Power distribution network fault detection method applicable to distributed power supply

Abstract

The present invention relates to the field of power distribution network fault detection, and in particular to a power distribution network fault detection method applicable to a distributed power supply. Potential fault modes and rules are extracted from a large amount of real-time monitoring data by using multiple data models, achieving accurate detection and diagnosis of various fault types; data collected in real time is processed and analyzed, so that fault signals can be found in a timely manner and corresponding alarm and notification mechanisms can be triggered, improving the response speed and processing efficiency of faults; continuous data iteration and model updating are performed, so that the change of a system and the condition of different types of faults can be adapted, improving the robustness and accuracy of detection; and faults are detected and diagnosed in a timely manner, so that a power grid operation strategy is effectively optimized, the energy utilization efficiency is improved, and the energy consumption and the operation and maintenance costs are reduced.

Description

A distribution network fault detection method suitable for distributed power generation Technical Field

[0001] The present invention relates to the field of distribution network fault detection, and in particular to a distribution network fault detection method applicable to distributed power sources. Background Art

[0002] With the development of renewable energy technology and the improvement of environmental awareness, the scale and proportion of distributed power sources in distribution networks are constantly increasing. Traditional centralized power sources are gradually shifting to distributed power sources. The introduction of distributed power sources has led to more types of faults in distribution networks, requiring accurate detection and diagnosis based on the characteristics of distributed power sources.

[0003] As an energy carrier directly connected to power users in the power system, the safe and reliable operation of the distribution network plays an important role in the normal operation of the entire power system. When a distribution network fails, it will affect the power consumption of power users and even cause a failure in the power system. Therefore, it is necessary to find the fault and solve it as soon as possible. The existing distribution network fault detection method usually adopts manual inspection and fault alarm methods, which are inefficient and cannot solve problems encountered in daily work and life in a timely manner.

[0004] Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a distribution network fault detection method suitable for distributed power sources. By monitoring the current, voltage and frequency fault risk points, the deviation between the current signal peak and the normal expected peak is calculated, the voltage variation rate is calculated to evaluate the instantaneous voltage variation, and the difference value range between the real-time frequency and the normal frequency is calculated to solve the problems raised in the above-mentioned background technology.

[0006] The technical solution adopted by the present invention is a distribution network fault detection method applicable to distributed power sources, comprising the following steps:

[0007] 101. Use smart meters and sensors to form a data acquisition unit to collect multi-dimensional power parameter data such as current, voltage and frequency in real time, use normalization algorithm to pre-process the data and use mean-variance method to monitor and correct abnormal data;

[0008] 102. Use feature extraction algorithms to extract fault features, extract current standard deviation features to reflect current anomalies, use band-stop filters to separate harmonic components, extract voltage total harmonic distortion rate features, set frequency offset reference benchmarks, and determine whether the frequency offset exceeds the normal range;

[0009] 103. Calculate the difference between the observed data and the normal distribution, establish a decision tree classification model, recursively differentiate through decision rules, and identify and classify distribution network faults through model output;

[0010] 104. Identify current, voltage, and frequency fault risk monitoring points, calculate the deviation between the current signal peak and the normal expected peak, calculate the voltage variation rate to evaluate the instantaneous voltage variation, calculate the difference value range between the real-time frequency and the normal frequency, and use the magnitude of the difference beyond the normal range to classify the fault level;

[0011] 105. Determine the specific cause of the fault based on the output of the model, and use the fault location algorithm in combination with the output of the model to locate the specific location of the fault;

[0012] 106. Connect sensors and monitoring equipment in each distributed power source and distribution network, trigger the alarm mechanism through visual display based on the results of fault diagnosis and location and the cloud data center test report, monitor the operating status and parameter settings of the equipment in real time, and perform remote configuration, upgrade, and diagnosis operations.

