Photoacoustic feature combined particle identification method and system based on KNN algorithm

Abstract

The invention relates to the technical field of partial discharge detection of power equipment, and discloses a KNN algorithm-based photoacoustic feature combined particle identification method and system, optical and acoustic signals caused by particles in different shapes under alternating voltage are obtained through experiments, and mathematical modeling is performed on particle motion discharge based on particle motion and discharge theories; and then, time sequence distribution characteristics of optical and ultrasonic signals triggered by the particles are mastered, a particle type identification algorithm based on a KNN model is constructed, and particle shape identification and mixed particle component identification are realized. In conclusion, the method combines an optical detection method and an ultrasonic method to form photoacoustic combined detection, excavates combined feature information and establishes a particle discharge combined detection method, and has important significance on application of a particle discharge optical detection method and multi-parameter combined detection; and the method has outstanding value for maintaining safe and stable operation of power equipment and even a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of partial discharge detection of electric power equipment, and in particular to a particle recognition method and system based on a KNN algorithm and a combination of photoacoustic features. Background Art

[0002] During the production, assembly and operation of GIS, factors such as production process, assembly process and operation mode will lead to various insulation defects inside GIS, and will further lead to insulation degradation during long-term operation, and even cause serious insulation accidents. Metal free particles are one of the most common and serious insulation defects inside GIS. There are two main sources of them. One is the exogenous particles generated by GIS machining and cutting before operation and not completely cleaning during assembly, and the other is the endogenous particles generated by long-term repeated operation of switch equipment. Under the action of strong electric field, metal free particles will have phenomena such as motion discharge or adsorption, which will lead to serious insulation faults such as breakdown or flashing. According to incomplete statistics, the insulation failure rate caused by particulate foreign matter is as high as 20%, and according to the statistical data of the State Grid, 7 of the 42 faults that occurred from 2010 to 2013 detected particulate residues. In 2017, a 1000kV bus switch of a UHV power station failed, and in 2018, the main transformer switch of a 1000kV power station lost phase. Afterwards, the reasons may be caused by foreign foreign particles. It can be seen that the free metal particles inside GIS seriously threaten the safety of equipment and even the stable operation of the power system, and its harmfulness cannot be underestimated.

[0003] In order to solve the above problems and realize the detection of free metal particles and meet the requirements of safe operation of power equipment, a variety of particle discharge detection methods have been developed. At present, the actual application methods of particle discharge detection in traditional GIS factories or on-site are ultrasonic method and ultra-high frequency method. However, due to signal attenuation and interference from the external environment, the detection effect and diagnostic accuracy of traditional methods are not good. On the other hand, with the development of weak light detection technology, optical measurement method has gradually been applied in GIS partial discharge detection, and a relatively mature signal feature and diagnostic method have been formed. However, the complex structure and reserved installation sites inside the equipment also limit the use of optical measurement method, and a reliable analysis method has not yet been formed for particle discharge. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a particle identification method based on the photoacoustic feature combination of the KNN algorithm, which can improve the accuracy and efficiency of discharge component identification and provide a reliable technical means for fault diagnosis and maintenance in power systems. At the same time, by using the validation set to adjust the model parameters, overfitting is prevented, ensuring the generalization ability and stability of the model in practical applications.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a particle identification method based on the photoacoustic feature combination of the KNN algorithm, comprising: preparing a variety of metal particle models, performing local discharge tests on the particles, and collecting local discharge signal characteristics using ultrasonic sensors and optical sensors; establishing a particle type identification model based on the KNN algorithm, inputting the photoacoustic joint feature set into the model to realize the identification of different particle shapes; and inputting the mixed particle features into the model as a verification set to realize the identification of the mixed particle components.

