A power quality composite disturbance detection method, a detection system and a terminal

Abstract

The application provides a power quality composite disturbance detection method, a detection system and a terminal, relates to the technical field of power quality disturbance detection, extracts the characteristics of power quality disturbance in power distribution Internet of Things through a strong tracking filter, obtains a discrete form of a voltage signal, selects a maximum value of a fundamental wave amplitude, a minimum value of the fundamental wave amplitude, a fluctuation frequency, and a fading factor frequency as disturbance characteristics, extracts the disturbance characteristics, inputs the power quality composite disturbance characteristics into a support vector machine, and completes power quality composite disturbance detection of the power distribution Internet of Things. The application solves the problems that the current power quality composite disturbance detection method cannot accurately obtain power quality characteristics, the filtering effect of power data is not ideal, and the traditional method has the problems of low recognition rate, large detection deviation and poor detection effect. The application completes the detection of power quality composite disturbance of the power distribution Internet of Things.

Description

Technical Field

[0001] The present invention relates to the technical field of power quality disturbance detection, and in particular to a power quality composite disturbance detection method, a detection system and a terminal. Background Art

[0002] As the power grid gradually shifts to a smart grid that combines large power grids and microgrids based on clean energy, the power supply structure has changed; the number of nonlinear loads such as controllable semiconductor electrical equipment and precision electronic equipment that is sensitive to power quality continues to rise, and new problems such as the power quality in distributed power generation systems such as microgrids, the impact of intermittent power supply connection on power quality, and the power supply reliability of the combination of large power grids and microgrids have emerged one after another and continue to receive widespread attention from the scientific research community.

[0003] There are many methods for extracting the characteristics of power quality disturbance signals, such as short-time Fourier transform method, wavelet transform method, S transform method, etc.

[0004] The short-time Fourier transform (STFT) method extracts local features of power quality disturbance signals using a short-time window function. The STFT method is suitable for time-frequency analysis, simultaneously obtaining both the spectrum and time-domain information of power quality disturbance signals, making it suitable for detecting single transient abnormal disturbance events. While the STFT method can extract relevant feature sequences from the transformed matrix when extracting features from power quality disturbance signals, due to the limitation that the window function cannot be changed, this method is more suitable for extracting features from stationary signals.

[0005] The wavelet transform method improves on the short-time Fourier transform method for local feature extraction by incorporating scale and translation factors. This allows the window function to adapt to changes in scale, resulting in excellent adaptability in both the time and frequency domains, making it suitable for the classification and identification of transient abnormal power quality disturbances. However, the wavelet transform method is highly sensitive to noise, requiring noise reduction during preprocessing of the power quality disturbance signal. Furthermore, this method cannot accurately detect the onset and end times of complex power quality disturbances in the distribution Internet of Things.

[0006] The S-transform method adds a Gaussian window function to the existing translation and scaling characteristics of the wavelet basis function, allowing the width and height of the window function to be variable and independent of scale. This method achieves better feature extraction than the short-time Fourier transform and wavelet transform. However, the S-transform is not suitable for feature extraction of high-frequency power quality disturbance signals, nor is it suitable for classifying complex power quality disturbance events. It cannot accurately capture the characteristics of complex power quality disturbances, which reduces the method's recognition rate. Summary of the Invention

[0007] The present invention provides a method for detecting composite power quality disturbances, which solves the problems that the current composite power quality disturbance detection method cannot accurately obtain power quality characteristics and the power data filtering effect is not ideal, resulting in low recognition rate, large detection deviation and poor detection effect in traditional methods.

[0008] The power quality composite disturbance detection method includes:

[0009] Step 1: Extract the characteristics of power quality disturbances in the distribution Internet of Things through a strong tracking filter to obtain a discrete voltage signal;

[0010] Step 2: Select the maximum fundamental wave amplitude, the minimum fundamental wave amplitude, the number of fluctuations, and the frequency of the fading factor as disturbance features, and extract the disturbance features;

[0011] Step 3: Input the power quality composite disturbance characteristics into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

[0012] It should be further explained that step 1 also includes: configuring a discrete nonlinear system, which is expressed as:

[0013]

[0014] Where: x∈R n represents the state vector; f d :R P ×R n →R n represents a nonlinear function; k represents a discrete time variable; u∈R P represents the input vector; v(k) represents the process noise; y∈R m represents the output vector; h d :R n →R m represents the first-order continuous partial derivative; e(k) represents the observation noise;

[0015] Calculate the state estimate at time (k+1) using formula (2)

[0016]

[0017] Where: K represents the filter gain matrix; γ represents the residual sequence, and its calculation formula is

[0018]

[0019] Where: P(k+1|k) represents the prediction error covariance matrix; R and Q both represent variance; F represents the state transfer matrix; H represents the measurement matrix; Λ represents the fading factor matrix.

