Low-voltage distribution network operation state detection method based on fusion terminal

Abstract

The invention discloses a low-voltage power distribution network operation state detection method based on a fusion terminal, and relates to the technical field of power distribution networks, and the method comprises the steps that the fusion terminal obtains an original data set of a low-voltage power distribution network; performing data preprocessing on the original data set; performing feature extraction on the preprocessed data; performing feature fusion on the extracted features to obtain a fused feature vector, and dividing a training set and a test set; constructing a neural network classification model for training to obtain a state classification model, optimizing the state classification model, and inputting the fusion feature vector into the state classification model for classification; selecting abnormal state data for detection, and identifying an abnormal state type; and selecting normal state data for health assessment, and comprehensively evaluating the health state of the low-voltage distribution network through the health key indexes and the health assessment function. According to the invention, the monitoring precision of the low-voltage power distribution network is improved, the anomaly detection and health assessment capabilities are enhanced, and the resource configuration is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network, and in particular to a method for detecting the operating status of a low-voltage power distribution network based on a fusion terminal. Background Art

[0002] The current operation status detection of low-voltage distribution networks relies on a number of key technologies. Among them, smart meters have become an indispensable part of low-voltage distribution networks. Through data collection, communication and monitoring functions, they can obtain user power consumption information and the operation status of distribution equipment. Sensor technology is widely used in the detection of power parameters such as current, voltage, power factor, and physical quantities such as equipment temperature and humidity. Data communication technology realizes remote data transmission and centralized management. Data analysis and processing technology extracts useful information and judges the operation status through methods such as statistical analysis, pattern recognition and machine learning. Cloud computing and big data technology are applied to data storage, processing and analysis to realize centralized management, real-time monitoring and remote control of distribution network data. The comprehensive application of these technologies improves the safety, stability and operation efficiency of low-voltage distribution networks. Summary of the invention

[0003] A series of simplified concepts are introduced in the Summary of the Invention, which will be further described in detail in the Detailed Description of the Invention. The Summary of the Invention does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0004] Since the traditional low-voltage distribution network lacks effective real-time monitoring means, it is difficult to locate and handle faults in a timely and accurate manner after they occur. The existing distribution network monitoring method may not be able to fully reflect the operating status of the distribution network, resulting in difficulty in accurately carrying out operation and maintenance work. The present invention proposes a low-voltage distribution network operating status detection method based on a fusion terminal, and the above method includes: S10: The fusion terminal obtains the original data set of the low-voltage distribution network; S20: performing data preprocessing on the original data set to obtain a denoised and padded data set; S30: extracting features from the denoising and padding data set to obtain spectrum features, time domain features, and energy features; S40: performing feature fusion on the spectrum features, time domain features, and energy features to obtain fused feature vectors, and selecting part of the fused feature vectors to divide them into a training set and a test set; S50: constructing a neural network classification model, training the neural network classification model through the training set to obtain a state classification model, optimizing the state classification model through the test set, and inputting the fused feature vector into the state classification model for classification; S60: Select data classified as abnormal state for detection, and identify the type of abnormal state; S70: Select data classified as normal status for health assessment, and comprehensively evaluate the health status of the low-voltage distribution network by calculating key health indicators and health assessment functions; The original data set includes: power signal, power data, and power supply data; The power signal includes: current and voltage; The power data includes: rated active power of transformer, actual active power of transformer, and three-phase load; The power supply data includes: power supply radius, total power supply, total power sales, total power supply time, and voltage limit exceeding time; The partial fusion feature vectors are 40% randomly selected from the fusion feature vectors.

[0005] Further, the S20 includes: S21: dividing the missing data in the original data set into recent missing data and common missing data, using timestamp alignment interpolation to fill the recent missing data, and using Lagrange interpolation to fill the common missing data, to obtain a filled data set; S22: performing data denoising on the padded data set, performing denoising on the time domain coded signal by using a moving average filter, and performing denoising on the frequency domain coded signal by using a Butterworth low-pass filter, to obtain a denoised padded data set.