[0007] As a preferred solution of the particle identification method based on the photoacoustic feature combination of the KNN algorithm described in the present invention, wherein: the acquisition of local discharge signal features includes: the ultrasonic sensor uses a piezoelectric ultrasonic probe, the piezoelectric ceramic material used in the detection contact surface converts the mechanical signal into an electrical signal, the ultrasonic probe detects a frequency range of 15-70kHz, and a resonant frequency of 40kHz;

[0008] The optical sensor is a silicon photomultiplier tube, which collects spectral signals in the spectral range of 200 to 900 nm.

[0009] As a preferred solution of the particle identification method based on the photoacoustic feature combination of the KNN algorithm of the present invention, the KNN algorithm includes: determining the k value in the KNN algorithm, that is, selecting the first k samples that are most similar to the sample to be classified, for a new sample to be classified, calculating the distance between the feature vector of the new sample and each sample in the training set, sorting the samples in the training set according to the calculated distance, and finding the k samples that are most similar to the sample to be classified;

[0010] The categories of the k most similar samples are counted, and the category of the sample to be classified is determined by the principle of majority voting.

[0011] As a preferred solution of the particle identification method based on the photoacoustic feature combination of the KNN algorithm described in the present invention, wherein: the photoacoustic combined feature includes an optical feature and an ultrasonic feature;

[0012] Among them, the optical characteristics include optical timing characteristics and optical distribution characteristics; the optical timing characteristics include the maximum value of the spectral pulse, the average value of the spectral pulse and the pulse repetition rate; the optical distribution characteristics include the kurtosis and skewness of the amplitude distribution and the kurtosis and skewness of the time interval distribution;

[0013] Ultrasonic features include ultrasonic timing features and ultrasonic distribution features; ultrasonic timing features include pulse maximum value, pulse average value and pulse repetition rate; ultrasonic distribution features include amplitude distribution kurtosis, skewness and time interval distribution kurtosis, skewness.

[0014] As a preferred solution of the particle identification method based on the KNN algorithm combined with photoacoustic features described in the present invention, the pulse maximum value includes the maximum value of all pulses within the data acquisition time, the pulse average value is the average value of all pulses within the data acquisition time, and the pulse repetition rate is the number of pulses per unit time during the particle discharge process.

[0015] As a preferred solution of the particle identification method based on the KNN algorithm combined with the photoacoustic feature, the kurtosis includes a characteristic value describing the steepness or flatness of the distribution curve, and the calculation formula is:

[0016]

[0017] The skewness describes the characteristic value of the skewness or symmetry of the distribution curve, and the calculation formula is:

[0018]

[0019] Among them, I i is the ith signal pulse, I avg is the average value of the signal pulse, K represents the sample kurtosis value, S represents the sample skewness value, and n represents the total number of samples.

[0020] As a preferred solution of the particle identification method based on the KNN algorithm combined with photoacoustic features described in the present invention, the verification set includes adjusting the model parameters during the model training process to prevent overfitting. The verification set does not participate in the training of the model, but is used to evaluate the performance of the model during the training process and select the best model parameters.

[0021] Another object of the present invention is to provide a particle recognition system based on the photoacoustic feature combination of the KNN algorithm, which can recognize particles: the system collects the photoacoustic signals of metal particles during the local discharge process, and uses the KNN algorithm to analyze the signal characteristics to achieve accurate recognition of particles of different shapes and compositions.

[0022] As a preferred solution of the particle identification system based on the photoacoustic feature combination of the KNN algorithm described in the present invention, it includes: a particle model preparation module, a signal acquisition module, a feature extraction module, a KNN training and identification module, and a verification and optimization module;

[0023] The particle model preparation module is used to prepare various metal particle models to provide test samples for subsequent partial discharge tests;

[0024] The signal acquisition module includes an ultrasonic sensor and an optical sensor, which are used to detect particles in a partial discharge test and collect ultrasonic and optical signal characteristics;

[0025] The feature extraction module processes the ultrasonic and optical signals collected by the signal acquisition module to extract the photoacoustic joint features, including time series features and distribution features;

[0026] The KNN training and identification module uses the extracted photoacoustic joint features to train a particle type identification model based on the KNN algorithm and classifies and identifies new samples;

[0027] The verification and optimization module uses the verification set to evaluate the performance of the trained model, adjusts the model parameters to prevent overfitting, and ensures the accuracy and generalization ability of the model.

