A method and system for evaluating the carrying capacity of a distribution network

Abstract

The present invention relates to the technical field of power grid carrying capacity assessment, and in particular to a method and system for distribution network carrying capacity assessment. The method comprises the following steps: obtaining a noise reduction distribution network operation data set; calculating a three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; performing wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; correlating the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model; performing electro-thermal coupling modeling on the distribution transformer winding based on the photovoltaic-imbalance mapping relationship model to obtain a distribution transformer thermodynamic response model; performing finite element thermal field simulation on the distribution transformer thermodynamic response model to obtain winding temperature distribution data. The present invention compensates for the defect that traditional methods are difficult to accurately assess by accurately quantifying the impact of the intermittent output of the photovoltaic inverter on the three-phase imbalance of the distribution transformer.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid carrying capacity assessment, and in particular to a method and system for evaluating the carrying capacity of a distribution network. Background Art

[0002] Early distribution network capacity assessments mainly focused on evaluating the load capacity of traditional power grids under a single power supply mode. The assessment method was relatively simple, usually based on an equal distribution calculation of the total distribution transformer capacity, which could meet the grid operation needs at the time.

[0003] However, with the rapid development and widespread application of new energy generation technologies such as distributed photovoltaics, a large number of distributed photovoltaics have begun to be connected to low-voltage substations in older urban areas, and the assessment of the distribution network's carrying capacity faces new challenges. In the scenario of large-scale access to distributed photovoltaics, existing carrying capacity assessment methods have gradually exposed many shortcomings, making it difficult to accurately quantify the superimposed effect of the intermittent output of photovoltaic inverters on the three-phase imbalance of the distribution transformer. This is because the output of distributed photovoltaics is intermittent and fluctuating, and is greatly affected by environmental factors such as weather. After being connected to the grid, it interacts with traditional power loads, disrupting the balance of parameters such as the three-phase current and voltage of the distribution transformer, exacerbating the three-phase imbalance problem and causing problems such as local overheating of the distribution transformer winding. In addition, traditional models ignore the hidden overload risks caused by the differences in the access capacity of each phase under the self-generation and self-use mode of photovoltaic users. Summary of the Invention

[0004] Based on this, it is necessary for the present invention to provide a method and system for evaluating the carrying capacity of a distribution network to solve at least one of the above technical problems.

[0005] To achieve the above object, a method for evaluating the carrying capacity of a distribution network includes the following steps:

[0006] Step S1: Obtain the noise reduction distribution network operation data set;

[0007] Step S2: Calculate the three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; perform wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; and correlate the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model.

[0008] Step S3: Based on the photovoltaic-imbalance mapping relationship model, the distribution transformer winding is subjected to electro-thermal coupling modeling to obtain a thermodynamic response model of the distribution transformer; a finite element thermal field simulation is performed on the thermodynamic response model of the distribution transformer to obtain winding temperature distribution data; and a probability statistical analysis is performed on the winding temperature distribution data to obtain a critical temperature threshold of the distribution transformer;

[0009] Step S4: constructing a digital twin model of the substation topology; grouping users' electricity consumption characteristics based on the digital twin model of the substation topology to obtain user complementary characteristic grouping data; reconstructing phases based on the user complementary characteristic grouping data to obtain a user phase adjustment plan;

[0010] Step S5: Based on the distribution transformer critical temperature threshold and the user phase adjustment plan, a risk fusion assessment is performed on the low-voltage substation to obtain the comprehensive risk grading data of the substation; based on the comprehensive risk grading data of the substation, a risk-carrying capacity visualization mapping is performed on the low-voltage substation to obtain the visual warning result of the substation carrying capacity.

[0011] By obtaining a noise-reduced distribution network operation data set, the present invention can provide a high-quality data foundation for subsequent accurate evaluation, effectively filter out noise interference, and improve data reliability. In the three-phase imbalance analysis, the three-phase imbalance index is calculated based on the noise-reduced data, and the impact of the intermittent output of the photovoltaic inverter on the three-phase imbalance of the distribution transformer is accurately quantified, making up for the defect that the traditional method is difficult to accurately evaluate. The photovoltaic output fluctuation characteristic data is obtained through wavelet transform decomposition, and the correlation modeling with the three-phase imbalance characteristic data is used to help deeply understand the intrinsic relationship between photovoltaic output fluctuations and three-phase imbalance, and provide a theoretical basis for formulating targeted solutions.

