Power distribution network health assessment method, system and equipment and storage medium

Abstract

The invention relates to a power distribution network health assessment method. The method comprises the steps of obtaining electrical parameters and environmental parameters in a power distribution network in a to-be-monitored time period; identifying abnormal values in the electrical parameters and the environmental parameters by using an isolated forest model, and extracting the abnormal values to obtain to-be-analyzed monitoring data; and inputting the to-be-analyzed monitoring data into a health evaluation model for health evaluation, and evaluating the health condition of the power distribution network according to a health index output by the health evaluation model. The abnormal value is removed by using the isolated forest model, and the health index is obtained by using the health assessment model, so that the health assessment of the whole power distribution network is realized, and the assessment accuracy is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent distribution transformer terminals, and specifically to a method, system, device and storage medium for power distribution network health assessment. Background Art

[0002] In the power system, as a key link in power transmission, the stability of the operation state and the health status of the distribution network directly affect the overall reliability of the power grid. The traditional monitoring method of the distribution network mainly relies on regular manual inspections, simple monitoring devices and supporting master stations, and there are many deficiencies in this method. With the expansion of the scale and the complexity of the structure of the distribution network, there is an urgent need for a system that can monitor, analyze the operation state of the power grid in real time and conduct health assessment to improve the operation reliability and safety of the distribution network.

[0003] At present, some studies have proposed to use intelligent monitoring terminals and big data analysis technology to improve the monitoring ability of the power grid. For example, patent document CN119209918A proposes an intelligent power grid monitoring system, but this system can only monitor some key nodes of the power grid and cannot give an accurate health condition evaluation according to different device scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for power distribution network health assessment to solve the problems raised in the above background art.

[0005] The first aspect of the present invention provides a method for power distribution network health assessment, including:

[0006] S1. Obtain the electrical parameters and environmental parameters in the power distribution network during the period to be monitored;

[0007] S2. Apply the Isolation Forest model to identify the outliers in the electrical parameters and the environmental parameters. After extracting the outliers, obtain the monitoring data to be analyzed;

[0008] S3. Input the monitoring data to be analyzed into the health assessment model for health evaluation, and evaluate the health condition of the power distribution network according to the health index output by the health assessment model;

[0009] Among them, the health assessment model is established by applying the regression analysis method or the neural network algorithm, and voltage, current and temperature are used as characteristic parameters as the input data for training the health assessment model;

[0010] When the regression analysis method is adopted, the health index H is obtained by the following formula:

[0011]

[0012] Where: β0 is the intercept, β i is the regression coefficient, xi where \(i\) is the value of the characteristic parameter and \(n\) is the total number of characteristics;

[0013] When using the neural network algorithm, the health index \(H\) is obtained by the following formula:

[0014] \(H = f(X; w)\)

[0015] \(X = [x_1, x_2, \cdots, x_{ n \), where \(X\) is the characteristic parameter, \(w\) is the weight parameter in the network, and \(f\) is the activation function.

[0016] In a possible implementation, the step S2 includes:

[0017] Input the electrical parameters and the environmental parameters as data points into the trained isolation forest model. The isolation forest model makes predictions on the data points. When the prediction result is positive, the data point is a normal point; when the prediction result is negative, the data point is a to-be-determined point;

[0018] Apply the isolation forest model to score each to-be-determined point according to the distance between the to-be-determined point and the other points. If the score exceeds the anomaly threshold, the to-be-determined point is an outlier; otherwise, it is a normal value.

[0019] In a possible implementation, after the step S1, it further includes:

[0020] Perform noise reduction processing on the electrical parameters and the environmental parameters to remove the noise data in the electrical parameters and the environmental parameters.

[0021] In a possible implementation, the health assessment model is established using the neural network algorithm and includes:

[0022] Determine the neural network algorithm for establishing the health assessment model according to the parameter type of the data in the distribution network.

[0023] In a possible implementation, the determining the neural network algorithm for establishing the health assessment model according to the parameter type of the data in the distribution network includes:

[0024] When the parameter type is non-linear, use a feedforward neural network;

[0025] When the parameter type is image data, use a convolutional neural network;

[0026] When the parameter type is time series data, use a recurrent neural network.