[0028] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, steps of a particle identification method based on the combination of photoacoustic features and KNN algorithm are implemented.

[0029] A computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of a particle identification method based on a KNN algorithm and a combination of photoacoustic features are implemented.

[0030] Beneficial effects of the present invention: The present invention obtains light and sound signals caused by particles of different shapes under AC voltage through experiments, and mathematically models the particle motion and discharge based on the particle motion and discharge theory; then, the temporal distribution characteristics of optical and ultrasonic signals induced by particles are mastered, and a particle type recognition algorithm based on the KNN model is constructed, which realizes particle shape recognition and mixed particle composition recognition, and has outstanding value in maintaining the safe and stable operation of power equipment and even power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0032] Figure 1 A schematic flow chart of a particle identification method combining photoacoustic features based on a KNN algorithm is provided in accordance with an embodiment of the present invention.

[0033] Figure 2 A schematic diagram of a particle discharge photoacoustic joint experimental platform for a particle identification method combining photoacoustic features with a KNN algorithm provided in one embodiment of the present invention.

[0034] Figure 3A schematic diagram of a KNN algorithm flow chart of a particle identification method combining photoacoustic features based on a KNN algorithm is provided in accordance with an embodiment of the present invention.

[0035] Figure 4 A schematic diagram of particle identification results of a particle identification method combining photoacoustic features based on a KNN algorithm provided in one embodiment of the present invention.

[0036] Figure 5 A hybrid particle cost analysis result of a particle identification method combining photoacoustic features with a KNN algorithm provided in one embodiment of the present invention.

[0037] Figure 6 A schematic diagram of the working modules of a particle recognition system combining photoacoustic features based on a KNN algorithm provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0039] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0040] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0041] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0042] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0043] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0044] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides a particle identification method based on photoacoustic feature combination of KNN algorithm, including:

[0045] S1: Prepare various metal particle models, conduct partial discharge tests on the particles, and use ultrasonic sensors and optical sensors to collect partial discharge signal characteristics.

[0046] Furthermore, there are three shapes of particles in the metal particle model: granular (P), flake (L) and linear (W). Metal particle models are divided into two categories, namely single particle model and multi-particle model. The single particle model has particles of one shape, while the multi-particle model has particles of multiple shapes mixed.

[0047] The ultrasonic sensor uses a traditional piezoelectric ultrasonic probe. The piezoelectric ceramic material used on the detection contact surface converts the mechanical signal into an electrical signal. The ultrasonic probe has a detection frequency range of 15-70kHz and a resonant frequency of 40kHz. The optical sensor is a silicon photomultiplier tube that can collect spectral signals in the spectral range of 200 to 900nm.

[0048] S2: A particle type recognition model is established based on the KNN algorithm, and the photoacoustic joint feature set is input into the model to realize the recognition of different particle shapes.

[0049] Furthermore, KNN is a basic algorithm in machine learning. The basic idea of ​​the algorithm is that if the k most similar samples in the feature space belong to a certain category, then the sample belongs to this category. The algorithm requires three elements to be determined in advance: k value, distance measurement and category decision method.

[0050] k value: k value is a hyperparameter. KNN results depend largely on the choice of k value. If k value is too small, it will lead to overfitting and the model will be easily affected by noise points. If k value is too large, it will lead to underfitting and increase the training error. The k value is usually selected through cross-validation.

[0051] Distance measurement: KNN measures the similarity between two samples by distance. The main distance measurement methods are Euclidean distance and Manhattan distance, which can be expressed by formula (1) and formula (2) respectively.

[0052]

[0053] D(x,y)=∑|x i -y i | (2)

[0054] Category decision method: In the classification model, the majority voting method and weighted majority voting method are mainly used, that is, the class of the input example is determined by the majority class of the neighboring samples.