[0012] In the electro-thermal coupling analysis of distribution transformer windings, electro-thermal coupling modeling is performed based on the photovoltaic-imbalance mapping relationship model. Combined with finite element thermal field simulation, accurate winding temperature distribution data can be obtained, and then the critical temperature threshold of the distribution transformer can be determined. This effectively solves the problem that traditional models ignore the risk of hidden overload, leading to overheating of the distribution transformer winding and accelerated insulation aging. By constructing a digital twin model of the substation topology and grouping user power characteristics and reconstructing phases, the user phase distribution can be optimized, improving the power quality and operational stability of the distribution network. By conducting a risk fusion assessment based on the distribution transformer critical temperature threshold and the user phase adjustment plan, visual mapping and early warning of risk and carrying capacity are achieved, providing strong support for the upgrade, operation and maintenance, and fault prevention of the distribution network, ensuring the safe and stable operation of the distribution network in distributed photovoltaic access scenarios.

[0013] Preferably, the present invention further provides a system for evaluating the carrying capacity of a distribution network, which is used to execute the method for evaluating the carrying capacity of a distribution network as described above. The system for evaluating the carrying capacity of a distribution network comprises:

[0014] Data acquisition module, used to obtain noise reduction distribution network operation data set;

[0015] A modeling module is used to calculate the three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; perform wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; and associate the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model;

[0016] The electrothermal coupling simulation module is used to perform electrothermal coupling modeling of the distribution transformer winding based on the photovoltaic-imbalance mapping relationship model to obtain the distribution transformer thermodynamic response model; perform finite element thermal field simulation on the distribution transformer thermodynamic response model to obtain winding temperature distribution data; and perform probabilistic statistical analysis on the winding temperature distribution data to obtain the distribution transformer critical temperature threshold;

[0017] The optimization module is used to build a digital twin model of the substation topology; group users' electricity consumption characteristics based on the digital twin model to obtain user complementary characteristic grouping data; and reconstruct phases based on the user complementary characteristic grouping data to obtain user phase adjustment solutions.

[0018] The risk warning module is used to conduct risk fusion assessment of low-voltage substations based on the distribution transformer critical temperature threshold and the user phase adjustment plan to obtain comprehensive risk grading data of the substations; based on the comprehensive risk grading data of the substations, risk-carrying capacity visualization mapping of the low-voltage substations is performed to obtain visual warning results of the substation carrying capacity.

[0019] This invention uses a data acquisition module to acquire a noise-reduced distribution network operation dataset, providing an accurate and reliable data foundation for subsequent analysis and evaluation, effectively filtering out noise interference and ensuring data quality. The modeling module calculates three-phase imbalance indicators and extracts photovoltaic output fluctuation characteristics based on the noise-reduced data. This module constructs a photovoltaic-imbalance mapping model to accurately quantify the impact of photovoltaic output fluctuations on the distribution transformer's three-phase imbalance, addressing the shortcomings of traditional methods. The electrothermal coupling simulation module implements electrothermal coupling modeling of the distribution transformer windings. Finite element thermal field simulation is used to obtain winding temperature distribution data and the distribution transformer's critical temperature threshold. This addresses the problem of traditional models ignoring hidden overload risks and improves the accuracy of distribution transformer thermal status assessment. The optimization module constructs a digital twin model of the substation topology, groups user power consumption characteristics, and reconstructs phases to derive user phase adjustment plans, optimize the distribution network's three-phase balance, and improve power quality and operational efficiency. The risk warning module integrates the distribution transformer's critical temperature threshold and user phase adjustment plans to perform a fusion risk assessment of low-voltage substations, achieving visual risk-carrying capacity mapping and early warning. This provides a scientific basis for distribution network operation, maintenance, and upgrades, effectively mitigating risks and ensuring safe and stable operation. The system adopts a modular design, with each module having a clear division of labor and working together to form a complete distribution network carrying capacity assessment system. It has good scalability and maintainability, and can adapt to the assessment needs of distribution networks of different scales and complexities. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Other features, objects and advantages of the present invention will become more apparent from reading the detailed description made with reference to the following drawings:

[0021] Figure 1 A schematic flow chart of the steps of a method for evaluating the carrying capacity of a distribution network according to an embodiment is shown. DETAILED DESCRIPTION

[0022] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0023] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0024] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0025] To achieve this, please refer to Figure 1 The present invention provides a method for evaluating the carrying capacity of a distribution network, comprising the following steps:

[0026] Step S1: Obtain the noise reduction distribution network operation data set;

[0027] Step S2: Calculate the three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; perform wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; and correlate the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model.