[0020] It should be further explained that step 2 also includes:

[0021] The voltage signal in the discrete state can be expressed as

[0022]

[0023] Where: A0(k) represents the DC offset; Indicates the initial phase angle of the fundamental wave; A m (k) represents the amplitude corresponding to the fundamental wave; T represents the sampling time interval; e(k) represents the observation noise and conforms to the following formula:

[0024]

[0025] Where: R represents the covariance of noise;

[0026] The state variables are calculated using the following formula:

[0027]

[0028] Where: x3(k) represents the state variable of the fundamental voltage; x4(k) represents the state variable of the orthogonal component. The fundamental voltage amplitude is estimated by equation (7):

[0029]

[0030] The discrete voltage signal y(k) is obtained through the above analysis:

[0031]

[0032] It should be further explained that step 2 also includes: if k is used to represent the signal sampling point, then the maximum value feature C1=max[A1(k)] corresponding to the fundamental wave amplitude represents the increase in the fundamental wave amplitude compared with the original state when the amplitude-type disturbance occurs in the power distribution Internet of Things; the minimum value feature C2=min[A1(k)] corresponding to the fundamental wave amplitude represents the decrease in the fundamental wave amplitude compared with the original state when the amplitude-type disturbance occurs in the power distribution Internet of Things;

[0033] The number of fluctuations C3 represents the event that the time interval of a certain threshold d crossing the fundamental amplitude curve is longer than the time window τ1T. The number of fluctuations is usually divided into two types: positive crossing and negative crossing. The mean frequency of the fading factor C4 represents the number of fading factors in the time window τ1T. Its calculation formula is as follows:

[0034]

[0035] Where: λ represents the fading factor.

[0036] It should be further explained that step three also includes: performing nonlinear mapping on the inseparable samples in the input space, transforming them into a high-dimensional space, and completing data classification by solving a linear separable problem;

[0037] The classification plane D is expressed as

[0038] D=ω T x+b=0 (10)

[0039] Where: ω represents the connection weight; b represents the bias;

[0040] Set the sample set (x1,y1),…,(x n ,y n ), the samples in the sample set are all linearly separable, and the classification plane is used to segment the above samples, strengthening the isolation edge of positive and negative samples in the classification process, and obtaining the hyperplane y i (ω T x i +b)≥1; set the classification interval d x , which represents the distance between the hyperplane and the positive and negative samples, and its calculation formula is

[0041]

[0042] From the above analysis, we can know that the problem of constructing the optimal classification hyperplane U can be replaced by the problem of maximizing the classification interval:

[0043]

[0044] Solve equation (12) through equation (13):

[0045]

[0046] Where: α i Denotes the Lagrange multiplier, and replaces the above problem with the dual problem:

[0047]

[0048] Set the maximum value of the function Q(α):

[0049]

[0050] like is the optimal solution of the support vector machine, then there exists

[0051] According to the above process, the optimal classification function d(x) is constructed:

[0052]

[0053] Where: b * Represents the classification threshold.

[0054] It should be further explained that in the feature space, the kernel function K(x i ·x j ) instead of the inner product (x i ·x j ), and the optimal target classification function is expressed by the following formula:

[0055]

[0056] At this time, the kernel classification function of the support vector machine is expressed as

[0057]

[0058] When the sample is linearly inseparable, set the penalty factor C and the slack variable ξ i , and introduce the above two parameters into the support vector machine, then the objective function can be expressed as

[0059]

[0060] For the kernel function K(x i ·x j ), select the Gaussian radial basis kernel function, and its expression is as follows:

[0061] K(x i ·x j )=exp[-κ||x j -x i || 2 ] (20)

[0062] Where: κ represents the kernel function parameter. The obtained power quality composite disturbance features are input into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

[0063] The present invention also provides a power quality composite disturbance detection system, the system comprising: a feature extraction module and a power quality composite disturbance detection module;

[0064] The feature extraction module is used to extract the characteristics of power quality disturbances in the power distribution Internet of Things through a strong tracking filter to obtain a discrete voltage signal. The maximum fundamental amplitude, minimum fundamental amplitude, number of fluctuations, and frequency of the fading factor are selected as disturbance features and the disturbance features are extracted.

