A method and system for dynamic sound screening alarm of power transmission lines

Abstract

The present invention discloses a method and system for dynamic screening of transmission line sound alarms, the method comprising: collecting audio data within a set range of the transmission line, using a trained transmission line sound pre-classification model to predict whether the audio data is a sound type of concern; for audio data predicted to be a sound type of concern, judging whether it is an abnormal sound by strong transformation detection; wherein the threshold value when performing strong transformation detection is determined according to the ratio of the number of audio data predicted to be a sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds a set value, the threshold value is updated to enter the next counting round. The present invention realizes the update of the threshold value by dynamically updating the alarm excitation coefficient, thereby reasonably regulating the alarm situation and avoiding repeated alarms.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line monitoring, and in particular to a method and system for dynamic sound screening and alarming of power transmission lines. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Transmission lines are important carriers of electric energy transmission. In order to ensure the safe and reliable operation of transmission lines, abnormal sound signals during the operation of transmission lines (such as bird calls, thunder, blasting sounds, construction machinery sounds, etc.) are important parameters that need to be monitored.

[0004] At present, there are mainly two methods for detecting abnormal sound of transmission lines:

[0005] (1) Matching method: By establishing a standard abnormal sound library, the collected sound is then traversed and matched with the database, and the judgment result is output based on the comparison between the matching degree and the set threshold. However, as the standard library increases, the computational cost of this method will also increase, affecting the timeliness of detection.

[0006] (2) Model prediction method: First, construct an abnormal sound dataset, then use the dataset to train an abnormal sound detection model, and then predict the collected sound. However, the prediction accuracy of this method is affected by the amount of data in the abnormal sound dataset, and the existing method cannot dynamically adjust the alarm threshold according to the actual situation. When an abnormal sound appears and persists (such as continuous bird calls), repeated alarms may occur, resulting in reduced alarm effectiveness. Summary of the invention

[0007] In order to solve the above problems, the present invention proposes a method and system for dynamic screening alarm of transmission line sound, which dynamically updates the threshold of strong transformation detection through the alarm excitation coefficient, and can reasonably regulate the alarm situation of abnormal sound of the transmission line according to the real-time situation around the transmission channel, thereby effectively eliminating redundant alarms.

[0008] In some embodiments, the following technical solutions are adopted:

[0009] A transmission line sound dynamic screening alarm method, comprising:

[0010] Collect audio data within a set range of the transmission line, and use the trained transmission line sound pre-classification model to predict whether the audio data is a sound type of concern; for audio data predicted to be a sound type of concern, determine whether it is an abnormal sound through strong transformation detection;

[0011] Among them, the threshold value for strong change detection is determined according to the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds the set value, the threshold is updated and the next counting round is entered.

[0012] As a specific implementation, each threshold involved in strong transformation detection is the product of the threshold initial value and the alarm excitation coefficient; the alarm excitation coefficient is determined based on the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds.

[0013] In other embodiments, the following technical solutions are adopted:

[0014] A transmission line sound dynamic screening alarm system, comprising:

[0015] An abnormal sound pre-screening module is configured to collect audio data within a set range of the power transmission line, and use a trained transmission line sound pre-classification model to predict whether the audio data is a sound type of concern;

[0016] An abnormal sound judgment module is configured to judge whether the audio data predicted as the type of sound of interest is an abnormal sound through strong transformation detection;

[0017] Among them, the threshold value for strong change detection is determined according to the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds the set value, the threshold is updated and the next counting round is entered.

[0018] In other embodiments, the following technical solutions are adopted:

[0019] A terminal device comprises a processor and a memory, wherein the processor is used to implement instructions; the memory is used to store a plurality of instructions, wherein the instructions are suitable for being loaded by the processor and executing the above-mentioned abnormal sound dynamic screening alarm method for power transmission lines.