[0028] Step S3: Based on the photovoltaic-imbalance mapping relationship model, the distribution transformer winding is subjected to electro-thermal coupling modeling to obtain a thermodynamic response model of the distribution transformer; a finite element thermal field simulation is performed on the thermodynamic response model of the distribution transformer to obtain winding temperature distribution data; and a probability statistical analysis is performed on the winding temperature distribution data to obtain a critical temperature threshold of the distribution transformer;

[0029] Step S4: constructing a digital twin model of the substation topology; grouping users' electricity consumption characteristics based on the digital twin model of the substation topology to obtain user complementary characteristic grouping data; reconstructing phases based on the user complementary characteristic grouping data to obtain a user phase adjustment plan;

[0030] Step S5: Based on the distribution transformer critical temperature threshold and the user phase adjustment plan, a risk fusion assessment is performed on the low-voltage substation to obtain the comprehensive risk grading data of the substation; based on the comprehensive risk grading data of the substation, a risk-carrying capacity visualization mapping is performed on the low-voltage substation to obtain the visual warning result of the substation carrying capacity.

[0031] Preferably, step S1 includes the following steps:

[0032] Step S11: collecting three-phase current and voltage data of the low-voltage substation through smart meters to obtain the original data set of the distribution transformer operation;

[0033] Step S12: Collecting photovoltaic inverter output data for distributed photovoltaics to obtain a photovoltaic power generation original data set;

[0034] Step S13: Collect environmental factor data for the low-voltage area to obtain an original data set of environmental factors;

[0035] Step S14: performing time stamp calibration on the distribution transformer operation original data set, the photovoltaic power generation original data set, and the environmental factor original data set to obtain a unified time reference data set;

[0036] Step S15: classify and label the unified time reference dataset to obtain a multi-source heterogeneous label dataset;

[0037] Step S16: acquiring distribution network topology data; performing spatial correlation mapping on the multi-source heterogeneous label dataset and the distribution network topology data to obtain a spatiotemporal correlation power grid operation dataset;

[0038] Step S17: filtering outliers on the spatiotemporally correlated power grid operation dataset to obtain a noise-reduced distribution network operation dataset.