[0013] In a preferred embodiment, in step 101, a data acquisition unit is formed using a smart meter and sensors to collect multi-dimensional power parameter data such as current, voltage, and frequency in real time, and transmit the data to a central monitoring system via wireless communication for processing and analysis. Specifically, the following steps are included:

[0014] Step 1: Install smart meters and sensors at different nodes of distributed power sources, use smart meters and sensors to form data acquisition units, integrate communication and control information technology to deploy key nodes, establish a smart grid system, integrate communication and information exchange between different devices, control the comprehensive processing and scheduling of distributed power sources, collect multi-dimensional power parameter data such as current, voltage and frequency in real time, and transmit them to the central monitoring system for processing and analysis via wireless communication.

[0015] Step 2: Normalize the data obtained by the central monitoring system and perform abnormal data monitoring correction so that data of different dimensions have the same scale. The minimum-maximum normalization formula is:

[0016] Where Z represents the normalized data, x represents the original data, and x min Represents the minimum value of the original data, x max Represents the maximum value of the original data. The normalized data is identified as outliers through an anomaly detection algorithm, and corrected and eliminated according to the specific situation. The mean-variance method is used to perform anomaly detection analysis on the data. The above mean-variance method judges whether the data deviates from the normal based on the sample mean and standard deviation. The specific formula is:

[0017] Where n represents the number of data, S represents the mean of the data samples, and C represents the variance of the data samples. For each data point, the difference between it and the mean is calculated, and then compared with the standard deviation of the mean. When the difference exceeds the set threshold, the data point is marked as an outlier. The judgment data set is traversed and the data that meets the definition of outliers are marked. Obvious outliers are directly eliminated and replaced by interpolation. The interpolation method performs linear interpolation between the values ​​before and after the outlier according to the linear trend of the surrounding data.

[0018] In a preferred embodiment, in step 102, based on the data transmitted by the central monitoring system, a feature extraction algorithm is used to extract fault features. By monitoring the current data, voltage data, and frequency data in the power grid, current anomalies, voltage fluctuations, and frequency offsets are detected. The current standard deviation feature is extracted to reflect the current anomaly. The specific formula for the standard deviation is:

[0019] Where n represents the number of current data, S represents the mean value of the data samples, and C represents the variance value of the data samples. The harmonic components are separated by a band-stop filter to extract the voltage total harmonic distortion rate characteristics. The specific formula of the total harmonic distortion rate is:

[0020] Where T represents the total harmonic distortion rate of voltage, H represents the effective value of harmonic components, and x i Represents the voltage value at each sample point, and n represents the number of voltage data samples. It is used to analyze voltage fluctuation and stability, set the frequency offset reference benchmark, and determine whether the frequency offset exceeds the normal range.

[0021] In a preferred embodiment, in the above step 103, characteristic values ​​of current, voltage, and frequency are collected, and a corresponding fault mode recognition model is established according to the fault type of the distributed power system. The above fault types include current anomaly, voltage fluctuation, and frequency offset. The above fault mode recognition model is based on outlier detection of the normal distribution. By assuming that the observed data under normal conditions obeys the normal distribution, the difference between the observed data and the normal distribution is calculated. The specific calculation formula is:

[0022] Where Z represents the degree of deviation of the observed data point from the normal distribution, X represents the value of the observed current, voltage, and frequency data points, μ represents the mean of the normal distribution, and σ represents the standard deviation of the normal distribution. A threshold is set based on the degree of deviation to determine whether there are outliers for fault identification. A data set of known fault types is collected, including feature parameters and corresponding fault labels. Based on domain knowledge and data analysis, feature parameters related to the fault mode are selected to establish a decision tree classification model. The specific formula of the decision tree classification model is: L = f k (Ai ,θ i )

[0023] Where L represents the fault category of the decision tree node, k represents the number of feature vectors, and A i represents the feature vector of the i-th non-leaf node, θ i represents the threshold of the i-th non-leaf node, f k (x) represents the recursive loop function of the left and right child nodes, which is recursively differentiated through the decision rule, and the predicted classification result is the fault category of the leaf node. The decision rule is to compare A i ,θ i The fault characteristics are judged to enter the left and right node positions, and the data set is divided into a training set, a validation set, and a test set. The model is trained according to the training set, and the hyperparameters are tuned through the validation set. The performance of the model is evaluated by cross-validation using the test set. The cross-validation uses equal data subsets as the training set and the validation set to evaluate the model performance. The average value of the evaluation results is repeated multiple times as the model performance indicator. The extracted feature parameters are input into the model, and the distribution network faults are identified and classified through the model output.