[0065] The power quality composite disturbance detection module is used to input the power quality composite disturbance features into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

[0066] The present invention also provides a terminal for implementing a method for detecting a composite power quality disturbance, comprising:

[0067] A memory, used for storing a computer program and a power quality composite disturbance detection method;

[0068] The processor is used to execute the computer program and the power quality composite disturbance detection method to implement the steps of the power quality composite disturbance detection method.

[0069] It can be seen from the above technical solutions that the present invention has the following advantages:

[0070] The power quality composite disturbance detection method provided by the present invention establishes a discrete nonlinear system and a strong tracking filter, which are applied to the power distribution Internet of Things to extract the characteristics of the power quality composite disturbance. The extracted characteristics are input into a support vector machine to complete the detection of the power quality composite disturbance in the power distribution Internet of Things, thereby improving the method's recognition rate. This overcomes the drawback of the wavelet transform method, which requires noise reduction during the preprocessing of the power quality disturbance signal and cannot accurately detect the occurrence and end time of the power quality composite disturbance in the power distribution Internet of Things.

[0071] It also solves the problem that the S transform is not suitable for feature extraction of high-frequency power quality disturbance signals, is not suitable for classification of complex power quality abnormal disturbance events, and cannot accurately obtain the characteristics of power quality complex disturbances.

[0072] The present invention selects the maximum fundamental wave amplitude, the minimum fundamental wave amplitude, the number of fluctuations, and the frequency of the fading factor as disturbance features, uses a discrete nonlinear system and a strong tracking filter to extract the features and inputs them into a support vector machine to complete the detection of composite disturbances in the power quality of the distribution Internet of Things. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. 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 paying any creative work.

[0074] Figure 1 This is a flow chart of the power quality composite disturbance detection method;

[0075] Figure 2 Schematic diagram of the power quality composite disturbance detection system. DETAILED DESCRIPTION

[0076] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0077] The power quality composite disturbance detection method provided by the present invention is a detection method for power quality based on power quality disturbance detection. Detection involves obtaining information containing power quality disturbance signal characteristics through mathematical methods such as signal processing, and is a preparatory process for disturbance signal feature extraction. Feature selection involves calculating the original feature set from the detection results and extracting and selecting an effective feature subset using optimization techniques. Power quality disturbance detection not only provides disturbance event characteristics for power quality management, but also provides valuable reference information for power quality analysis, diagnosis, and fault location. It also provides a basis for judgment in power disturbance data correlation analysis and data mining.

[0078] There are two main types of research on power quality disturbance detection. One type only detects power quality disturbances. This type of research is mainly used to capture power quality disturbances, which can effectively discover the disturbance process and quickly analyze its characteristics. The other type combines detection and identification. This type of research not only needs to correctly detect the disturbance process and characteristics, but also needs to specifically determine the disturbance process to the type of disturbance event, which plays a decision-making auxiliary role in power quality management and disturbance event accountability.

[0079] The power quality composite disturbance detection method provided by the present invention is implemented based on a terminal, which may include mobile terminals such as mobile phones, smart phones, laptops, personal digital assistants (PDAs), tablet computers (PADs), etc., as well as fixed terminals such as digital TVs, desktop computers, etc.

[0080] A terminal may include a processor (CPU), a read-only memory (ROM), and a memory. The CPU, ROM, and memory are connected to each other via a bus. It may also include a keyboard, a mouse, an LCD display, and a speaker.

[0081] Among them, the memory is used to store the computer program and the power quality composite disturbance detection method; the processor is used to execute the computer program and the power quality composite disturbance detection method to implement the steps of the power quality composite disturbance detection method.