[0039] In this embodiment, Honeywell smart meters installed in low-voltage substations collect three-phase current and voltage data from each monitoring point within the substation in real time at preset 15-minute collection intervals. These smart meters utilize their built-in voltage and current transformers to accurately measure the voltage and current values ​​in the line and transmit the collected data to a local data acquisition terminal via the ZigBee wireless communication protocol. After collecting all three-phase current and voltage data for a specified time period, the terminal stores it as a structured distribution transformer operation raw data set in CSV format, containing the substation number, timestamp, and fields for phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage. For distributed photovoltaic power generation systems, the Modbus TCP / IP communication protocol is configured at the local communication interface of the photovoltaic inverter, and the configured data acquisition server runs SolarEdge's photovoltaic monitoring software. The server sends data read commands to the photovoltaic inverter every 10 minutes. Upon receiving the commands, the inverter packages its current operating parameters, such as output voltage, current, and power, in the Modbus protocol data format and sends them to the server. After receiving the data, the server decodes and parses it, extracting the real-time output data of the PV inverters and storing it as a raw PV power generation dataset. The data is saved in JSON format and includes key information such as the PV system number, acquisition time, output power, output current, and output voltage. To collect environmental factor data for the low-voltage substation, a data collection station (capable of simultaneously collecting multiple environmental parameters) is used. Multiple monitoring points are strategically located within the substation, with data collection stations installed in the east, south, west, north, and center of the substation, approximately 2 meters above the ground to avoid direct sunlight and rain. The collection stations are configured to collect environmental factor data such as ambient temperature, humidity, light intensity, and wind speed every 5 minutes. Each collection station transmits the collected data to a LoRa gateway installed near the substation via a built-in LoRa wireless communication module. The gateway collects the data from each collection station and transmits it to the local environmental monitoring server via an Ethernet interface. The server runs customized environmental data collection software to organize and store the received data, generating a raw environmental factor dataset. The data table contains fields such as monitoring point number, acquisition time, temperature, humidity, light intensity, and wind speed. The distribution transformer operation raw data set comes from smart meters installed in the substations. Data is collected every 15 minutes and includes fields such as the substation number, timestamp, and three-phase current and voltage values. The photovoltaic power generation raw data set comes from the photovoltaic inverters installed in the substations. This data is collected every 10 minutes using the Modbus TCP / IP protocol and contains information such as the photovoltaic system number, collection time, and output power. The environmental factor raw data set is collected every 5 minutes by environmental monitoring sensors deployed in the substations and includes environmental parameters such as monitoring point number, timestamp, temperature, and humidity. Timestamp calibration is performed on these three data sets.The timestamps collected by smart meters were used as the reference because they have a moderate collection frequency and high stability. For photovoltaic power generation data, a standard smart meter timestamp corresponding to each PV data point was calculated based on its 10-minute collection interval. For example, if the PV data was collected at 09:05:00 and the smart meter had a collection point at 09:00:00 and 09:15:00, the PV data's timestamp was adjusted to either 09:00:00 or 09:15:00, depending on the actual relevance of the data. For environmental factor data, its timestamp was matched to the nearest smart meter timestamp, using a 5-minute collection frequency. For example, if the environmental data was collected at 09:03:00 and the smart meter had a collection point at 09:00:00, the environmental data's timestamp was adjusted to 09:00:00. A Python script was written using the pandas library's datetime module and the merge_asof function to automatically calibrate the dataset's timestamps. The three types of calibrated data were merged into a unified chronological order to generate a unified time-base dataset containing all data fields. Data classification and labeling were performed using the Python programming language, along with the pandas and numpy libraries, for this timestamp-calibrated unified time-base dataset. Based on the data source and attributes, the data was divided into three categories: three-phase current and voltage data, photovoltaic power generation data, and environmental factor data. For the three-phase current and voltage data, the pandas groupby function was used to group them by substation number, in accordance with power industry standards. A custom labeling function was then used to label the three-phase current and voltage values ​​for each substation, such as "Substation Phase A Current" and "Substation Phase B Voltage," clearly identifying the specific meaning of each data point. For photovoltaic power generation data, the data was categorized by PV system number, and the numpy where function was used to label the PV output power. For example, power greater than 80% of the rated power was labeled "high power output," between 30% and 80% was labeled "medium power output," and below 30% was labeled "low power output." Environmental factor data are classified according to the monitoring point number and environmental parameter type. With the help of pandas's apply function, labels such as "ambient temperature" and "ambient humidity" are added to parameters such as temperature and humidity. The temperature range is divided according to the actual application scenario. For example, below 0℃ is marked as "low temperature", 0℃-20℃ is marked as "suitable temperature", and above 20℃ is marked as "high temperature". After the above processing, a multi-source heterogeneous label data set is finally obtained, and each data point has a clear classification and label. The distribution network topology data containing rich details is obtained. These data are stored in the Geographic Information System (GIS) format, and detailed information such as the line direction, equipment connection relationship, and geographic coordinates within the substation is recorded. The distribution network topology data is loaded using QGIS software.In QGIS, by importing a shapefile of topological data, the geographic distribution of lines, transformers, meters, and other equipment within the substation area is displayed. A multi-source, heterogeneous labeled dataset containing multi-dimensional data such as timestamp, substation area number, three-phase current and voltage values, PV output, and environmental factors is imported into the QGIS attribute table. By setting join fields, such as using the substation area number and device number as the association key, the labeled dataset is spatially associated with the topological data. In the QGIS map view, each specific device location is associated with a corresponding operational data label. For example, at a transformer location, not only is its geographic location and connectivity visible, but also its associated three-phase current and voltage data, output data for connected PV users, and environmental factor data from surrounding environmental monitoring points can be viewed. This associated data is exported as a new GIS data file (e.g., in geojson format) to generate a spatiotemporally correlated power grid operation dataset. This spatiotemporally correlated power grid operation dataset is then filtered for outliers using the Python programming language, coupled with the pandas and scipy libraries, to generate a de-noised distribution network operation dataset. Pandas was used to read the spatiotemporal correlation dataset and load it into a data frame. For the three-phase current and voltage data, reasonable threshold ranges were set based on power industry standards and historical operating experience. For example, the normal range for three-phase current is generally between 0 and the rated current (assuming 500A). If the current value of a data point exceeds 500A or is below 0A, it is initially identified as an outlier. For voltage data, the normal range is generally within ±10% of the rated voltage (assuming the rated voltage is 220V, the normal range is 198V-242V). The mask function in Pandas was used to mark data exceeding the threshold as outliers. Next, the .medfilt function (median filter) in the signal module in the Scipy library was used to process the data, with a window size of 5. This means that the values ​​of the two data points before and after each data point are considered to determine and replace outliers. For photovoltaic power generation data, we used the Pandas between function to filter out data outside the normal range based on the system's maximum output power (for example, a PV system with a rated power of 30 kW generally has an output range of no more than 30 kW). Median filtering was also applied to correct this. For environmental factor data, taking temperature as an example, we set a reasonable temperature range based on local climate conditions (for example, historical extreme temperatures between -15°C and 40°C). We used the Pandas query function to identify data points outside this range and filled them with linear interpolation (using the Pandas interpolate method). This process resulted in a denoised distribution network operation dataset.