[0024] In a preferred embodiment, in step 104, the current, voltage, and frequency fault risk monitoring points are divided according to the extracted features and the actual situation, and different fault levels are defined using the fault index. Specifically, the following steps are included:

[0025] Step 1: At the current fault risk monitoring point, the fault index calculation algorithm is used to obtain the fault index. The cloud data center specifies the normal expected value, traverses the current characteristic data points, compares the size relationship between the current data point and its adjacent data points, and determines whether the current data point is a peak point after iterative comparison if the current data point is greater than the previous data point and the next data point. The deviation between the current signal peak value and the normal expected peak value is calculated. The specific formula for calculating the peak deviation is: P = P a -P b

[0026] Where P represents the peak deviation, P a Indicates the current signal data peak value, P b It indicates the normal expected peak value, evaluates the fault severity, and uses the fault index to define the current fault level, including mild fluctuation fault, moderate shunt fault, and severe leakage fault.

[0027] Step 2: At the voltage fault risk monitoring point, traverse the voltage characteristic data points to obtain the voltage waveform value. The cloud data center obtains the voltage waveform reference value and calculates the voltage variation rate to evaluate the instantaneous voltage variation. The specific formula for calculating the voltage variation rate is:

[0028] Where TV represents the voltage change rate, V max Indicates the maximum value of the voltage waveform, V min Indicates the minimum value of the voltage waveform, V n It indicates the voltage fluctuation reference value, characterizes voltage instability, and uses the voltage variation rate to determine the voltage fault level, including short-term voltage variation and long-term voltage variation.

[0029] Step 3: At the frequency failure risk monitoring point, the cloud data center specifies a baseline value for the normal frequency, obtains the system's real-time frequency data, and calculates the difference between the real-time frequency and the normal frequency. The specific calculation formula is: F = |F r -F n |

[0030] Where F represents the frequency variation range, F r Indicates the real-time frequency, F n Indicates normal frequency. Set a threshold to determine whether the frequency fault change exceeds the normal range. The fault level is divided according to the degree of exceeding the normal range, including mild frequency jitter fault, moderate frequency imbalance fault, and severe frequency loss fault.

[0031] In a preferred embodiment, in step 105, the possible fault types are analyzed based on the output of the model, and the specific cause of the fault is determined in combination with the actual situation. The possible fault locations are analyzed based on the output of the model and the system topology. The fault location algorithm is specifically formulated as follows: S = V * T

[0032] Where S represents the distance between the fault location and the current, voltage, and frequency measurement points; V represents the transmission speed of the current, voltage, and frequency signals; and T represents the time it takes to transmit the current, voltage, and frequency measurement points. Combined with the model output, the specific location of the fault is located. Based on the specific cause of the fault, corresponding repair suggestions are provided and corresponding repair measures are implemented, including adjusting the distribution network configuration and repairing the network connection. The test fault location is determined, and the cloud data center records the test results, including the test steps, test data, and test conclusions.

[0033] In a preferred embodiment, in the above step 106, a wireless communication protocol is used to connect sensors and monitoring equipment in each distributed power supply and distribution network, and real-time monitoring is performed to obtain relevant data of the distributed power supply distribution network, including power supply current, voltage and frequency. Data is exchanged with the cloud data center to obtain test reports and fault analysis results. According to the results of the fault diagnosis and positioning and the cloud data center test report, an alarm mechanism is triggered through a visual display, and the distributed power supply and distribution network are remotely operated and controlled by the remote monitoring center. The operating status and parameter settings of the equipment are monitored in real time, and remote configuration, upgrade, and diagnostic operations are performed. The power supply is remotely switched and maintenance measures for standby equipment are dispatched.