[0006] Further, the S30 includes: S31: converting the power signal data in the denoised and padded data set from the time domain to the frequency domain through fast Fourier transform to obtain energy distribution information at different frequencies; S32: Performing spectrum analysis on the energy distribution information, and extracting spectrum features according to energy corresponding frequencies; S33: decomposing the time series signal in the denoising and padding data set into wavelet coefficients of different scales and frequencies through wavelet transform, calculating the mean, variance, skewness, and kurtosis of the wavelet coefficients, and obtaining time domain features; S34: Calculate the total signal energy of the time domain signal in the denoised and padded data set, and perform spectrum density integration on the frequency domain signal in the denoised and padded data set to obtain energy features.

[0007] Furthermore, the S40 includes: S41: performing feature fusion on the spectrum features, time domain features, and energy features to obtain a fused feature vector; S42: Randomly divide the partial fusion feature vectors into a training set and a test set according to a ratio of 8:2.

[0008] Furthermore, the S50 includes: S51: Determine the number of input layer nodes of the neural network classification model according to the dimension of the fused feature vector; S52: passing through three hidden layers, wherein the hidden layers are fully connected layers and use a ReLU activation function; S53: The output layer uses the Sigmoid activation function to divide the output results into abnormal state and normal state; S54: using the training set to train the neural network classification model to obtain a state classification model; S55: optimizing the state classification model by the testing machine; S56: Input the fused feature vector into a state classification model to obtain abnormal state data and normal state data.

[0009] Furthermore, the S60 includes: S61: According to the characteristics of low-voltage distribution network, three types of abnormal conditions are defined, including: current mutation, frequency deviation, and voltage fluctuation; S62: Calculate the local density of the abnormal state data and its neighborhood data using a local abnormal factor, average the ratio of the local density of the neighborhood data to the local density of the abnormal state data, and obtain the LOF value of the abnormal state data; S63: Identify the type causing the abnormal state according to the LOF value of the abnormal state data.

[0010] Furthermore, the S70 includes: S71: Five key indicators are defined based on the characteristics of low-voltage distribution networks, including: power supply radius, comprehensive line loss rate, transformer load rate, user voltage qualification rate, and three-phase load imbalance; S72: Setting power supply radius constraint conditions of low voltage distribution lines; S73: solving the comprehensive line loss rate of the low voltage distribution network; S74: solving the transformer load factor of the low voltage distribution network; S75: Calculate the voltage qualification rate of the low voltage distribution network; S76: solving the three-phase load imbalance of the low-voltage distribution network; S77: Calculate the weights of the five key indicators using the Lagrange optimal multiplier method to obtain a health assessment function, and perform a comprehensive health evaluation of the low-voltage distribution network on the normal state data based on the health assessment function.

[0011] The beneficial effects of the present invention are:

[0012] 1. The present invention provides a low-voltage distribution network operation status detection method based on a fusion terminal, which effectively removes noise and outliers by interpolating and denoising the collected original data set, ensuring the accuracy and reliability of subsequent analysis, thereby improving the accuracy of low-voltage distribution network monitoring and management.

[0013] 2. The present invention provides a method for detecting the operating status of a low-voltage distribution network based on a fusion terminal. By performing anomaly detection through a local anomaly factor method, the present invention can accurately identify abnormal conditions such as current mutations and frequency deviations, and can help to promptly discover and resolve potential problems, thereby ensuring the stable operation of the low-voltage distribution network.

[0014] 3. The present invention provides a method for detecting the operating status of a low-voltage distribution network based on a fusion terminal. By utilizing fast Fourier transform and wavelet transform to extract spectrum, time domain and energy features, and performing feature fusion, the accuracy and efficiency of state classification are improved, thereby providing strong support for the intelligent management of the low-voltage distribution network.

[0015] 4. The present invention provides a method for detecting the operating status of a low-voltage distribution network based on a fusion terminal. By calculating key indicators such as the comprehensive line loss rate, transformer load, and three-phase load imbalance, a comprehensive assessment of the health status of the distribution network is performed. Managers can better understand the operating status of the distribution network and improve the scientificity and efficiency of management decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a method for detecting the operating status of a low-voltage distribution network based on a fusion terminal of the present invention.

[0017] Figure 2 It is a data preprocessing schematic diagram of a low-voltage distribution network operating status detection method based on a fusion terminal of the present invention.

[0018] Figure 3 It is a feature extraction schematic diagram of a low-voltage distribution network operating status detection method based on a fusion terminal of the present invention.

[0019] Figure 4 It is a characteristic fusion schematic diagram of a low-voltage distribution network operation status detection method based on a fusion terminal of the present invention.