[0082] The present invention addresses the problems of low recognition rate, large detection deviation and poor detection effect in traditional power quality disturbance detection methods in power distribution Internet of Things power quality composite disturbance detection. A power quality composite disturbance detection method is proposed, in which Figure 1 As shown,

[0083] S101, extracting the characteristics of power quality disturbances in the power distribution Internet of Things through a strong tracking filter to obtain a discrete voltage signal;

[0084] Strong tracking filters are a new type of filter for nonlinear systems. They make the residual sequences orthogonal at each step, extracting all useful information from the residual sequences for estimating the current system state. They are also robust to model parameter mismatches and offer strong real-time tracking capabilities for sudden changes in state and parameters.

[0085] For the strong tracking filter of the present invention, discrete nonlinear system processing can be performed based on the strong tracking filter, which is specifically expressed as

[0086]

[0087] Where: x∈R n represents the state vector; f d :R P ×R n →R n represents a nonlinear function; k represents a discrete time variable; u∈R P represents the input vector; v(k) represents the process noise; y∈R m represents the output vector; h d :R n →R m represents the first-order continuous partial derivative; e(k) represents the observation noise.

[0088] Calculate the state estimate at time (k+1) using formula (2)

[0089]

[0090] Where: K represents the filter gain matrix; γ represents the residual sequence, and its calculation formula is

[0091]

[0092] Where: P(k+1|k) represents the prediction error covariance matrix; R and Q both represent variance; F represents the state transfer matrix; H represents the measurement matrix; Λ represents the fading factor matrix.

[0093] S102, selecting the maximum fundamental wave amplitude, the minimum fundamental wave amplitude, the number of fluctuations, and the frequency of the fading factor as disturbance features, and extracting the disturbance features;

[0094] The voltage signal in the discrete state can be expressed as

[0095]

[0096] Where: A0(k) represents the DC offset; Indicates the initial phase angle of the fundamental wave; A m (k) represents the amplitude corresponding to the fundamental wave; T represents the sampling time interval; e(k) represents the observation noise and conforms to the following formula:

[0097]

[0098] Where: R represents the covariance of the noise.

[0099] The state variables are calculated using the following formula:

[0100]

[0101] Where: x3(k) represents the state variable of the fundamental voltage; x4(k) represents the state variable of the orthogonal component. The fundamental voltage amplitude is estimated by equation (7):

[0102]

[0103] The discrete voltage signal y(k) is obtained through the above analysis:

[0104]

[0105] For the power quality composite disturbance detection method provided by the present invention, the present invention selects the maximum fundamental wave amplitude, the minimum fundamental wave amplitude, the number of fluctuations, and the frequency of the fading factor as disturbance characteristics.

[0106] If k represents the signal sampling point, the maximum value feature C1=max[A1(k)] corresponding to the fundamental wave amplitude indicates the increase in the fundamental wave amplitude compared with the original state when an amplitude-type disturbance occurs in the power distribution Internet of Things; the minimum value feature C2=min[A1(k)] corresponding to the fundamental wave amplitude indicates the decrease in the fundamental wave amplitude compared with the original state when an amplitude-type disturbance occurs in the power distribution Internet of Things.

[0107] The number of fluctuations C3 represents the event that the time interval of a certain threshold d crossing the fundamental amplitude curve is longer than the time window τ1T. The number of fluctuations is usually divided into two types: positive crossing and negative crossing. The mean frequency of the fading factor C4 represents the number of fading factors in the time window τ1T. Its calculation formula is as follows:

[0108]

[0109] Where: λ represents the fading factor.

[0110] S103: Input the power quality composite disturbance characteristics into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

[0111] In the power quality composite disturbance detection method provided by the present invention, the main purpose of the support vector machine is to complete nonlinear mapping of inseparable samples in the input space, transform them into a high-dimensional space, and complete data classification by solving linear separable problems.

[0112] The classification plane D is expressed as

[0113] D=ω T x+b=0 (10)

[0114] Where: ω represents the connection weight; b represents the bias.