[0020] In other embodiments, the following technical solutions are adopted:

[0021] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor of a terminal device and executing the above-mentioned method for dynamically screening and warning abnormal sounds of a power transmission line.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) The present invention combines a neural network prediction method with a strong transformation detection method. The strong transformation detection method can dynamically control the number of alarms and filter out duplicate and weak alarms. The neural network prediction method can improve the prediction accuracy. The combined application of the two methods can be deployed at the front end, and the number of alarms can be controlled while achieving accurate front-end alarms, effectively eliminating redundant and repeated alarms.

[0024] (2) The present invention calculates an alarm excitation coefficient based on the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds, and calculates the threshold involved in the strong transformation detection method based on the alarm excitation coefficient; the threshold is updated by dynamically updating the alarm excitation coefficient, so that the alarm situation can be reasonably adjusted according to the real-time situation around the transmission channel to avoid repeated and weak alarms.

[0025] (3) The method of the present invention can reduce the number of alarms, improve the effectiveness of alarms, save the heavy work of manual inspection, and improve work efficiency. At the same time, it can eliminate interference / abnormal sounds in the transmission line channel environment at the front end, saving the computing power required for back-end analysis.

[0026] Other features and advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic diagram of the structure of a time-delay neural network in an embodiment of the present invention;

[0028] Figure 2 Schematic diagram of a strong transformation detection process in an embodiment of the present invention;

[0029] Figure 3 The present invention is a flowchart of a method for dynamically screening and warning abnormal sounds of power transmission lines in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0031] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0032] It should be noted that the abnormal sounds of the power transmission line in this embodiment at least include: bird calls, construction machinery sounds, blasting sounds and thunder, etc., which are respectively used as the types of hidden danger sounds of concern. Of course, those skilled in the art can also set the types of abnormal sounds of the power transmission line according to actual needs.

[0033] Embodiment 1

[0034] In one or more embodiments, a method for dynamic screening and alarming of transmission line sounds is disclosed, which combines strong change detection with environmental sound classification to jointly achieve the screening and appropriate alarm functions of hidden danger sounds around the transmission channel.

[0035] This embodiment divides the types of hidden danger sounds around the transmission channel according to the hidden danger standards, obtains a transmission line sound pre-classification model based on a time-delay neural network through multiple rounds of training optimization, and then uses the model to predict the collected sounds. For the audio determined to be the type of hidden danger sound of concern in the transmission channel, strong transformation detection is performed. The strong transformation detection uses three features, sound pressure level, frame maximum energy, and frequency domain average energy, as judgment indicators. The audio is finally determined to be a hidden danger sound and an alarm is issued, and at the same time, it is passed to other downstream tasks such as back-end analysis. The neural network prediction method and the strong transformation detection method of this embodiment form a joint processing mechanism through the proposed alarm excitation coefficient, and dynamically adjusts the detection threshold of the screening alarm based on the alarm excitation coefficient, thereby ensuring the effectiveness of the transmission channel scene alarm on the basis of accurate classification and achieving appropriate alarms.

[0036] Specifically, combined Figure 3 The method of this embodiment specifically includes the following process:

[0037] (1) Collect audio data within the set range of the transmission line, extract the Mel-spectrogram features, and input them into the trained transmission line sound pre-classification model to predict whether the audio data is the sound type of interest.

[0038] In this embodiment, an audio database is first constructed, in which audio data files with sound type labels are stored; for example, if an audio data file contains both bird calls and the sounds of construction machinery, then its sound type labels are bird calls and construction machinery; as a specific example, the audio duration of each audio data file is usually around 10 seconds.

[0039] This embodiment uses a strict scoring mechanism to select and classify these audio data files.