[0027] In a possible implementation, the evaluating the health condition of the distribution network according to the health index output by the health assessment model includes:

[0028] Compare the health index with a preset health status range to obtain the health status range into which the health index falls, and label the real-time health status of the distribution network with the label of this health status range.

[0029] The second aspect of the present invention provides a distribution network health assessment system, including:

[0030] An acquisition module, configured to obtain the physical addresses of FTU devices to be software-updated, and number each FTU device according to the physical addresses;

[0031] A processing module, configured to identify outliers in the electrical parameters and the environmental parameters by applying an isolation forest model, and obtain the monitoring data to be analyzed after extracting the outliers;

[0032] An analysis module, configured to input the monitoring data to be analyzed into a health assessment model for health evaluation, and evaluate the health condition of the distribution network according to the health index output by the health assessment model.

[0033] The third aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the distribution network health assessment method described in the first aspect of the present invention is implemented.

[0034] The fourth aspect of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the distribution network health assessment method described in the first aspect of the present invention is implemented.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] 1. By using an isolation forest model to remove outliers and using the health index H obtained from the health assessment model, the health assessment of the entire distribution network is realized, and the assessment accuracy is significantly improved.

[0037] 2. Through intelligent health assessment and fault prediction, this system can identify potential faults in advance and send out warning signals. Compared with the traditional system that can only respond after a fault occurs, the warning mechanism of this system is more forward-looking, can effectively reduce the probability of faults occurring, and ensure the safe operation of the distribution network.

[0038] 3. Through real-time monitoring, accurate assessment and early warning, the safety and reliability of the distribution network operation are significantly improved, the risk of large-scale power outages caused by equipment failures is reduced, and the stability of power supply is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1Schematic flow diagram of the distribution network health assessment method of the present invention;

[0040] Figure 2 Schematic structural diagram of the distribution network health assessment system of the present invention;

[0041] Figure 3 Schematic diagram of a computer device in an embodiment of the present invention. Detailed implementation manners

[0042] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0043] It should be noted that the serial numbers assigned to the components in the embodiments of the present invention itself, such as "first", "second", etc., are only used to distinguish the described objects and do not have any sequential or technical meanings.

[0044] As Figure 1 shown, a distribution network health assessment method includes:

[0045] S1. Obtain electrical parameters and environmental parameters in the distribution network during the period to be monitored;

[0046] Specifically, in the present invention, a distribution switch monitoring terminal (FTU) installed at key nodes of the distribution network is used to collect electrical parameters (such as voltage, current, power) and environmental parameters (such as temperature, humidity) of the power grid in real time. The FTU is equipped with high-precision sensors and an embedded data processor. Then, electrical parameters and environmental parameters are obtained from the FTU device by using Ethernet, optical fiber communication or wireless communication (such as 4G / 5G).

[0047] After obtaining the electrical parameters and environmental parameters, it is necessary to perform noise reduction processing on these two parameters to remove noise so as not to affect the accuracy of subsequent distribution network health assessment. Among them, the noise mainly comes from the following two aspects:

[0048] Sensor error: The distribution switch monitoring terminal (FTU) obtains electrical parameters through current or voltage sensors. If the sensor is affected by electromagnetic interference or hardware failure, some random and irregular voltage fluctuation data may be generated. For example, when the actual voltage is 220V, the sensor may occasionally report 225V or 215V, and this kind of fluctuation belongs to noise.

[0049] Communication interference: During the data transmission process in the power grid, if the transmission line is affected by the external environment (such as lightning, electromagnetic waves), it may lead to packet loss or errors. The data generated in this way may show irregular sudden changes and belong to noise.

[0050] S2. Use the Isolation Forest model to identify outliers in the electrical parameters and the environmental parameters. After extracting the outliers, the monitoring data to be analyzed is obtained.

[0051] Specifically, take the electrical parameters and the environmental parameters as data points and input them into the trained Isolation Forest model. The Isolation Forest model predicts the data points. When the prediction result is positive, the data point is a normal point; when the prediction result is negative, the data point is a suspicious point.