[0040] The present invention can comprehensively obtain the operating data of low-voltage substations through the collection of smart meters, photovoltaic inverter output data, and environmental factor data, covering multiple aspects such as power quality, photovoltaic output, and environmental impact, providing a rich and comprehensive information basis for subsequent carrying capacity assessment. By performing timestamp calibration and data classification and labeling, the time base is unified, so that multi-source heterogeneous data can be analyzed in the same time frame, improving the accuracy and reliability of data integration, and avoiding data errors caused by time inconsistency. By combining the distribution network topology data for spatial correlation mapping, the spatiotemporal dimension of the data is further enriched, which helps to more accurately locate and analyze problems in the distribution network. The denoised data set is obtained by filtering outliers, which effectively removes noise and interference in the data and improves data quality.

[0041] Preferably, step S2 includes the following steps:

[0042] Step S21: performing three-phase current separation and extraction on the noise reduction distribution network operation data set to obtain three-phase current characteristic data;

[0043] Step S22: Calculating the three-phase current imbalance degree according to the three-phase current characteristic data to obtain a current imbalance index;

[0044] Step S23: performing three-phase voltage separation and extraction on the noise reduction distribution network operation data set to obtain three-phase voltage characteristic data;

[0045] Step S24: Calculating the three-phase voltage imbalance according to the three-phase voltage characteristic data to obtain a voltage imbalance index;

[0046] Step S25: extracting zero-sequence current data from the noise reduction distribution network operation data set to obtain zero-sequence current characteristic data;

[0047] Step S26: Calculate the zero-sequence current ratio based on the zero-sequence current characteristic data and the three-phase current characteristic data to obtain a neutral line load index;

[0048] Step S27: performing time series aggregation on the current imbalance index, the voltage imbalance index, and the neutral line load index to obtain three-phase imbalance characteristic data;

[0049] Step S28: performing wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; correlating the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model.

[0050] In this embodiment, MATLAB software is used to extract three-phase currents from a de-noised distribution network operating dataset. The de-noised data is imported into the MATLAB workspace. The data contains information such as the timestamp, substation number, and three-phase currents (phase A, B, and C) and three-phase voltages. In MATLAB, the "find" function is used to locate the columns containing the three-phase current data. Matrix operations are then used to extract the A, B, and C current data and store them in three separate arrays. For example, assuming the original data matrix is ​​"data_matrix," where the third column contains the A phase current, the fourth column contains the B phase current, and the fifth column contains the C phase current. By executing the statement "Ia=data_matrix(:,3);Ib=data_matrix(:,4);Ic=data_matrix(:,5);," the three-phase current characteristic data is obtained, namely, the arrays Ia, Ib, and Ic, each containing time series data for the corresponding phase current. Based on the extracted three-phase current characteristic data, the three-phase current imbalance is calculated using the formula recommended by the power industry standard. The specific operation is as follows: In MATLAB software, the three-phase current characteristic data Ia, Ib, and Ic are combined into a matrix Imatrix=[Ia,Ib,Ic]. The current imbalance is calculated using the following formula:

[0051] Current imbalance = ;

[0052] In MATLAB, the calculation of this formula is implemented by writing a custom function. For example, write the function "current_imbalance=calculate_current_imbalance(Ia,Ib,Ic)". Inside the function, use the "max" function to calculate the maximum difference between the three-phase currents, and use the "mean" function to calculate the average value of the three-phase currents. The current imbalance is calculated according to the above formula. The calculation results are stored in the array "current_imbalance" to obtain the current imbalance index at each time point. Use MATLAB software to separate and extract the three-phase voltages of the distribution network operation data set after noise reduction. Locate the column where the three-phase voltage data is located from the data set. Assume that the 6th column of the original data matrix "data_matrix" is the phase A voltage, the 7th column is the phase B voltage, and the 8th column is the phase C voltage. Execute:

[0053] The statement "Va=data_matrix(:,6);Vb=data_matrix(:,7);Vc=data_matrix(:,8);" extracts the three-phase voltage data into the arrays Va, Vb, and Vc, respectively. This process clearly separates the three-phase voltage characteristic data. Using MATLAB software, the three-phase voltage imbalance is calculated based on the extracted three-phase voltage characteristic data. Referring to the power industry standard, the following formula is used:

[0054] Voltage imbalance = ;

[0055] In MATLAB, write a custom function:

[0056] "voltage_imbalance = calculate_voltage_imbalance(Va,Vb,Vc)". Within this function, the "max" function is used to calculate the maximum difference between the three-phase voltages, and the "mean" function is used to calculate the average of the three-phase voltages. Substituting this into the formula, the voltage imbalance is calculated. The results are stored in the array "voltage_imbalance", yielding the voltage imbalance index at each time point. Using MATLAB software, zero-sequence current data is extracted from the noise-reduced distribution network operation dataset. The original dataset contains information such as the three-phase currents (Ia, Ib, and Ic). According to the definition of zero-sequence current, it is equal to the vector sum of the three-phase currents: I0 = 3Ia + Ib + Ic. In MATLAB, the three-phase current characteristic data (Ia, Ib, and Ic) are combined into a matrix. The statement "I0 = (Ia + Ib + Ic) / 3;" is executed to calculate the zero-sequence current characteristic data, I0, and store it in array I0. Through this process, the zero-sequence current characteristic data is accurately extracted. The zero-sequence current proportion is calculated using MATLAB software. Given the zero-sequence current characteristic data I0 and the three-phase current characteristic data Ia, Ib, and Ic, the formula for calculating the zero-sequence current ratio is:

[0057] Zero sequence current ratio = ;

[0058] In MATLAB, write a custom function:

[0059] "neutral_load_ratio=calculate_neutral_load_ratio(I0,Ia,Ib,Ic)". Within the function, use the "max" function to calculate the maximum value of the three-phase currents, divide the zero-sequence current I0 by this maximum value, and obtain the zero-sequence current ratio. The calculation results are stored in the array "neutral_load_ratio" to obtain the neutral line load index. Use Python language with the pandas and numpy libraries to perform time series aggregation on the current imbalance index, voltage imbalance index, and neutral line load index. The above three indicator data are read into pandas Data Frames respectively. Each Data Frame contains a timestamp and the corresponding indicator value. Use the pandas merge function to merge the three Data Frames into a comprehensive Data Frame with the timestamp as the key. Use the rolling function to set the time window (for example, set it to a rolling window of 1 hour) and calculate the mean and variance of each indicator in each window to reflect the time series variation pattern of the three-phase imbalance characteristics. The results are stored as a new Data Frame and exported as a CSV file to obtain the three-phase unbalance characteristic data. For the detailed implementation process of step S28, please refer to the sub-steps of step S28.