[0034] Technical effects and advantages of the present invention:

[0035] The present invention utilizes multiple data models to extract potential failure modes and patterns from a large amount of real-time monitoring data, thereby achieving accurate detection and diagnosis of various types of failures. By processing and analyzing the real-time collected data, it is possible to promptly detect failure signals and trigger corresponding alarm and notification mechanisms, thereby improving the response speed and processing efficiency to failures. Adaptive adjustments are made according to the operating status of the distribution network and environmental changes. Through continuous data iteration and model updates, it is possible to adapt to system changes and different types of failure conditions, thereby improving the robustness and accuracy of detection. By promptly detecting and diagnosing failures, the grid operation strategy is effectively optimized, energy utilization efficiency is improved, and energy consumption and operation and maintenance costs are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] FIG1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0037] 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.

[0038] Example 1

[0039] This embodiment provides a distribution network fault detection method applicable to a distributed power supply as shown in FIG1 , which specifically includes the following steps:

[0040] 101. Use smart meters and sensors to form a data acquisition unit to collect multi-dimensional power parameter data such as current, voltage and frequency in real time, use normalization algorithm to pre-process the data and use mean-variance method to monitor and correct abnormal data;

[0041] Furthermore, a data acquisition unit is formed by using smart meters and sensors to collect multi-dimensional power parameter data such as current, voltage and frequency in real time, and transmit it to the central monitoring system through wireless communication for processing and analysis. The specific content includes the following steps:

[0042] Step 1: Install smart meters and sensors at different nodes of distributed power sources, use smart meters and sensors to form data acquisition units, integrate communication and control information technology to deploy key nodes, establish a smart grid system, integrate communication and information exchange between different devices, control the comprehensive processing and scheduling of distributed power sources, collect multi-dimensional power parameter data such as current, voltage and frequency in real time, and transmit them to the central monitoring system for processing and analysis via wireless communication.

[0043] Step 2: Normalize the data obtained by the central monitoring system and perform abnormal data monitoring and correction so that data of different dimensions have the same scale. The above minimum-maximum normalization formula is:

[0044] Where Z represents the normalized data, x represents the original data, and x min Represents the minimum value of the original data, x max Represents the maximum value of the original data. The normalized data is identified as outliers through an anomaly detection algorithm, and corrected and eliminated according to the specific situation. The mean-variance method is used to perform anomaly detection analysis on the data. The above mean-variance method judges whether the data deviates from the normal based on the sample mean and standard deviation. The specific formula is:

[0045] Where n represents the number of data, S represents the mean of the data samples, and C represents the variance of the data samples. For each data point, the difference between it and the mean is calculated, and then compared with the standard deviation of the mean. When the difference exceeds the set threshold, the data point is marked as an anomaly. The judgment data set is traversed and the data that meets the definition of an outlier are marked. Obvious outliers are directly eliminated and replaced by interpolation. The above interpolation method performs linear interpolation between the values ​​before and after the outlier according to the linear trend of the surrounding data.

[0046] 102. Use feature extraction algorithms to extract fault features, extract current standard deviation features to reflect current anomalies, use band-stop filters to separate harmonic components, extract voltage total harmonic distortion rate features, set frequency offset reference benchmarks, and determine whether the frequency offset exceeds the normal range;

[0047] Furthermore, based on the data transmitted by the central monitoring system, a feature extraction algorithm is used to extract fault features. By monitoring the current data, voltage data, and frequency data in the power grid, current anomalies, voltage fluctuations, and frequency offsets are detected. The current standard deviation feature is extracted to reflect current anomalies. The specific formula for the above current standard deviation is:

[0048] Where n represents the number of current data, S represents the mean value of the data samples, and C represents the variance value of the data samples. The harmonic components are separated by a band-stop filter to extract the voltage total harmonic distortion rate characteristics. The specific formula for the above total harmonic distortion rate is:

[0049] Where T represents the total harmonic distortion rate of voltage, H represents the effective value of harmonic components, and x i Represents the voltage value at each sample point, and n represents the number of voltage data samples. It is used to analyze voltage fluctuation and stability, set the frequency offset reference benchmark, and determine whether the frequency offset exceeds the normal range.