[0055] It should be noted that the photoacoustic joint features include optical features and ultrasonic features;

[0056] Among them, the optical characteristics include optical timing characteristics and optical distribution characteristics; the optical timing characteristics include the maximum value of the spectral pulse, the average value of the spectral pulse and the pulse repetition rate; the optical distribution characteristics include the kurtosis and skewness of the amplitude distribution and the kurtosis and skewness of the time interval distribution;

[0057] Ultrasonic features include ultrasonic timing features and ultrasonic distribution features; ultrasonic timing features include pulse maximum value, pulse average value and pulse repetition rate; ultrasonic distribution features include amplitude distribution kurtosis, skewness and time interval distribution kurtosis, skewness.

[0058] Furthermore, the pulse maximum value is the maximum value of all pulses within the data acquisition time, the pulse average value is the average value of all pulses within the data acquisition time, and the pulse repetition rate is the number of pulses per unit time during the particle discharge process.

[0059] Kurtosis is a characteristic value that describes the steepness or flatness of the distribution curve. The calculation formula is:

[0060]

[0061] Skewness is a characteristic value that describes the skewness or symmetry of the distribution curve. The calculation formula is:

[0062]

[0063] Among them, I i is the ith signal pulse, I avgis the average value of the signal pulse, K represents the sample kurtosis value, S represents the sample skewness value, and n represents the total number of samples.

[0064] S3: The mixed particle features are input into the model as a validation set to identify the mixed particle components.

[0065] Furthermore, the validation set is used to adjust model parameters during model training to prevent overfitting. The validation set does not participate in the training of the model, but is used to evaluate the performance of the model and select the best model parameters during the training process.

[0066] Adopt adaptive parameter adjustment strategy to optimize the k value in KNN algorithm. Define temperature parameter T, which gradually decreases with the increase of iteration number. The parameter adjustment formula is expressed as:

[0067]

[0068] Where Δk is the adjustment step of k value, T0 is the initial temperature, k old represents the parameter before adjustment, k new Represents the adjusted parameters, and the sigmoid function ensures that the adjustment process is smooth and convergent.

[0069] It should be noted that the weighted accuracy WAcc is used to evaluate the model performance, taking into account the importance of samples of different categories. The formula is as follows:

[0070]

[0071] Where C is the number of categories, Acc i is the accuracy of the i-th category, w i is the weight of the ith category, which can be set according to the relative importance of the category in the data set.

[0072] The loss function is used to select the k value that maximizes L(k) as the optimal model parameter, which is defined as follows:

[0073] L(k)=WAcc-λ·P(k)

[0074]

[0075] Where P(k) is the parameter complexity penalty term, λ is the penalty coefficient, and μ is a small positive number to avoid the denominator being zero.

[0076] Example 2, reference Figure 2 - Figure 5 , which is an embodiment of the present invention, provides a particle identification method based on the combination of photoacoustic features and KNN algorithm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0077] The study selected three types of particles: granular (P), flake (L) and linear (W). The mixed particle sample was a mixture of particles of three different shapes. Figure 2 It is a combined photoacoustic test platform for particle discharge, in which SiPM is for optical signal acquisition and AE is for ultrasonic signal acquisition. The collected optical and acoustic signals are processed and analyzed to extract signal features. The optical features include optical timing features and optical distribution features, in which the optical timing features include the maximum value of the spectral pulse, the average value of the spectral pulse and the pulse repetition rate. The optical distribution features include the kurtosis and skewness of the amplitude distribution and the kurtosis and skewness of the time interval distribution. The ultrasonic features include ultrasonic timing features and ultrasonic distribution features, in which the ultrasonic timing features include the maximum value of the pulse, the average value of the pulse and the pulse repetition rate. The ultrasonic distribution features include the kurtosis and skewness of the amplitude distribution and the kurtosis and skewness of the time interval distribution.