[0040] Specifically, two rules for distinguishing normal sounds from abnormal sounds are provided:

[0041] ① Loose distinction rules between abnormal and normal data: According to the audio data file and its type label, any audio data with a label of 1 in the hidden danger sound type of concern for the transmission line and all other types not concerned are not 1 are placed in the abnormal category, and the rest are placed in the normal category. That is, when any one or more of the hidden danger sound types of concern exist in a certain audio data, and there is no hidden danger sound type that is not concerned, then the audio data is classified into the abnormal category, and other audio data that do not meet the above conditions are all classified into the normal category.

[0042] ② Strict distinction rules between abnormal and normal data: According to the audio data file and its type label, each individual hidden danger sound type of concern for the transmission line is labeled as 1, and the rest of the audio data that are not 1 are placed in the abnormal category, and the rest are all placed in the normal category. That is: when only one hidden danger sound type of concern exists in a certain audio data, and the other hidden danger sound types of concern and the hidden danger sound types of no concern do not exist, then the audio data is classified into the abnormal category, and other audio data that do not meet the above conditions are all classified into the normal category. This method can select various hidden danger sound types of concern separately, avoiding mixed intersections between different sound categories, which affects the model prediction performance.

[0043] Since the amount of data in the audio database of this embodiment is still insufficient for the training requirements of the neural network model, some data enhancements are performed on the basis of the data set in the audio database, including random volume perturbation, in which each audio is randomly selected between -15dB and 15dB to generate a volume perturbation decibel value for corresponding adjustment; random speed perturbation, in which the speed perturbation value ranges from 0.9 to 1.1, and the audio is randomly perturbed by a perturbation value within the range, and the audio speed becomes faster or slower according to the adjustment; spectrum enhancement, in which the extracted acoustic features are randomly masked from the two dimensions of time and frequency, which increases the difficulty of the data and improves the generalization of the model; random displacement perturbation, in which the audio is randomly moved forward or backward by a few milliseconds along the time axis, and the movement value range is set between -5ms and 5ms; data resampling enhancement, in which the audio data is resampled according to several set sampling rates to improve the sampling diversity of the audio data, and the specific sampling rates include 8kHz, 32kHz, 44.1kHz, and 48kHz.

[0044] As an example, the transmission line sound pre-classification model of this embodiment uses a time-delay neural network (TDNN); TDNN is composed of a convolutional layer, a batch normalization layer, a pooling layer, and a fully connected layer. The Relu activation function is used to add nonlinear characteristics to the model to ensure that the model has sufficient feature expression capabilities. TDNN is different from the model structure that only uses one frame of features. It is a neural network that can contain multiple frames. In this way, the time-delay neural network is equivalent to delaying the weights. If this method is used between the hidden layer and the output layer, the overall weight is greatly reduced, which is convenient for training. If you want to extract a feature expression with 5 frames of context for a total of 11 frames, generally speaking, the 11 frames of features will be directly spliced ​​together to form a 11*D (D is the feature dimension of a frame) feature, and then the model is allowed to learn the 11*D feature map. The TDNN processing method is different. TDNN will process a narrower temporal context than 11 frames in the initial layer, and then extract the 11 frames of feature context features by sending them to a deeper network; the specific schematic diagram is as follows Figure 1 shown.

[0045] Of course, those skilled in the art may also select other neural network types as needed, such as BP neural network, convolutional neural network, etc.

[0046] This embodiment uses the constructed audio database to train the neural network model, and uses the trained neural network model for training. The input of the neural network model is the Mel-spectrogram feature of the audio data, and the output is whether the audio data is the sound type of concern.

[0047] Among them, the extraction of the Mel spectrum graph first pre-emphasizes the loaded audio. The high-frequency component of the audio signal has a small intensity, while the low-frequency component has a large intensity. Signal pre-emphasis is to let the signal pass through a high-pass filter so that the intensity of the high-frequency and low-frequency components of the signal does not differ too much, in order to compensate for the excessive attenuation of the high-frequency part of the channel. Pre-emphasis generally uses a first-order high-pass filter:

[0048] H(z)=1-μz -1

[0049] Among them, μ is the pre-emphasis coefficient, and its general value range is 0.9<α<1.0, and is usually 0.97.