[0020] Figure 5 It is a schematic diagram of a state classification model of a low-voltage distribution network operation state detection method based on a fusion terminal of the present invention.

[0021] Figure 6 It is a schematic diagram of abnormal state detection of a low-voltage distribution network operation state detection method based on a fusion terminal of the present invention.

[0022] Figure 7It is a health assessment schematic diagram of a low-voltage distribution network operation status detection method based on a fusion terminal of the present invention. DETAILED DESCRIPTION

[0023] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0024] 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; it is obvious that the embodiments in the specification are only part of the embodiments of the present invention, rather than all of the embodiments.

[0025] See also Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 , the embodiment of the present invention includes: See also Figure 1 , is a flow chart of a method for detecting the operating status of a low-voltage distribution network based on a fusion terminal provided in an embodiment of the present invention, which may specifically include: steps S10 to S70.

[0026] S10: The fusion terminal obtains the original data set of the low-voltage distribution network.

[0027] Further, the original data set includes: power signal, power data, power supply data; The power signal includes: current and voltage; The power data includes: rated active power of transformer, actual active power of transformer, and three-phase load; The power supply data includes: power supply radius, total power supply, total power sales, total power supply time, and voltage limit exceeding time; The partial fusion feature vectors are 40% randomly selected from the fusion feature vectors.

[0028] Exemplarily, the original data set of the low-voltage distribution network can be obtained from the fusion terminal. The original data set includes: current, voltage, rated active power of transformer, actual active power of transformer, three-phase load, power supply radius, total power supply, total power sales, total power supply time, and voltage over-limit time, forming a data set.

[0029] S20: performing data preprocessing on the original data set to obtain a denoised and padded data set.

[0030] See also Figure 2, which is a data preprocessing diagram of a low-voltage distribution network operating status detection method based on a fusion terminal provided in an embodiment of the present invention, which may specifically include: steps S21 to S22.

[0031] S21: dividing the missing data in the original data set into recent missing data and common missing data, using timestamp alignment interpolation to fill the recent missing data, and using Lagrange interpolation to fill the common missing data, to obtain a filled data set; S22: performing data denoising on the padded data set, denoising the time domain coded signal by using a moving average filter, and denoising the frequency domain coded signal by using a Butterworth low-pass filter, to obtain a denoised padded data set; Exemplarily, the missing data in the original data set are divided into recent missing data and common missing data, and the recent missing data and the common missing data are filled according to various existing data filling methods to obtain a filled data set, such as: K-Nearest Neighbors, matrix decomposition, interpolation method, Random Forest, Generative Adversarial Networks (GANs), etc.

[0032] Exemplarily, the recent missing data and the common missing data are supplemented by an interpolation method to obtain the supplemented data set. The specific steps of the interpolation method are as follows: S211: fill in the recent missing data by aligning and interpolating the timestamps; The timestamp alignment interpolation calculation formula is: ; in, For time Insert the value at For timestamp The known value at For timestamp The known value at S212: Lagrange interpolation is used to complete the common missing data; The Lagrange interpolation completion calculation formula is: ; ; in, is the interpolation result, is the Lagrangian basis function, is the degree of the interpolation polynomial, For the The function value of the data point, For the The value of the data point, For the The value of the data point; Exemplarily, the padded data set is denoised by a filtering denoising method to obtain the denoised padded data set. The specific steps of the filtering denoising method are as follows: S221: De-noising the time-domain coded signal using a moving average filter; The moving average filter calculation formula is: y [ i ]= 1 MP ∑ j =0 MP -1 x i + j ; in, y [ i ] is the output signal, is the input signal, is the number of data points to be averaged; S222: De-noising the frequency domain coded signal using a Butterworth low-pass filter; The Butterworth low-pass filter calculation formula is: ; in, is the frequency response relationship between the input signal and the output signal, is a complex frequency domain variable, is the cut-off frequency, is the order of the Butterworth low-pass filter; S30: performing feature extraction on the denoising and padding data set to obtain spectrum features, time domain features, and energy features.

[0033] See also Figure 3 , which is a schematic diagram of feature extraction of a low-voltage distribution network operating status detection method based on a fusion terminal provided in an embodiment of the present invention, which may specifically include: steps S31 to S34.