[0115] Set the sample set (x1,y1),…,(x n ,y n ), the samples in the sample set are all linearly separable, and the classification plane is used to segment the above samples, strengthening the isolation edge of positive and negative samples in the classification process, and on this basis, the hyperplane y is obtained i (ω T x i +b)≥1; set the classification interval d x , which represents the distance between the hyperplane and the positive and negative samples, and its calculation formula is

[0116]

[0117] From the above analysis, we can know that the problem of constructing the optimal classification hyperplane U can be replaced by the problem of maximizing the classification interval:

[0118]

[0119] Solve equation (12) through equation (13):

[0120]

[0121] Where: α i Denotes the Lagrange multiplier, and replaces the above problem with the dual problem:

[0122]

[0123] Set the maximum value of the function Q(α):

[0124]

[0125] like is the optimal solution of the support vector machine, then there exists

[0126] According to the above process, the optimal classification function d(x) is constructed:

[0127]

[0128] Where: b * Indicates the classification threshold. In the feature space, the kernel function K(x i ·x j ) instead of the inner product (x i ·x j ), and the optimal target classification function is expressed by the following formula:

[0129]

[0130] At this time, the kernel classification function of the support vector machine is expressed as

[0131]

[0132] When the sample is linearly inseparable, set the penalty factor C and the slack variable ξ i , and introduce the above two parameters into the support vector machine, then the objective function can be expressed as

[0133]

[0134] For the kernel function K(x i ·x j ), the proposed method uses the Gaussian radial basis kernel function, which is expressed as follows:

[0135] K(x i ·x j )=exp[-κ||x j -x i || 2 ] (20)

[0136] Where κ represents the kernel function parameter. The acquired power quality composite disturbance features are input into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

[0137] In this way, the power quality composite disturbance detection method provided by the present invention establishes a discrete nonlinear system and a strong tracking filter, applies it to the power distribution Internet of Things, and extracts the power quality composite disturbance characteristics;

[0138] The present invention selects the maximum fundamental wave amplitude, the minimum fundamental wave amplitude, the number of fluctuations, and the frequency of the fading factor as disturbance features, uses a discrete nonlinear system and a strong tracking filter to extract the features and inputs them into a support vector machine to complete the detection of composite disturbances in the power quality of the distribution Internet of Things.

[0139] Based on the above power quality composite disturbance detection method, the present invention also provides a power quality composite disturbance detection system, such as Figure 2 As shown, the system includes: a feature extraction module and a power quality composite disturbance detection module;

[0140] The feature extraction module is used to extract the characteristics of power quality disturbances in the power distribution Internet of Things through a strong tracking filter to obtain a discrete voltage signal. The maximum fundamental amplitude, minimum fundamental amplitude, number of fluctuations, and frequency of the fading factor are selected as disturbance features and the disturbance features are extracted.

[0141] The power quality composite disturbance detection module is used to input the power quality composite disturbance features into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

[0142] Specifically, the present invention extracts the characteristics of composite power quality disturbances by establishing a discrete nonlinear system and a strong tracking filter. The maximum fundamental amplitude, minimum fundamental amplitude, number of fluctuations, and frequency of the fading factor are selected as disturbance features, and the extracted features are input into a support vector machine to detect composite power quality disturbances in the distribution Internet of Things. This solves the problem that current composite power quality disturbance detection methods cannot accurately obtain power quality characteristics and have unsatisfactory power data filtering effects. It also avoids the low recognition rate, large detection deviation, and poor detection effect of traditional methods.

[0143] The units and algorithm steps of each example described in the embodiments disclosed in the power quality composite disturbance detection method and system provided by the present invention can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0144] The block diagrams shown in the accompanying drawings of the power quality composite disturbance detection method and system provided by the present invention are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0145] The power quality composite disturbance detection method and system provided by the present invention are the units and algorithm steps of each example described in combination with the embodiments disclosed herein, and can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0146] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting composite power quality disturbances, characterized in that: Methods include: Step 1: Extract the characteristics of power quality disturbances in the distribution Internet of Things through a strong tracking filter to obtain a discrete voltage signal; Configure the discrete nonlinear system, which is expressed as: （1） Where: represents the state vector; represents a nonlinear function; represents a discrete-time variable; represents the input vector; represents process noise; represents the output vector; represents the first-order continuous partial derivative; represents the observation noise; Calculate using formula (2) The estimated state at time : （2） Where: represents the filter gain matrix; Represents the residual sequence, and its calculation formula is （3） Where: represents the forecast error covariance matrix; 、 Both represent variance; represents the state transition matrix; represents the measurement matrix; represents the vanishing factor matrix; Step 2: Select the maximum fundamental wave amplitude, the minimum fundamental wave amplitude, the number of fluctuations, and the frequency of the fading factor as disturbance features, and extract the disturbance features; The voltage signal in the discrete state can be expressed as （4） Where: Indicates DC offset; Indicates the initial phase angle of the fundamental wave; Indicates the amplitude corresponding to the fundamental wave; Indicates the sampling time interval; represents the observation noise and conforms to the following formula: （5） Where: represents the covariance of the noise; The state variables are calculated using the following formula: （6） Where: The state variable representing the fundamental voltage; Represents the state variable of the orthogonal component; the fundamental voltage amplitude is estimated by formula (7) : （7） The voltage signal in discrete form is obtained through the above analysis : （8） Step 3: Input the power quality composite disturbance characteristics into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