[0050] Next, frame division and windowing are performed. After the signal is preprocessed, since the audio signal has the short-term stable characteristics and in order to obtain the relationship between the frequency and time of the audio signal through Fourier transform, the signal needs to be divided into several small segments according to time. A segment is called a frame. The frame length is usually 20 to 40ms. After comprehensive consideration, 25ms is selected and the frame shift is 15ms. There is an overlap of 10ms between the two frames to ensure a smooth transition between the two frames to prevent information leakage. The signal processing method requires that the signal is continuous, but it is disconnected in the middle during frame division. In order to meet the conditions, the divided frame data needs to be multiplied by a piece of data of the same length. This piece of data is the data within the entire period of the window function, changing from the minimum to the maximum, and then to the minimum. This operation is called windowing. On the one hand, it makes the global more continuous and avoids the Gibbs effect. On the other hand, it makes the originally non-periodic speech signal show some characteristics of the periodic function, which is convenient for subsequent frequency domain analysis.

[0051] Then, the power spectrum is obtained, and Fourier transform is performed on each frame. The amplitude of the obtained Fourier transform is taken, squared, and then divided by the set number of Fourier points to obtain the power spectrum.

[0052] Finally, the above data is passed through the Mel filter bank. The starting point of each filter in the Mel filter bank is at the midpoint of the previous filter. The number of filters is set to 64 here. In actual operation, each filter is point-multiplied with the power spectrum, and the result is the energy in the frequency band, that is, the Mel spectrum graph.

[0053] In the specific audio data prediction process, the audio data within the set range of the transmission line is collected by a sound collection module arranged on the front-end device of the transmission line.

[0054] The acquired audio data is preprocessed, the mel-spectrogram of the preprocessed audio data is extracted, the mel-spectrogram is input into the trained neural network model, and the output is whether the audio data is the sound type of interest.

[0055] (2) for audio data predicted to be of the sound type of interest, determine whether it is an abnormal sound through strong transformation detection;

[0056] Among them, the threshold value for strong change detection is determined based on the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds the set value, the threshold is updated and the next counting round is entered.

[0057] Since the performance of abnormal sound and sound in stable state is very different, the strong change detection method can detect the sudden abnormal sound in the current stable environment. Figure 2,The specific process of the strong transformation detection method is as follows:

[0058] 1) Obtain the i-th frame data of the audio data, set i=1; perform strong transformation initialization to calculate the sound pressure level (SPL) and frame maximum energy (FME) of the frame data, which are used as the dynamic reference value of the sound pressure level and the dynamic reference value of the frame maximum energy respectively;

[0059] 2) Determine whether the time limit of the audio data has been reached, that is, whether all data frames of the audio data have been determined; if not, obtain the i+1th frame data of the audio data, perform strong transformation initialization to calculate the sound pressure level value and frame maximum energy value of the frame data, and enter 3); if so, the determination ends;

[0060] 3) calculating the difference between the sound pressure level value of the i+1th frame data and the sound pressure level dynamic reference value, recorded as the first difference; and the difference between the frame maximum energy value of the i+1th frame data and the frame maximum energy dynamic reference value, recorded as the second difference;

[0061] If the first difference and the second difference do not exceed the first threshold and the second threshold set respectively, the i+1 frame data is not an abnormal sound, and the sound pressure level value and the frame maximum energy value calculated from the i+1 frame data are updated as dynamic reference values; otherwise i=i+1, return 2);

[0062] If the first difference and the second difference both exceed the first threshold and the second threshold set respectively, the (i+1)th frame data is an abnormal sound; proceed to 4);

[0063] 4) Calculate the frequency domain average energy (FDAE) of the audio data. If the frequency domain average energy exceeds a set third threshold, the audio data is determined to be an abnormal sound; otherwise, it is a normal sound.