[0052] Use the Isolation Forest model to score each suspicious point according to the distance between the suspicious point and the other points. If the score exceeds the outlier threshold, the suspicious point is an outlier; otherwise, it is a normal value.

[0053] Isolation Forest is an anomaly detection algorithm based on tree structure, which is specifically used to identify outliers in a dataset. Its core idea is that outliers are more likely to be isolated because they are far from other data points. This algorithm is efficient in processing high-dimensional data and does not require labeled data, so it is very suitable for detecting outliers in power grid equipment data in the power grid health assessment system.

[0054] When training the Isolation Forest model, first, collect and organize the operation data of power grid equipment (such as voltage, current, temperature, etc.). These data usually contain multiple features, such as:

[0055] Voltage (V), current (A), temperature (°C), power (W)

[0056] Preprocess this data to ensure that there are no missing values or incorrect values. For missing values, you can choose to delete or fill them.

[0057] Then, determine the parameters in the Isolation Forest model: Maximum depth (max_samples): Define the maximum depth of each tree. A larger depth can analyze the data more deeply, but it will increase the computational cost.

[0058] Outlier ratio (contamination): Set the ratio of outliers in the dataset. This ratio helps the model set the judgment criteria for outliers. Usually, it can be estimated based on historical fault data. For example, if 10% of the data is set as outliers, that is, contamination = 0.1.

[0059] Random state: Sets the seed of the random number generator to ensure reproducibility of experimental results.

[0060] Finally, start training the Isolation Forest model.

[0061] When performing outlier identification, the Isolation Forest model generates an "anomaly score" for each data point, usually in the range of -1 to 1, where -1 indicates an outlier and 1 indicates normal. According to the set contamination parameter, when the score of an outlier is usually greater than a certain threshold, the model will mark it as an outlier. If contamination is set to 0.1 (i.e., 10% of the data set are outliers), the model will automatically calculate a threshold based on this proportion. If the score of a data point is higher than this threshold, it will be judged as normal data; if it is lower than this threshold, it will be judged as an outlier.

[0062] For example, assume that the normal voltage range of the power grid is between 180V and 240V, and during a certain monitoring, the sensor suddenly reports a voltage value of 500V. The Isolation Forest algorithm will find that this 500V point is very far from other points and can be easily isolated in all trees, so its path length is short, and the model will mark it as an outlier.

[0063] Normal data: Voltage 220V, current 30A; Voltage 210V, current 35A.

[0064] Abnormal data: Voltage 500V, current 30A (outlier).

[0065] S3. Input the monitoring data to be analyzed into the health assessment model for health evaluation, and evaluate the health status of the distribution network according to the health index output by the health assessment model;

[0066] Among them, the health assessment model is established using regression analysis or neural network algorithm, and voltage, current, and temperature are used as characteristic parameters as the input data for training the health assessment model;

[0067] When using regression analysis, the health index H is obtained by the following formula:

[0068]

[0069] Where: β0 is the intercept, β i is the regression coefficient, x i is the value of the characteristic parameter i, and n is the total number of characteristics;

[0070] When using neural network algorithm, the health index H is obtained by the following formula:

[0071] H = f(X; w)

[0072] X = [x1, x2,..., x n , where X is a characteristic parameter, w is a weight parameter in the network, and f is an activation function.

[0073] By monitoring the real-time change of the health index H, the health status of the distribution network can be predicted. That is, if H gradually decreases over time, there will be problems with the health status of the distribution network. When making predictions, decision tree algorithms and others can be used.

[0074] Regression analysis is a method in statistics used to study the relationship between variables. In equipment health assessment, the relationship between equipment operating parameters and health status can be established through regression analysis. In the present invention, the following regression models can be used for regression analysis: linear regression, support vector regression (SVR), ridge regression, etc. Specifically, the regression analysis steps are as follows:

[0075] 1. Data preparation:

[0076] Collect the operating data of the equipment, such as voltage, current, temperature, equipment aging condition, etc.