[0078] Before the experiment, the inside of the chamber was evacuated to vacuum by a vacuum pump, and then filled with SF6 gas to 0.3MPa. This process is a gas washing operation. After three gas washing operations, the SF6 gas pressure in the chamber was maintained at 0.3MPa and no leakage was ensured, which ensured that a gas insulation environment was established inside the experimental chamber. The step-up method was used for the experiment and the pressure gradient was set to 2kV. At the beginning of the experiment, the pressure was first applied to the particle discharge starting voltage Us, and then maintained for 1 minute to ensure that all particles and electrodes fully exchanged charges. The optical signal and ultrasonic signal were collected for 50s, and then the pressure was applied to the nearest even-numbered voltage level (for example, from 20.5kV or 21.5kV to 22kV), and the signal was collected for 50s after 1 minute. After that, the pressure-holding-collection operation was repeated with a gradient of 2kV until the electrode self-discharge voltage was reached, and finally the pressure was reduced to 0 and the insulating gas was released, and this group of experiments ended. Other types of particles were tested according to the above scheme to ensure that they were all studied at the same voltage level.

[0079] After determining the three elements of k value, distance measurement and category decision method, the specific algorithm flow of KNN is as follows: Figure 3 As shown in the figure, in this paper, the k value is selected as 3, the distance metric is Manhattan distance, the category decision method is weighted majority voting, and the weight is inversely proportional to the distance, that is, the closer the distance, the greater the weight. The training sets are imported into the KNN model, and the model verification method is selected as five-fold cross validation. At the same time, in order to verify the rationality and effectiveness of the photoacoustic joint detection, the recognition results of the optical and ultrasonic signal (LI, AE) individual features and the photoacoustic joint features (AE+LI) are compared. The output results are as follows: Figure 4 As shown. Figure 4When the optical signal is input alone, the recognition accuracy of granular particles is the highest, which is 95.3%, followed by flake-shaped particles at 92.9% and linear particles at 85.9%; when the ultrasonic signal is input alone, the recognition accuracy of granular and linear particles increases to 97.3% and 91.3% respectively, while the recognition accuracy of flake-shaped particles decreases to 89.9%. Therefore, for the KNN model, the ultrasonic signal features of granular and linear particles are more significant, while the optical signal features of flake-shaped particles are more significant; and when the photoelectric signal features are jointly input, the recognition accuracy of the three types of particles increases significantly, reaching 99.2%, 100% and 92.9% respectively.

[0080] Based on the KNN model trained with photoacoustic joint features, the unlabeled validation set was input into the model to verify the model's ability to recognize the components of mixed particles. For different types of mixed particles, the first two types of particle recognition types of the model (with the largest proportion) were listed. Figure 5 The mixed particle verification results show that the KNN model can effectively identify the components of the mixed particle model composed of two types of shaped particles. For the three types of fully mixed particles, the KNN model can identify the relatively high-risk flake and linear particle components.

[0081] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0082] Embodiment 3, the third embodiment of the present invention, is different from the first two embodiments in that:

[0083] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0084] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0085] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0086] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0087] Example 4, reference Figure 6 , which is an embodiment of the present invention, provides a particle identification system based on photoacoustic feature combination of KNN algorithm, characterized by: comprising a particle model preparation module 1, a signal acquisition module 2, a feature extraction module 3, a KNN training and identification module 4, and a verification and optimization module 6;

[0088] Particle model preparation module 1, used to prepare various metal particle models to provide test samples for subsequent partial discharge tests;

[0089] Signal acquisition module 2, including ultrasonic sensors and optical sensors, used to detect particles in partial discharge tests and collect ultrasonic and optical signal characteristics;

[0090] Feature extraction module 3 processes the ultrasonic and optical signals collected by the signal acquisition module and extracts the photoacoustic joint features, including time series features and distribution features;

[0091] KNN training and identification module 4, based on the KNN algorithm, uses the extracted photoacoustic joint features to train the particle type identification model and classify and identify new samples;

[0092] Verification and optimization module 5 uses the validation set to evaluate the performance of the trained model and adjust the model parameters to prevent overfitting and ensure the accuracy and generalization ability of the model.