2. The power quality composite disturbance detection method according to claim 1, characterized in that: Step 2 also includes: Represents the signal sampling point, then the maximum value characteristic corresponding to the fundamental amplitude , which represents the increase in the fundamental amplitude compared with the original state when an amplitude disturbance occurs in the power distribution Internet of Things; the minimum value characteristic corresponding to the fundamental amplitude , which represents the decrease in the amplitude of the fundamental wave compared with the original state when an amplitude-type disturbance occurs in the power distribution Internet of Things; Number of fluctuations Indicates a threshold The time interval of the fundamental amplitude curve is higher than the time window The number of fluctuations is usually divided into two types: positive crossing and negative crossing; the mean frequency of the fading factor Indicates the fading factor in the time window The number in is calculated as follows: （9） Where: represents the fading factor.

3. The power quality composite disturbance detection method according to claim 1, characterized in that: Step 3 also includes: performing nonlinear mapping on the inseparable samples in the input space, transforming them into a high-dimensional space, and completing data classification by solving the linear separable problem; Classification plane Expressed as （10） Where: represents the connection weight; Indicates bias; Setting up sample sets , the samples in the sample set are all linearly separable, and the classification plane is used to segment the above samples, strengthening the isolation edge of positive and negative samples in the classification process, and obtaining the hyperplane ; Set the classification interval , which represents the distance between the hyperplane and the positive and negative samples, and its calculation formula is （11） From the above analysis, we can see that the maximum classification interval problem can be used instead of the optimal classification hyperplane Build issues: （12） Solve equation (12) through equation (13): （13） Where: Denotes the Lagrange multiplier, and replaces the above problem with the dual problem: （14） Set the maximum value of the function : （15） like is the optimal solution of the support vector machine, then there exists ; According to the above process, the optimal classification function is constructed : （16） Where: Represents the classification threshold.

4. The power quality composite disturbance detection method according to claim 3, characterized in that: In the feature space, the kernel function Replace inner product , and the optimal target classification function is expressed by the following formula: （17） At this time, the kernel classification function of the support vector machine is expressed as （18） When the sample is linearly inseparable, set the penalty factor and slack variables , and introduce the above two parameters into the support vector machine, then the objective function can be expressed as （19） For the kernel function in support vector machine , select the Gaussian radial basis kernel function, its expression is as follows: （20） Where: Represents the kernel function parameters, and the obtained power quality composite disturbance features are input into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

5. A power quality composite disturbance detection system, characterized in that: The system adopts the power quality composite disturbance detection method according to any one of claims 1 to 4; The system includes: a feature extraction module and a power quality composite disturbance detection module; The feature extraction module is used to extract the characteristics of power quality disturbances in the power distribution Internet of Things through a strong tracking filter to obtain a discrete voltage signal. The maximum fundamental amplitude, minimum fundamental amplitude, number of fluctuations, and frequency of the fading factor are selected as disturbance features and the disturbance features are extracted. The power quality composite disturbance detection module is used to input the power quality composite disturbance features into the support vector machine to complete the power quality composite disturbance detection of the distribution Internet of Things.

6. A terminal device for implementing a method for detecting a composite power quality disturbance, characterized in that: include: A memory, used for storing a computer program and a power quality composite disturbance detection method; A processor is configured to execute the computer program and the method for detecting a composite power quality disturbance, so as to implement the steps of the method for detecting a composite power quality disturbance as claimed in any one of claims 1 to 4.

Images