[0034] S31: converting the power signal data in the denoised and padded data set from the time domain to the frequency domain through fast Fourier transform to obtain energy distribution information at different frequencies; S32: Performing spectrum analysis on the energy distribution information, and extracting spectrum features according to energy corresponding frequencies; S33: decomposing the time series signal in the denoising and padding data set into wavelet coefficients of different scales and frequencies through wavelet transform, calculating the mean, variance, skewness, and kurtosis of the wavelet coefficients, and obtaining time domain features; S34: calculating the total signal energy of the time domain signal in the denoised and padded data set, and performing spectrum density integration on the frequency domain signal in the denoised and padded data set to obtain energy features; Exemplarily, the power signal data in the denoised and padded data set is converted into a frequency domain representation by fast Fourier transform, and the fast Fourier transform calculation formula is: ,in, For the frequency The spectrum value at is the first data points, is the sequence length, The energy distribution information at different frequencies is obtained, the spectrum analysis is performed on the energy distribution information, and the spectrum features are extracted according to the energy corresponding frequency.

[0035] Exemplarily, the time series signal in the denoised and padded data set is decomposed into wavelet coefficients of different scales and frequencies by wavelet transform, and the wavelet transform calculation formula is: ,in, is the wavelet coefficient, is the input signal, is the complex conjugate of the wavelet basis function, is the time variable, is the scale parameter, is the translation parameter, and the mean, variance, skewness and kurtosis of the wavelet coefficients are calculated to obtain the time domain characteristics.

[0036] Exemplarily, the total signal energy is calculated for the time domain signal in the denoised and padded data set, and the total signal energy calculation formula is: ,in, is the total energy of the time domain signal, is the time domain signal The value of the sampling point, is the number of sampling points of the signal. The frequency domain signal in the denoised and padded data set is subjected to spectral density integration, and the spectral density integration calculation formula is: ,in, is the total energy of the frequency domain signal, For frequency The spectral density at and is the frequency range of the integration.

[0037] S40: performing feature fusion on the spectrum features, time domain features, and energy features to obtain fused feature vectors, and selecting a portion of the fused feature vectors to divide into a training set and a test set.

[0038] See also Figure 4 , which is a feature fusion schematic diagram of a low-voltage distribution network operation status detection method based on a fusion terminal provided in an embodiment of the present invention, which may specifically include: steps S41 to S42.

[0039] S41: performing feature fusion on the spectrum features, time domain features, and energy features to obtain a fused feature vector; S42: randomly dividing the partial fusion feature vectors into a training set and a test set according to a ratio of 8:2; Exemplarily, the extracted spectrum features, time domain features, and energy features are fused to obtain a fused feature vector, such as concatenation, weighted average, cross, stacking, voting, feature selection, etc.

[0040] Exemplarily, the extracted spectrum features, time domain features, and energy features are fused by cascading to obtain a fused feature vector, and the fused feature vector is: X =[ X spectral , X time , X Energy ] ,in, is the fusion feature vector, is the spectrum feature, is the time domain feature, Energy characteristics.

[0041] S50: constructing a neural network classification model, training the neural network classification model through the training set to obtain a state classification model, optimizing the state classification model through the test set, and inputting the fused feature vector into the state classification model for classification.

[0042] See also Figure 5 , which is a schematic diagram of a state classification model of a low-voltage distribution network operation state detection method based on a fusion terminal provided in an embodiment of the present invention, and may specifically include: steps S51 to S56.

[0043] S51: Determine the number of input layer nodes of the neural network classification model according to the dimension of the fused feature vector; S52: passing through three hidden layers, wherein the hidden layers are fully connected layers and use a ReLU activation function; S53: The output layer uses the Sigmoid activation function to divide the output results into abnormal state and normal state; S54: using the training set to train the neural network classification model to obtain a state classification model; S55: optimizing the state classification model by the testing machine; S56: inputting the fused feature vector into a state classification model to obtain abnormal state data and normal state data; Exemplarily, a state classification model is obtained by establishing a neural network classification model and training it. The number of input layer nodes of the neural network classification model is determined by the dimension of the fused feature vector, and three hidden layers are used, wherein the hidden layer adopts a fully connected layer and uses a ReLU activation function, wherein the ReLU activation function is: ,in, is the input value, To take 0 and The maximum value in the output layer uses the Sigmoid activation function, which is ,in, The sigmoid activation function is used as an input value to perform binary classification, and the output result is divided into abnormal state and normal state. The training set is used to train the neural network classification model to obtain a state classification model, and the state classification model is optimized by the test machine. The fused feature vector is input into the state classification model to obtain abnormal state data and normal state data.