[0050] 103. Calculate the difference between the observed data and the normal distribution, establish a decision tree classification model, recursively differentiate through decision rules, and identify and classify distribution network faults through model output;

[0051] Furthermore, characteristic values ​​of current, voltage, and frequency are collected, and a corresponding fault pattern recognition model is established according to the fault type of the distributed power system. The above-mentioned fault types include current anomaly, voltage fluctuation, and frequency deviation. The above-mentioned fault pattern recognition model is based on outlier detection of the normal distribution. By assuming that the observed data under normal conditions obeys the normal distribution, the difference between the observed data and the normal distribution is calculated. The calculation formula is:

[0052] Where Z represents the degree of deviation of the observed data point from the normal distribution, X represents the value of the observed current, voltage, and frequency data points, μ represents the mean of the normal distribution, and σ represents the standard deviation of the normal distribution. A threshold is set based on the degree of deviation to determine whether there are outliers for fault identification. A data set of known fault types is collected, including feature parameters and corresponding fault labels. Based on domain knowledge and data analysis, feature parameters related to the fault mode are selected to establish a decision tree classification model. The calculation formula of the decision tree classification model is: L = f k (A i ,θ i )

[0053] Where L represents the fault category of the decision tree node, k represents the number of feature vectors, and A i represents the feature vector of the i-th non-leaf node, θ i represents the threshold of the i-th non-leaf node, fk (x) represents the recursive loop function of the left and right child nodes, which is recursively differentiated through the decision rule, and the predicted classification result is the fault category of the leaf node. The decision rule is to compare A i ,θ i Determine the position of the left and right nodes where the fault features enter, divide the data set into training set, validation set and test set, train the model according to the training set, and tune the hyperparameters through the validation set. Use the test set to use cross-validation to evaluate the performance of the model. Cross-validation uses equal data subsets as training sets, and uses the validation set to evaluate the model performance. Repeat the process multiple times and use the average value of the evaluation results as the model performance indicator. Input the extracted feature parameters into the model, and identify and classify distribution network faults through the model output.

[0054] 104. Identify current, voltage, and frequency fault risk monitoring points, calculate the deviation between the current signal peak and the normal expected peak, calculate the voltage variation rate to evaluate the instantaneous voltage variation, calculate the difference value range between the real-time frequency and the normal frequency, and use the magnitude of the difference beyond the normal range to classify the fault level;

[0055] Further, based on the extracted features and actual conditions, the current, voltage, and frequency fault risk monitoring points are divided, and different fault levels are defined using the fault index. The specific steps include the following:

[0056] Step 1: At the current fault risk monitoring point, the fault index calculation algorithm is used to obtain the fault index. The cloud data center specifies the normal expected value, traverses the current characteristic data points, compares the size relationship between the current data point and its adjacent data points, and determines whether the current data point is a peak point after iterative comparison if the current data point is greater than the previous data point and the next data point. The deviation between the current signal peak value and the normal expected peak value is calculated. The specific formula for calculating the peak deviation is: P = P a -P b

[0057] Where P represents the peak deviation, P a Indicates the current signal data peak value, P b It indicates the normal expected peak value, evaluates the fault severity, and uses the fault index to define the current fault level, including mild fluctuation fault, moderate shunt fault, and severe leakage fault.

[0058] Step 2: At the voltage fault risk monitoring point, traverse the voltage characteristic data points to obtain the voltage waveform value. The cloud data center obtains the voltage waveform reference value and calculates the voltage variation rate to evaluate the instantaneous voltage variation. The specific formula for calculating the voltage variation rate is:

[0059] Where TV represents the voltage change rate, V max Indicates the maximum value of the voltage waveform, Vmin Indicates the minimum value of the voltage waveform, V n It indicates the voltage fluctuation reference value, characterizes voltage instability, and uses the voltage variation rate to determine the voltage fault level, including short-term voltage variation and long-term voltage variation.