[0077] Obtain the labeled dataset for comparison, such as the health status score of the equipment (e.g., health index), or the health status of the equipment obtained through expert evaluation (such as normal, slightly abnormal, severely abnormal).

[0078] 2. Feature selection:

[0079] According to actual needs, select the parameters related to the health status of the equipment (such as current, voltage, temperature, etc.) as input features.

[0080] 3. Regression model training:

[0081] Use regression methods such as linear regression, ridge regression, or support vector regression (SVR) to establish a health assessment model through the training dataset. The health assessment model determines the health index or the health status of the equipment by monitoring the operating parameters of the equipment.

[0082] In the present invention, a method of determining the neural network algorithm for establishing the health assessment model according to the parameter type of the data in the distribution network is adopted. This method can perform targeted analysis according to the data types generated by different distribution networks and different FTU devices, avoiding the situation where a single neural network cannot obtain accurate analysis results and thus cannot correctly evaluate the health status of the distribution network.

[0083] The neural networks used for different types of data are described in detail below:

[0084] When the parameter type is non - linear, a feed - forward neural network (FNN) is adopted; the feed - forward neural network is suitable for general regression problems and is a basic method for equipment health assessment. It performs non - linear transformations on data from the input layer through several hidden layers and finally outputs a health index. When the relationship between the health status of the equipment and operating parameters such as voltage and current is complex and non - linear, the FNN can accurately calculate the health index.

[0085] When the parameter type is image data, a convolutional neural network is adopted; the convolutional neural network is suitable for data with spatio - temporal correlation and is suitable for processing image data in the power grid. It automatically extracts spatial features (such as the time - series waveform of the power grid) from the data through the convolutional layer, enabling more accurate health assessment. However, compared with the feed - forward neural network, the convolutional neural network requires more computing power. It is applicable to the health assessment of power grid data with spatio - temporal structure.

[0086] When the parameter type is time - series data, a recurrent neural network is adopted. The recurrent neural network is suitable for time - series data, especially the long - term operation history data of equipment. Through feedback connections, it can learn the time - dependent relationships in equipment data, such as long - term voltage fluctuations and load changes. However, for short - term data analysis, the accuracy of the recurrent neural network will decrease. When the health status of the equipment is affected by long - term trends or periodic fluctuations (such as power grid load changes), the RNN can capture these dependencies and predict future health indices.

[0087] When evaluating the health of the distribution network according to the health index output by the health assessment model, the following method can be adopted:

[0088] Compare the health index with a preset health status interval to obtain the health status interval into which the health index falls, and label the real - time health status of the distribution network with the label of this health status interval.

[0089] The health status intervals can be determined according to the following criteria:

[0090] The intervals of the health index H are: normal (>80), attention (60 - 80), warning (40 - 60), danger (<40). When the health index of the equipment is in an abnormal interval, a warning signal corresponding to the label of the interval will be sent.

[0091] As Figure 2 shown, the second aspect of the present invention provides a distribution network health assessment system, including:

[0092] An acquisition module 10, configured to obtain the physical addresses of FTU devices to be software - updated, and number each FTU device according to the physical addresses;

[0093] A processing module 20, configured to identify outliers in the electrical parameters and the environmental parameters by applying an isolation forest model, and obtain monitoring data to be analyzed after extracting the outliers.

[0094] An analysis module 30, configured to input the monitoring data to be analyzed into a health assessment model for health evaluation, and evaluate the health condition of the distribution network according to the health index output by the health assessment model.

[0095] The distribution network health assessment system of the present invention may further include a user interface module, which can graphically display the operation status and health condition of the power grid with the health index. Users can remotely access the system through a monitoring terminal or a mobile device to view real-time monitoring data and health assessment reports, and support remote operation and parameter adjustment of devices.

[0096] In one embodiment, as Figure 3 shown, a computer device 40 is provided, including a memory 42, a processor 41, and a computer program 43 stored in the memory 42 and executable on the processor 41. When the processor 41 executes the computer program 43, the steps in the data processing method of the above embodiment are implemented. To avoid repetition, they are not described here again. Alternatively, when the processor 41 executes the computer program 43, the functions of each module in the above embodiment of the distribution network health assessment system are implemented. To avoid repetition, they are not described here again.