[0064] In this embodiment, it is first determined whether the difference between the sound pressure level value and the frame maximum energy value of each frame of audio data and the corresponding dynamic reference value exceeds the respective set thresholds. If so, it is preliminarily determined that there is an abnormal sound in the audio data. As long as there is an abnormal sound, the frequency domain average energy of the audio data is calculated, and based on the size of the frequency domain average energy and the third threshold, it is determined whether the audio data is an abnormal sound.

[0065] In this embodiment, the thresholds for strong change detection include the first threshold, the second threshold and the third threshold mentioned above; these three thresholds are dynamically adjusted by the alarm excitation coefficient; the calculation formula of the alarm excitation coefficient is as follows:

[0066]

[0067] Here, yc represents the number of audio data predicted to be the sound type of interest, and asdyc represents the number of audio data determined to be abnormal sounds.

[0068] During the entire detection process, when the number of alarms reaches the set number (for example, 1000), the alarm excitation coefficient is recalculated. After the alarm excitation coefficient is updated, the number of audio data predicted as the sound type of concern and the number of audio data judged as abnormal sounds will also be reset and re-enter the next counting round.

[0069] The alarm excitation coefficient of this embodiment is mainly used to update the three thresholds of strong change detection. The calculation of the three thresholds is as follows:

[0070] λ SPL1 =γ*λ SPL

[0071] λ FME1 =γ*λ FME

[0072] λ FDAE1 =γ*λ FDAE

[0073] Among them, λ SPL , FME and λ FDAE Respectively represent the initial values ​​of the first threshold, the second threshold and the third threshold; λ SPL1 , FME1 and λ FDAE1 They respectively represent the first threshold, the second threshold and the third threshold updated according to the alarm excitation coefficient γ.

[0074] By dynamically adjusting the three thresholds through the alarm incentive coefficient γ, some repeated alarms can be avoided, thereby controlling the amount of hidden danger alarms and keeping them at an appropriate level. This can not only ensure the rationality and standardization of alarms, but also prevent monitoring staff from suffering a heavy workload due to repeated and frequent alarms.

[0075] Embodiment 2

[0076] In one or more real-time modes, a transmission line sound dynamic screening alarm system is disclosed, comprising:

[0077] An abnormal sound pre-screening module is configured to collect audio data within a set range of the power transmission line, and use a trained transmission line sound pre-classification model to predict whether the audio data is a sound type of concern;

[0078] An abnormal sound judgment module is configured to judge whether the audio data predicted as the type of sound of interest is an abnormal sound through strong transformation detection;

[0079] Among them, the threshold value for strong change detection is determined according to the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds the set value, the threshold is updated and the next counting round is entered.

[0080] The specific implementation of the above modules is the same as that in Example 1 and will not be described in detail.

[0081] Embodiment 3

[0082] In one or more embodiments, a terminal device is disclosed, including a server, the server including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method for dynamic screening and alarming of abnormal sound of power transmission lines in Embodiment 1 when executing the program. For the sake of brevity, it will not be described here.

[0083] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0084] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0085] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software.

[0086] Embodiment 4

[0087] In one or more embodiments, a computer-readable storage medium is disclosed, in which a plurality of instructions are stored, wherein the instructions are suitable for being loaded by a processor of a terminal device and executing the method for dynamically screening and warning abnormal sounds of power transmission lines described in Embodiment 1.

[0088] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for dynamic sound screening and alarming of a power transmission line, characterized in that: include: Collect audio data within a set range of the transmission line, and use the trained transmission line sound pre-classification model to predict whether the audio data is a sound type of concern; for audio data predicted to be a sound type of concern, determine whether it is an abnormal sound through strong transformation detection; Among them, the threshold value for strong change detection is determined according to the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds the set value, the threshold is updated and the next counting round is entered.