[0060] Step 3: At the frequency failure risk monitoring point, the cloud data center specifies a baseline value for the normal frequency, obtains the system's real-time frequency data, and calculates the difference between the real-time frequency and the normal frequency. The specific calculation formula is: F = |F r -F n |

[0061] Where F represents the frequency variation range, F r Indicates the real-time frequency, F n Indicates normal frequency. Set a threshold to determine whether the frequency fault change exceeds the normal range. The fault level is divided according to the degree of exceeding the normal range, including mild frequency jitter fault, moderate frequency imbalance fault, and severe frequency loss fault.

[0062] 105. Determine the specific cause of the fault based on the output of the model, and use the fault location algorithm in combination with the output of the model to locate the specific location of the fault;

[0063] Furthermore, based on the results of the model output, the possible fault types are analyzed, and the specific causes of the faults are determined in combination with the actual situation. Based on the results of the model output and the system topology, the possible fault locations are analyzed, and the fault location algorithm is used. The specific formula of the fault location algorithm is: S = V * T

[0064] Where S represents the distance between the fault location and the current, voltage, and frequency measurement points; V represents the transmission speed of the current, voltage, and frequency signals; and T represents the time it takes to transmit the current, voltage, and frequency measurement points. Combined with the model output, the specific location of the fault is located. Based on the specific cause of the fault, corresponding repair suggestions are provided and corresponding repair measures are implemented, including adjusting the distribution network configuration and repairing network connections, to determine the test fault location.

[0065] 106. Connect sensors and monitoring equipment in each distributed power source and distribution network, trigger alarm mechanisms through visual display based on the results of fault diagnosis and location and cloud data center test reports, monitor the operating status and parameter settings of the equipment in real time, and perform remote configuration, upgrade, and diagnosis operations;

[0066] Furthermore, wireless communication protocols are used to connect sensors and monitoring equipment in various distributed power sources and distribution networks, and relevant data of the distributed power distribution network, including power current, voltage and frequency, is monitored in real time. Data is exchanged with the cloud data center to obtain test reports and fault analysis results. According to the results of fault diagnosis and positioning and the test report of the cloud data center, an alarm mechanism is triggered through visual display, and the distributed power source and distribution network are remotely operated and controlled by the remote monitoring center. The operating status and parameter settings of the equipment are monitored in real time, and remote configuration, upgrade, and diagnosis operations are performed. The power supply is remotely switched and maintenance measures for standby equipment are dispatched.

[0067] The formula in the present invention is a formula that is closest to the actual situation obtained by removing the dimension and taking its numerical calculation by collecting a large amount of data and performing software simulation. The preset proportional coefficient in the formula is set by technical personnel in this field according to the actual situation or obtained through large-scale data simulation.

[0068] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

[0069] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A distribution network fault detection method applicable to distributed power sources, characterized in that: The following steps are involved:

101. Use the data acquisition unit to collect the multi-dimensional power parameter data of current, voltage and frequency in real time, and perform pre-processing and abnormal data monitoring and correction on the data; 102. Extract fault features, extract current standard deviation features to reflect current abnormalities, use band-stop filters to separate harmonic components, extract voltage total harmonic distortion rate features, set frequency offset reference benchmarks, and determine whether the frequency offset exceeds the normal range; 103. Calculate the difference between the observed data and the normal distribution, establish a decision tree classification model, recursively differentiate through decision rules, and identify and classify distribution network faults through model output; 104. Divide the current, voltage and frequency fault risk monitoring points, calculate the deviation between the current signal peak and the normal expected peak, calculate the voltage change rate to evaluate the instantaneous voltage change, calculate the difference value range between the real-time frequency and the normal frequency, and divide the fault level according to the size beyond the normal range; 105. Determine the specific cause of the fault based on the output of the model, and use the fault location algorithm in combination with the output of the model to locate the specific location of the fault; 106. Connect the sensors and monitoring equipment in each distributed power source and distribution network, trigger the alarm mechanism according to the results of fault diagnosis and positioning, monitor the operating status and parameter settings of the equipment in real time, and perform remote configuration, upgrade and diagnosis operations.