[0097] In one embodiment, a readable storage medium is provided. The readable storage medium stores a computer program 43. When the computer program 43 is executed by the processor 41, the steps in the data processing method of the above embodiment are implemented. To avoid repetition, they are not described here again. Alternatively, when the processor 41 executes the computer program 43, the functions of each module in the above embodiment of the data processing device are implemented. To avoid repetition, they are not described here again.

[0098] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memories (ROM), programmable ROM

[0099] (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (SynchlinK) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), among others.

[0100] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the above examples of the division of each functional unit and module are given. In actual applications, the above functions can be allocated to different functional modules, sub-modules, and units according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distribution network health assessment method, characterized in that: include: S1. Obtain electrical parameters and environmental parameters in the distribution network during the monitoring period; S2. Using an isolation forest model to identify abnormal values ​​in the electrical parameters and the environmental parameters, and extracting the abnormal values ​​to obtain monitoring data to be analyzed; S3, inputting the monitoring data to be analyzed into a health assessment model for health assessment, and evaluating the health status of the distribution network according to the health index output by the health assessment model; Wherein, the health assessment model is established by using regression analysis method or neural network algorithm, and voltage, current and temperature are used as characteristic parameters as input data for health assessment model training; When the regression analysis method is used, the health index H is obtained by the following formula: Among them: β0 is the intercept, β i is the regression coefficient, x i is the value of feature parameter i, n is the total number of features; When the neural network algorithm is used, the health index H is obtained by the following formula: H = f(X; w) X=[x1,x2,...,x n ], X is the feature parameter, w is the weight parameter in the network, and f is the activation function.

2. The distribution network health assessment method according to claim 1, characterized in that: The step S2 comprises: The electrical parameters and the environmental parameters are input as data points into a trained isolation forest model, and the isolation forest model predicts the data points. When the prediction result is a positive number, the data point is a normal point, and when the prediction result is a negative number, the data point is a pending point; The isolation forest model is applied to score each pending point according to the distance between the pending point and the other points. If the score exceeds the abnormal threshold, the pending point is an abnormal value, otherwise it is a normal value.

3. The distribution network health assessment method according to claim 1, characterized in that: After step S1, the method further includes: The electrical parameters and the environmental parameters are subjected to noise reduction processing to remove noise data in the electrical parameters and the environmental parameters.

4. The distribution network health assessment method according to claim 1, characterized in that: The health assessment model is established using a neural network algorithm, including: A neural network algorithm for establishing the health assessment model is determined according to the parameter type of the data in the distribution network.

5. The distribution network health assessment method according to claim 4, characterized in that: Determining the neural network algorithm for establishing the health assessment model according to the parameter type of the data in the distribution network includes: When the parameter type is nonlinear, a feedforward neural network is used; When the parameter type is image data, a convolutional neural network is used; When the parameter type is time series data, a recursive neural network is used.

6. The distribution network health assessment method according to claim 1, characterized in that: The health status of the distribution network is evaluated according to the health index output by the health assessment model, including: The health index is compared with a preset health status interval to obtain the health status interval into which the health index falls, and the real-time power distribution network health status is marked with a label of the health status interval.

7. A distribution network health assessment system, used to perform the distribution network health assessment according to claims 1-4, characterized in that: include: The acquisition module is used to obtain the physical address of the FTU device to be updated, and number each FTU device according to the physical address; An analysis module is used to determine the starting FTU device for software update according to the number, and to establish an update path tree diagram and an update local area network for each of the FTU devices. A processing module is used for the cloud server to send a software update data packet and an update device list to the starting FTU device according to the update path tree diagram. After the starting FTU device performs software update, the software update data packet and the update device list are sent to the next level child node in the update path tree diagram. After the next level child node performs software update, the above steps are repeated until all nodes in the update path tree diagram complete the software update.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the distribution network health assessment method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the distribution network health assessment method according to any one of claims 1 to 6 is implemented.

Images