[0093] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A particle identification method based on the photoacoustic feature combination of the KNN algorithm, characterized by: include, Prepare various metal particle models, conduct partial discharge tests on the particles, and use ultrasonic sensors and optical sensors to collect partial discharge signal characteristics; A particle type recognition model was established based on the KNN algorithm, and the photoacoustic joint feature set was input into the model to realize the recognition of different particle shapes. The mixed particle features are input into the model as a validation set to achieve the identification of the mixed particle components.

2. The particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in claim 1, characterized in that: The characteristics of collecting partial discharge signals include: the ultrasonic sensor uses a piezoelectric ultrasonic probe, the piezoelectric ceramic material used in the detection contact surface converts the mechanical signal into an electrical signal, the ultrasonic probe detection frequency range is 15-70kHz, and the resonance frequency is 40kHz; The optical sensor is a silicon photomultiplier tube, which collects spectral signals in the spectral range of 200 to 900 nm.

3. The particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in claim 2, characterized in that: The KNN algorithm includes determining the k value in the KNN algorithm, that is, selecting the first k samples that are most similar to the sample to be classified, and for a new sample to be classified, calculating the distance between the feature vector of the new sample and each sample in the training set, and sorting the samples in the training set according to the calculated distance to find the k samples that are most similar to the sample to be classified; The categories of the k most similar samples are counted, and the category of the sample to be classified is determined by the principle of majority voting.

4. The particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in claim 3, characterized in that: The photoacoustic joint features include: Optical and ultrasonic characteristics; Among them, the optical characteristics include optical timing characteristics and optical distribution characteristics; the optical timing characteristics include the maximum value of the spectral pulse, the average value of the spectral pulse and the pulse repetition rate; the optical distribution characteristics include the kurtosis and skewness of the amplitude distribution and the kurtosis and skewness of the time interval distribution; Ultrasonic features include ultrasonic timing features and ultrasonic distribution features; ultrasonic timing features include pulse maximum value, pulse average value and pulse repetition rate; ultrasonic distribution features include amplitude distribution kurtosis, skewness and time interval distribution kurtosis, skewness.

5. The particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in claim 4, characterized in that: The pulse maximum value includes the maximum value of all pulses within the data acquisition time, the pulse average value is the average value of all pulses within the data acquisition time, and the pulse repetition rate is the number of pulses per unit time during the particle discharge process.

6. The particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in claim 5, characterized in that: The kurtosis includes a characteristic value that describes whether the distribution curve is steep or flat, and the calculation formula is: The skewness describes the characteristic value of the skewness or symmetry of the distribution curve, and the calculation formula is: Among them, I i is the ith signal pulse, I avg is the average value of the signal pulse, K represents the sample kurtosis value, S represents the sample skewness value, and n represents the total number of samples.

7. The particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in claim 6, characterized in that: The validation set includes adjusting model parameters during model training to prevent overfitting. The validation set does not participate in the training of the model, but is used to evaluate the performance of the model and select the best model parameters during the training process.

8. A system using the particle identification method based on the photoacoustic feature combination of the KNN algorithm as claimed in any one of claims 1 to 7, characterized in that: It includes particle model preparation module, signal acquisition module, feature extraction module, KNN training and identification module, and verification and optimization module; The particle model preparation module is used to prepare various metal particle models to provide test samples for subsequent partial discharge tests; The signal acquisition module includes an ultrasonic sensor and an optical sensor, which are used to detect particles in a partial discharge test and collect ultrasonic and optical signal characteristics; The feature extraction module processes the ultrasonic and optical signals collected by the signal acquisition module to extract the photoacoustic joint features, including time series features and distribution features; The KNN training and identification module uses the extracted photoacoustic joint features to train a particle type identification model based on the KNN algorithm and classifies and identifies new samples; The verification and optimization module uses the verification set to evaluate the performance of the trained model, adjusts the model parameters to prevent overfitting, and ensures the accuracy and generalization ability of the model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images