preset Spearman correlation coefficient threshold to obtain the subjective evaluation data.

[0146] In an exemplary embodiment, the simulated noise data is associated with the actual noise annoyance, and the parameter training module 703 is also used to use Xavier initialization to initialize the model parameters of the original electrical equipment noise annoyance prediction model to obtain the electrical equipment noise annoyance prediction model to be trained; the subjective evaluation data and the objective evaluation data in the training data set are input into the electrical equipment noise annoyance prediction model to be trained to obtain the predicted noise annoyance corresponding to the simulated noise data; based on the difference between the predicted noise annoyance and the actual noise annoyance, and the loss function pre-set for the electrical equipment noise annoyance prediction model to be trained, the corresponding model loss value is obtained; the model parameters are updated according to the model loss value until the loss function converges, and the electrical equipment noise annoyance prediction model to be trained is used as the initial electrical equipment noise annoyance prediction model.

[0147] In an exemplary embodiment, the parameter training module 703 is also used to initialize the candidate model hyperparameters of the initial electrical equipment noise annoyance prediction model, the ant colony initial pheromones of each path, the maximum number of iterations and the global error; according to the pheromone concentration of the ant colony initial pheromones of each path, the node selection probability influence function, the influence factor and the expected influence factor, the transition probability of the current candidate model hyperparameter to the next candidate model hyperparameter is obtained, and the model hyperparameter combination of each path is obtained based on the transition probability; the subjective evaluation data and the objective evaluation data of the verification data set are input into the initial electrical equipment noise annoyance prediction model to obtain the predicted noise annoyance of the simulated noise data; when the difference between the predicted noise annoyance and the actual noise annoyance is not greater than the global error, the ant colony initial pheromones of each path are updated, and the current model hyperparameter combination of each path is used as the current optimal solution, and the current model hyperparameter combination of each path is used as the new model hyperparameter combination of each path, and the above process is repeated until the number of iterations reaches the maximum number of iterations, and the initial electrical equipment noise annoyance prediction model at this time is determined as the target electrical equipment noise annoyance prediction model.

[0148] In one embodiment, the data acquisition module 701 also includes: a simulation unit, which is used to obtain the current electric field force of the sample electrical device, and set the current electric field force in a pre-constructed finite element model of the sample electrical device to obtain the surface vibration acceleration of the sample electrical device; set the surface vibration acceleration as a simulation boundary condition in the pre-constructed simulation model of the sample electrical device, and perform sound field simulation in the simulation model to obtain simulated noise data of the sample electrical device.

[0149] In an exemplary embodiment, the data set construction module 702 also includes: a data processing unit, which is used to remove outliers in the objective evaluation data, subjective evaluation data, and objective evaluation data, and to supplement the missing values ​​in the objective evaluation data, subjective evaluation data, and objective evaluation data, to obtain cleaned subjective evaluation data and objective evaluation data; standardize the subjective evaluation data and objective evaluation data after the objective evaluation data is cleaned to obtain standardized subjective evaluation data and objective evaluation data; and use the subjective evaluation data and objective evaluation data after the objective evaluation data is standardized to construct a training data set, a verification data set, and a test data set.

[0150] Each module in the above-mentioned electrical equipment noise annoyance prediction model training device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0151] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG8. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store simulated noise data, subjective evaluation data and objective evaluation data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for training a noise annoyance prediction model for electrical equipment is implemented.

[0152] Those skilled in the art will understand that the structure shown in FIG. 8 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0153] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the method for training a model for predicting noise annoyance of electrical equipment of the above embodiment is implemented.

[0154] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the electrical equipment noise annoyance prediction model training method of the above embodiment is implemented.

[0155] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the electrical equipment noise annoyance prediction model training method of the above embodiment.

[0156] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0157] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0158] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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, they should be considered to be within the scope of this application.

[0159] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for training a prediction model for noise annoyance of electrical equipment, characterized in that: The method comprises: Acquiring simulated noise data of sample electrical equipment, and acquiring subjective evaluation data corresponding to the simulated noise data; Based on the simulated noise data and the subjective evaluation data, obtaining objective evaluation data associated with the simulated noise data, and constructing a training data set, a validation data set, and a test data set using the subjective evaluation data and the objective evaluation data; Using the training data set, training the model parameters of the electrical equipment noise annoyance prediction model to be trained to obtain an initial electrical equipment noise annoyance prediction model, and based on the initial electrical equipment noise annoyance prediction model and the verification data set, training the model hyperparameters of the initial electrical equipment noise annoyance prediction model by an ant colony optimization algorithm to obtain a target electrical equipment noise annoyance prediction model; The test data set is input into the target electrical equipment noise annoyance prediction model, and a performance test is performed on the target electrical equipment noise annoyance prediction model. If the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

2. The method according to claim 1, characterized in that The step of acquiring objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data includes: Performing FFT transformation on the simulated noise data to obtain a simulated noise spectrum, and acquiring original objective evaluation data associated with the simulated noise data based on the simulated noise spectrum; Performing correlation analysis on the original objective evaluation data by SPSS to obtain a Pearson correlation coefficient associated with the original objective evaluation data; Based on the Pearson correlation coefficient, redundant objective evaluation data in the original objective evaluation data are removed to obtain initial objective evaluation data; The objective evaluation data is acquired according to the initial objective evaluation data and the subjective evaluation data.