2. A method for dynamic sound screening and alarming of a power transmission line according to claim 1, characterized in that: When training the transmission line sound pre-classification model, an audio database is constructed; the audio database stores a plurality of audio data; If the audio data contains only one or more types of sounds of concern, the audio data is classified as abnormal; the rest are classified as normal.

3. A method for dynamic sound screening and alarming of a power transmission line according to claim 1, characterized in that: When training the transmission line sound pre-classification model, an audio database is constructed; the audio database stores a plurality of audio data; If the audio data contains only one type of sound of concern, the audio data is classified as abnormal; the rest are classified as normal.

4. A method for dynamic sound screening and alarming of a power transmission line according to claim 2 or 3, characterized in that: The audio data in the audio database is preprocessed, including: random volume perturbation, random speed perturbation, spectrum enhancement, random displacement perturbation and data resampling enhancement.

5. A method for dynamic sound screening and alarming of a power transmission line according to claim 1, characterized in that: Determine whether it is an abnormal sound through strong transformation detection, specifically: (1) Obtain the i-th frame data of the audio data, and let i=1; Perform strong transformation initialization to calculate the sound pressure level value and the frame maximum energy value of the frame data, which are used as the sound pressure level dynamic reference value and the frame maximum energy dynamic reference value respectively; (2) Determine whether the set time limit has been reached. If not, obtain the i+1th frame of audio data, perform strong transformation initialization to calculate the sound pressure level value and frame maximum energy value of the frame data, and enter (3); if so, the determination ends; (3) calculating the difference between the sound pressure level value of the i+1th frame data and the sound pressure level dynamic reference value, and the difference between the frame maximum energy value of the i+1th frame data and the frame maximum energy dynamic reference value; If at least one of the two differences does not exceed the first threshold and the second threshold set respectively, the i+1 frame data is not an abnormal sound, and the sound pressure level value and the frame maximum energy value calculated from the i+1 frame data are updated as dynamic reference values; otherwise, i=i+1, and return to (2); If both differences exceed the first threshold and the second threshold set respectively, the i+1th frame data is an abnormal sound, and the process goes to (4); (4) Calculate the frequency domain average energy of the audio data. If the frequency domain average energy exceeds a set third threshold, it is determined to be an abnormal sound and an alarm is issued; otherwise, it is a normal sound.

6. A method for dynamic sound screening and alarming of a power transmission line according to claim 1 or 5, characterized in that: The threshold for strong change detection is determined based on the ratio of the number of audio data predicted to be the sound type of interest to the number of audio data determined to be abnormal sound, specifically: Each threshold involved in strong change detection is the product of the threshold initial value and the alarm excitation coefficient; the alarm excitation coefficient is determined based on the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds.

7. A method for dynamic sound screening and alarming of a power transmission line according to claim 6, characterized in that: The alarm excitation coefficient is specifically: Here, yc represents the number of audio data predicted to be the sound type of interest, and asdyc represents the number of audio data determined to be abnormal sounds.

8. A transmission line sound dynamic screening alarm system, characterized in that: include: An abnormal sound pre-screening module is configured to collect audio data within a set range of the power transmission line, and use a trained transmission line sound pre-classification model to predict whether the audio data is a sound type of concern; An abnormal sound judgment module is configured to judge whether the audio data predicted as the type of sound of interest is an abnormal sound through strong transformation detection; Among them, the threshold value for strong change detection is determined according to the ratio of the number of audio data predicted to be the sound type of concern to the number of audio data judged to be abnormal sounds; when the number of audio data judged to be abnormal sounds exceeds the set value, the threshold is updated and the next counting round is entered.

9. A terminal device, comprising a processor and a memory, wherein the processor is used to implement instructions; and the memory is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing the method for dynamic screening and alarming of abnormal sounds of power transmission lines as described in any one of claims 1-7.

10. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing the method for dynamic screening and alarming of abnormal sounds of power transmission lines as described in any one of claims 1-7.

Images