[0044] S60: Select data classified as abnormal state for detection, and identify the type of abnormal state.

[0045] See also Figure 6 , is a schematic diagram of abnormal state detection of a low-voltage distribution network operation state detection method based on a fusion terminal provided in an embodiment of the present invention, which may specifically include: steps S61 to S63.

[0046] S61: According to the characteristics of low-voltage distribution network, three types of abnormal conditions are defined, including: current mutation, frequency deviation, and voltage fluctuation; S62: Calculate the local density of the abnormal state data and its neighborhood data using a local abnormal factor, average the ratio of the local density of the neighborhood data to the local density of the abnormal state data, and obtain the LOF value of the abnormal state data; S63: Identify the type causing the abnormal state according to the LOF value of the abnormal state data;

[0047] Exemplarily, according to the characteristics of the low-voltage distribution network, the abnormal state types are divided into three situations, including: current mutation, frequency offset, and voltage fluctuation. The local density is calculated using a local abnormal factor for the abnormal state data and its neighborhood data. The local density calculation formula is: ,in, is the data point for which the local density is to be calculated, For data points The neighboring data points of For data points arrive The distance For data points The local reachable density of To obtain the maximum value, the ratio of the local density of the neighborhood data to the local density of the abnormal state data is averaged to obtain the LOF value of the abnormal state data. The LOF value calculation formula is: ,in, For data points The set of neighborhood data points, For data points The type causing the abnormal state is identified according to the LOF value of the abnormal state data.

[0048] S70: Select data classified as normal status for health assessment, and comprehensively evaluate the health status of the low-voltage distribution network by calculating key health indicators and health assessment functions.

[0049] See also Figure 7 , which is a health assessment diagram of a low-voltage distribution network operation status detection method based on a fusion terminal provided in an embodiment of the present invention, and may specifically include: steps S71 to S77.

[0050] S71: Five key indicators are defined based on the characteristics of low-voltage distribution networks, including: power supply radius, comprehensive line loss rate, transformer load rate, user voltage qualification rate, and three-phase load imbalance; S72: Setting power supply radius constraint conditions of low voltage distribution lines; S73: solving the comprehensive line loss rate of the low voltage distribution network; S74: solving the transformer load factor of the low voltage distribution network; S75: Calculate the voltage qualification rate of the low voltage distribution network; S76: solving the three-phase load imbalance of the low-voltage distribution network; S77: weighting the five key indicators by Lagrange optimal multiplier method to obtain a health assessment function, and comprehensively evaluating the health of the low-voltage distribution network on the normal state data according to the health assessment function; Exemplarily, five key indicators are defined according to the characteristics of the low-voltage distribution network, including: power supply radius, comprehensive line loss rate, transformer load, user voltage qualification rate, and three-phase load imbalance, and the power supply radius constraint conditions of the low-voltage distribution line are set. The power supply radius constraint conditions are: ,in, The comprehensive line loss rate of the low-voltage distribution network is solved for the power supply radius. The calculation formula of the comprehensive line loss rate is: ,in, is the comprehensive line loss rate, is the total power supply of the target area, The transformer load factor of the low-voltage distribution network is calculated using the following formula: ,in, is the transformer load factor, is the rated active power of the transformer, is the actual active power of the transformer. The voltage qualification rate of the low-voltage distribution network is solved, and the voltage qualification rate calculation formula is: ,in, is the voltage qualification rate, is the total power supply time, The three-phase load imbalance degree is calculated for the low-voltage distribution network. The three-phase load imbalance degree calculation formula is: ,in, is the three-phase load imbalance, To select the maximum value, For the low-pressure side outlet of the station area Phase load, For the low-pressure side outlet of the station area Phase load, For the low-pressure side outlet of the station area The five key indicators are weighted by Lagrange optimal multiplier method to obtain a health assessment function, and the normal state data is comprehensively evaluated for the health of the low-voltage distribution network according to the health assessment function.