3. The method according to claim 2, characterized in that The step of obtaining the objective evaluation data according to the initial objective evaluation data and the subjective evaluation data includes: Using the SPSS to perform correlation analysis on the initial objective evaluation data and the subjective evaluation data to obtain a linear correlation coefficient between the initial objective evaluation data and the subjective evaluation data; The objective evaluation data is obtained by removing the initial objective evaluation data whose linear correlation coefficient is less than a preset linear correlation coefficient threshold.

4. The method according to claim 3, characterized in that The obtaining of subjective evaluation data corresponding to the simulated noise data includes: Obtaining an adjective group of annoyance levels of sample users for the simulated noise data; Scoring the annoyance level of each of the simulated noise data based on each of the annoyance level adjective groups to obtain a plurality of original subjective evaluation data; Performing correlation analysis on the original subjective evaluation data by using the SPSS to obtain a Spearman correlation coefficient associated with the original subjective evaluation data; The original subjective evaluation data whose Spearman correlation coefficient is less than a preset Spearman correlation coefficient threshold is removed to obtain the subjective evaluation data.

5. The method according to claim 1, characterized in that The simulated noise data is associated with an actual noise annoyance level; The method of using the training data set to train the model parameters of the electrical equipment noise annoyance prediction model to be trained to obtain an initial electrical equipment noise annoyance prediction model includes: Using Xavier initialization to initialize the model parameters of the original electrical equipment noise annoyance prediction model to obtain the electrical equipment noise annoyance prediction model to be trained; Inputting the subjective evaluation data and the objective evaluation data in the training data set into the noise annoyance prediction model for the electrical equipment to be trained to obtain a predicted noise annoyance corresponding to the simulated noise data; Based on the difference between the predicted noise annoyance level and the actual noise annoyance level, and the loss function preset for the noise annoyance level prediction model of the electrical equipment to be trained, a corresponding model loss value is obtained; The model parameters are updated according to the model loss value until the loss function converges, and the electrical equipment noise annoyance prediction model to be trained is used as the initial electrical equipment noise annoyance prediction model.

6. The method according to claim 5, characterized in that The method of training the model hyperparameters of the initial electrical equipment noise annoyance prediction model by using an ant colony optimization algorithm based on the initial electrical equipment noise annoyance prediction model and the verification data set to obtain a target electrical equipment noise annoyance prediction model includes: Initializing candidate model hyperparameters of the initial electrical equipment noise annoyance prediction model, ant colony initial pheromones of each path, maximum number of iterations and global error; According to the pheromone concentration of the initial pheromone of the ant colony, the node selection probability influence function, the influence factor and the expected influence factor of each path, the transition probability of the current candidate model hyperparameter to the next candidate model hyperparameter is obtained, and the model hyperparameter combination of each path is obtained based on the transition probability; Inputting the subjective evaluation data and the objective evaluation data of the verification data set into the initial electrical equipment noise annoyance prediction model to obtain the predicted noise annoyance of the simulated noise data; When the difference between the predicted noise annoyance and the actual noise annoyance is not greater than the global error, the initial pheromone of the ant colony of each path is updated, and the current model hyperparameter combination of each path is used as the current optimal solution, and the current model hyperparameter combination of each path is used as the new model hyperparameter combination of each path. The above process is repeated until the number of iterations reaches the maximum number of iterations, and the initial electrical equipment noise annoyance prediction model at this time is determined as the target electrical equipment noise annoyance prediction model.

7. The method according to claim 1, characterized in that The step of obtaining simulated noise data of a sample electrical device comprises: Acquiring a current electric field force of the sample electrical device, and setting the current electric field force in a pre-constructed finite element model of the sample electrical device to obtain a surface vibration acceleration of the sample electrical device; The surface vibration acceleration is set as a simulation boundary condition in a pre-built simulation model of a sample electrical device, and a sound field simulation is performed in the simulation model to obtain simulated noise data of the sample electrical device.

8. The method according to claim 1, characterized in that The step of using the subjective evaluation data and the objective evaluation data to construct a training data set, a verification data set, and a test data set includes: Removing outliers from the subjective evaluation data and the objective evaluation data, and supplementing missing values ​​from the subjective evaluation data and the objective evaluation data, to obtain cleaned subjective evaluation data and objective evaluation data; Performing standardization processing on the cleaned subjective evaluation data and objective evaluation data to obtain standardized subjective evaluation data and objective evaluation data; The standardized subjective evaluation data and objective evaluation data are used to construct a training data set, a validation data set, and a test data set.

9. A device for training a prediction model of noise annoyance of electrical equipment, characterized in that: The device comprises: A data acquisition module, used to acquire simulated noise data of sample electrical equipment, and to acquire subjective evaluation data corresponding to the simulated noise data; A data set construction module, used to obtain objective evaluation data associated with the simulated noise data based on the simulated noise data and the subjective evaluation data, and to construct a training data set, a verification data set and a test data set using the subjective evaluation data and the objective evaluation data; a parameter training module, for training the model parameters of the electrical equipment noise annoyance prediction model to be trained using the training data set to obtain an initial electrical equipment noise annoyance prediction model, and for training the model hyperparameters of the initial electrical equipment noise annoyance prediction model using an ant colony optimization algorithm based on the initial electrical equipment noise annoyance prediction model and the verification data set to obtain a target electrical equipment noise annoyance prediction model; A performance testing module is used to input the test data set into the target electrical equipment noise annoyance prediction model, perform a performance test on the target electrical equipment noise annoyance prediction model, and if the result of the performance test indicates that the target electrical equipment noise annoyance prediction model passes the performance test, the target electrical equipment noise annoyance prediction model is used as the trained electrical equipment noise annoyance prediction model.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images