2. A distribution network fault detection method applicable to distributed power sources according to claim 1, characterized in that: In step 101, a data acquisition unit is formed by using a smart meter and a sensor, and a normalization algorithm is used to pre-process the data and a mean-variance method is used to monitor and correct abnormal data. The normalization algorithm is minimum-maximum normalization, and its formula is: Where Z represents the normalized data, x represents the original data, and x min Represents the minimum value of the original data, x max Represents the maximum value of the original data, and the formula of the mean variance method is: Where n represents the number of data, S represents the mean of the data sample, and C represents the variance of the data sample.

3. A distribution network fault detection method applicable to distributed power sources according to claim 1, characterized in that: In step 102, the current standard deviation is calculated as: Where n represents the number of current data, S represents the mean value of data samples, and C represents the variance value of data samples; The total harmonic distortion formula is: Where T represents the total harmonic distortion rate of voltage, H represents the effective value of harmonic components, and x i represents the voltage value of each sample point, and n represents the number of samples of voltage data.

4. The method for detecting a fault in a distribution network applicable to a distributed power source according to claim 1, characterized in that: In step 103, a corresponding fault pattern recognition model and a decision tree classification model are established. Fault types include current anomalies, voltage fluctuations, and frequency deviations. A threshold is set according to the degree of deviation to determine whether there are outliers for fault identification. The decision rules are recursively differentiated to predict the classification results as the fault categories of leaf nodes. Distribution network faults are identified and classified through model output.

5. A distribution network fault detection method applicable to distributed power sources according to claim 4, characterized in that: The fault mode recognition model is based on outlier detection of normal distribution, and its specific calculation formula is: Where Z represents the degree of deviation of the observed data point relative to the normal distribution, X represents the value of the observed current, voltage and frequency data points, μ represents the mean of the normal distribution, σ represents the standard deviation of the normal distribution, and the specific formula of the decision tree classification model is: L=f k (A i ,i i ) Where L represents the fault category of the decision tree leaf node, k represents the number of feature vectors, and A i represents the feature vector of the i-th non-leaf node, θ i represents the threshold of the i-th non-leaf node, f k (x) represents the recursive loop function of the left and right child nodes.

6. A distribution network fault detection method applicable to distributed power sources according to claim 1, characterized in that: In step 104, the calculation formula for the deviation between the current signal peak value and the normal expected peak value is: P = P a -P b Where P represents the peak deviation, P a Indicates the current signal data peak value, P b Indicates the normal expected peak value, and the voltage change rate calculation formula is: Where TV represents the voltage change rate, V max Indicates the maximum value of the voltage waveform, V min Indicates the minimum value of the voltage waveform, V n Indicates the voltage wave reference value, the difference range between the real-time frequency and the normal frequency is calculated as follows: F=|F r -F n | Where F represents the frequency variation range, F r Indicates the real-time frequency, F n Indicates normal frequency.

7. A distribution network fault detection method applicable to distributed power sources according to claim 1, characterized in that: In step 105, a fault location algorithm is used in combination with the results of the model output to locate the specific location where the fault occurs according to the results of the model output and the system topology.

8. A distribution network fault detection method applicable to distributed power sources according to claim 7, characterized in that: The specific formula of the fault location algorithm is: S = V*T Where S represents the distance between the fault location and the current, voltage and frequency measurement points, V represents the current, voltage and frequency signal transmission speed, and T represents the time to transmit the current, voltage and frequency measurement points.

9. The method for detecting a fault in a distribution network applicable to a distributed power source according to claim 1, characterized in that: In step 106, sensors and monitoring equipment in each distributed power source and distribution network are connected, and based on the results of fault diagnosis and location and the cloud data center test report, an alarm mechanism is triggered through a visual display, and the remote monitoring center is used to monitor the operating status and parameter settings of the equipment in real time, and remote configuration, upgrade, and diagnosis operations are performed.

Images