[0051] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal, characterized in that: include: S10: The fusion terminal obtains the original data set of the low-voltage distribution network; S20: performing data preprocessing on the original data set to obtain a denoised and padded data set; S30: extracting features from the denoising and padding data set to obtain spectrum features, time domain features, and energy features; S40: performing feature fusion on the spectrum features, time domain features, and energy features to obtain fused feature vectors, and selecting part of the fused feature vectors to divide into a training set and a test set; S50: constructing a neural network classification model, training the neural network classification model through the training set to obtain a state classification model, optimizing the state classification model through the test set, and inputting the fused feature vector into the state classification model for classification; S60: Select data classified as abnormal state for detection, and identify the type of abnormal state; S70: Select data classified as normal status for health assessment, and comprehensively evaluate the health status of the low-voltage distribution network by calculating key health indicators and health assessment functions; The original data set includes: power signal, power data, and power supply data; The power signal includes: current and voltage; The power data includes: rated active power of transformer, actual active power of transformer, and three-phase load; The power supply data includes: power supply radius, total power supply, total power sales, total power supply time, and voltage limit exceeding time; The partial fusion feature vectors are 40% randomly selected from the fusion feature vectors; The classification includes: abnormal state data, normal state data; The abnormal state types include: current mutation, frequency deviation, and voltage fluctuation.

2. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal according to claim 1, characterized in that: The S20 includes: S21: dividing the missing data in the original data set into recent missing data and common missing data, using timestamp alignment interpolation to fill the recent missing data, and using Lagrange interpolation to fill the common missing data, to obtain a filled data set; S22: performing data denoising on the padded data set, performing denoising on the time domain coded signal by using a moving average filter, and performing denoising on the frequency domain coded signal by using a Butterworth low-pass filter, to obtain a denoised padded data set.

3. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal according to claim 1, characterized in that: The S30 includes: S31: converting the power signal data in the denoised and padded data set from the time domain to the frequency domain through fast Fourier transform to obtain energy distribution information at different frequencies; S32: Performing spectrum analysis on the energy distribution information, and extracting spectrum features according to energy corresponding frequencies; S33: decomposing the time series signal in the denoising and padding data set into wavelet coefficients of different scales and frequencies through wavelet transform, calculating the mean, variance, skewness, and kurtosis of the wavelet coefficients, and obtaining time domain features; S34: Calculate the total signal energy of the time domain signal in the denoised and padded data set, and perform spectrum density integration on the frequency domain signal in the denoised and padded data set to obtain energy features.

4. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal according to claim 1, characterized in that: The S40 comprises: S41: performing feature fusion on the spectrum features, time domain features, and energy features to obtain a fused feature vector; S42: Randomly divide the partial fusion feature vectors into a training set and a test set according to a ratio of 8:

2.

5. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal according to claim 1, characterized in that: The S50 comprises: S51: Determine the number of input layer nodes of the neural network classification model according to the dimension of the fused feature vector; S52: passing through three hidden layers, wherein the hidden layers are fully connected layers and use a ReLU activation function; S53: The output layer uses the Sigmoid activation function to divide the output results into abnormal state and normal state; S54: using the training set to train the neural network classification model to obtain a state classification model; S55: Optimizing the state classification model through the test set; S56: Input the fused feature vector into a state classification model to obtain abnormal state data and normal state data.

6. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal according to claim 1, characterized in that: The S60 includes: S61: According to the characteristics of low-voltage distribution network, three types of abnormal conditions are defined, including: current mutation, frequency deviation, and voltage fluctuation; S62: Calculate the local density of the abnormal state data and its neighborhood data using a local abnormal factor, average the ratio of the local density of the neighborhood data to the local density of the abnormal state data, and obtain the LOF value of the abnormal state data; S63: Identify the type causing the abnormal state according to the LOF value of the abnormal state data.

7. A method for detecting the operating status of a low-voltage distribution network based on a fusion terminal according to claim 1, characterized in that: The S70 includes: S71: Five key indicators are defined based on the characteristics of low-voltage distribution networks, including: power supply radius, comprehensive line loss rate, transformer load rate, user voltage qualification rate, and three-phase load imbalance; S72: Setting power supply radius constraint conditions of low voltage distribution lines; S73: solving the comprehensive line loss rate of the low voltage distribution network; S74: solving the transformer load factor of the low voltage distribution network; S75: Calculate the voltage qualification rate of the low voltage distribution network; S76: solving the three-phase load imbalance of the low-voltage distribution network; S77: Calculate the weights of the five key indicators using the Lagrange optimal multiplier method to obtain a health assessment function, and perform a comprehensive health evaluation of the low-voltage distribution network on the normal state data based on the health assessment function.

Images