[0060] The present invention extracts three-phase current and voltage separately and extracts zero-sequence current data from the noise-reduced distribution network operation data, calculates the current imbalance index, voltage imbalance index, and neutral line load index respectively, and then performs time-series aggregation to form three-phase imbalance characteristic data, thereby achieving a comprehensive, detailed, and dynamic quantitative description of the three-phase imbalance state of the distribution network. This effectively solves the problem that traditional methods are difficult to accurately quantify the superimposed effect of the intermittent output of photovoltaic inverters on the three-phase imbalance of the distribution transformer, and improves the accuracy and reliability of the distribution network imbalance state assessment. At the same time, the noise-reduced data is processed using wavelet transform decomposition, which can accurately extract the photovoltaic output fluctuation characteristic data and associate it with the three-phase imbalance characteristic data to construct a photovoltaic-imbalance mapping relationship model. This not only deeply reveals the intrinsic connection between photovoltaic output fluctuation and distribution transformer three-phase imbalance, but also provides a strong basis for the subsequent targeted formulation of three-phase imbalance control measures and the accurate assessment of the impact of distributed photovoltaic access on the distribution network carrying capacity, which helps to improve the operating stability and carrying capacity of the distribution network in the distributed photovoltaic access scenario.

[0061] Preferably, step S28 includes the following steps:

[0062] Step S281: performing wavelet multi-scale decomposition on the photovoltaic output data in the noise reduction distribution network operation data set to obtain a photovoltaic output multi-scale component set;

[0063] Step S282: Identify the fluctuation characteristics of the photovoltaic output multi-scale component set to obtain photovoltaic output fluctuation characteristic data;

[0064] Step S283: performing time-frequency domain feature conversion on the photovoltaic output fluctuation characteristic data to obtain photovoltaic fluctuation time-frequency characteristic data;

[0065] Step S284: Acquire an original data set of environmental factors; associate the photovoltaic fluctuation time-frequency characteristic data with the original data set of environmental factors to obtain photovoltaic-meteorological correlation characteristic data;

[0066] Step S285: Time-aligning the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain photovoltaic-imbalance time series correlation data;

[0067] Step S286: performing hybrid mutual information entropy and cross spectral density joint feature extraction on the photovoltaic-imbalance time series correlation data to obtain photovoltaic-imbalance feature data; and constructing a photovoltaic-imbalance mapping relationship model based on the photovoltaic-imbalance feature data;

[0068] Step S287: Performing Monte Carlo testing on the photovoltaic-imbalance mapping relationship model to obtain a model reliability assessment result, wherein each set of scenarios is repeated 100 times; performing parameter calibration on the photovoltaic-imbalance mapping relationship model based on the model reliability assessment result to obtain a calibrated photovoltaic-imbalance mapping relationship model.

[0069] In this embodiment, MATLAB software was used to perform wavelet multi-scale decomposition on the PV output data from the noise-reduced distribution network operation dataset. The specific steps are as follows: The PV output data was imported into the MATLAB workspace in a time series format, containing timestamps and corresponding PV output values. The Daubechies wavelet (db4) was selected as the mother wavelet for a four-layer wavelet decomposition. The decomposition was performed using MATLAB's "wavedec" function. For example, the command "[c,l]=wavedecPVOutputData,4,'db4');" was executed, where "c" is the decomposed coefficient and "l" is the length of each layer's coefficient. The "appcoef" and "detcoef" functions were used to extract the approximate and detail coefficients of each layer, respectively, to obtain a multi-scale component set of PV output, containing the PV output variation characteristics at different time scales (e.g., hourly and minute-level). Based on the multi-scale component set of PV output, MATLAB software was used to identify fluctuation characteristics. The specific steps were as follows: For each scale component, the standard deviation (std) and variance (var) were calculated to quantify the degree of fluctuation. For example, for the first layer of detail coefficients, the command:

[0070] "std_value = std(detail_coeff1); var_value = var(detail_coeff1);". If the standard deviation is greater than a set threshold (e.g., 0.5), the fluctuation at that scale is considered significant. Simultaneously, time-domain waveforms of the components at each scale are plotted to visually identify the fluctuation characteristics by observing the fluctuations in the waveforms. For high-frequency components (e.g., layers 1 and 2), if the waveform changes sharply, it is considered a short-term, intense fluctuation. For low-frequency components (e.g., layers 3 and 4), if the waveform changes slowly, it is considered a long-term, trend-based fluctuation. The fluctuation characteristics at each scale are compiled into PV output fluctuation characteristic data. The PV output fluctuation characteristic data is converted to the time-frequency domain using Python and the scipy library. The specific steps are as follows: Import the fluctuation characteristic data into the Python environment as a one-dimensional array. Perform a short-time Fourier transform using the scipy.signal.stft function, with the following parameters: a window length of 1024 samples, an overlap length of 512 samples, and a Hanning window ('hann') as the window type. For example, execute "f,t,Zxx=signal.stft(fluctuation characteristic data,fs=sampling frequency,window='hann',nperseg=1024,noverlap=512)" to obtain the frequency axis f, time axis t, and the corresponding time-frequency domain matrix Z. Take the absolute value of Z to obtain the amplitude spectrum, square it and normalize it to obtain the energy spectrum, and thus obtain the photovoltaic fluctuation time-frequency characteristic data. Based on the photovoltaic fluctuation time-frequency characteristic data, combined with the original environmental factor data set, correlation analysis is performed. The original environmental factor data set contains timestamp, temperature, humidity, and sunshine intensity information. The data comes from the meteorological station in the substation area and is stored in CSV file format. Use Python and the pandas library to read the environmental factor data and execute:

[0071] "env_data=pd.read_csv('env_data.csv')" is used to time-align it with the photovoltaic fluctuation time-frequency characteristic data. Using the timestamp as the keyword, use the pandas.merge function to merge the two parts of data, for example, "merged_data=pd.merge(photovoltaic time-frequency data,env_data,on='timestamp')". Calculate the correlation coefficient between each environmental factor and photovoltaic output fluctuation. Use the pandas.corr function to find meteorological factors that are significantly correlated with photovoltaic fluctuations, such as sunshine intensity and temperature, to finally obtain photovoltaic-meteorological correlation characteristic data. Time-align the photovoltaic output fluctuation characteristic data and the three-phase imbalance characteristic data. Check whether the timestamp format of the two data is consistent to ensure that they have the same time zone and time accuracy. Use Python's pandas library to load the photovoltaic fluctuation characteristic data and the three-phase imbalance characteristic data into two Data Frames respectively. Use the pandas.merge_asof function to merge based on the timestamp, allowing a certain time tolerance (such as 5 minutes) to ensure that the data points are accurately matched in the time dimension. For example, execute:

[0072] The PV-imbalance time series correlation data is obtained by executing "aligned_data = pd.merge_asof(PV fluctuation data.sort_index(), three-phase imbalance data.sort_index(), on = 'timestamp', tolerance = pd.Timedelta(minutes = 5), direction = 'nearest')". Based on this PV-imbalance time series correlation data, Python and the sklearn and scipy libraries are used to perform a hybrid mutual information entropy and cross-spectral density joint feature extraction. The specific operation is as follows: the mutual information entropy between the PV fluctuation features and the three-phase imbalance features is calculated, and the sklearn.feature_selection.mutual_info_regression function is used to quantify the nonlinear correlation between the two. Furthermore, a cross-spectral density analysis is performed, using the scipy.signal.csd function to extract frequency-domain related features. For example, executing "f,Pxy = signal.csd(PV fluctuation data, three-phase imbalance data, fs = sampling frequency, window = 'hann', nperseg = 1024)" yields a cross-spectral density matrix. Mutual information entropy and cross-spectral density features were fused and dimensionality reduced using principal component analysis (PCA) to obtain PV-imbalance feature data. Based on these feature data, a PV-imbalance mapping model was constructed using TensorFlow. A multi-layer perceptron (MLP) network architecture was selected, with 64 neurons in the hidden layer, ReLU as the activation function, and a linear activation function in the output layer. Mean squared error (MSE) was used as the loss function when compiling the model, and Adam was selected as the optimizer. The training dataset consisted of 80% of the training data set and 20% of the validation data set. Training was performed for 100 epochs to obtain a preliminary PV-imbalance mapping model. Monte Carlo testing was performed to validate the constructed PV-imbalance mapping model. The specific steps were as follows: 100 different scenario data sets were prepared, each containing PV output fluctuation characteristics and three-phase imbalance characteristics. The data were derived from historical operating data sets. Random seeds were generated using the Python numpy library to ensure randomness and independence of each sample. The model was run 100 times for each scenario, and the predicted and actual values ​​were recorded. Calculate the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²) for each scenario to assess model stability and accuracy. For example, execute "mae = np.mean(np.abs(predicted result - actual value))" to obtain the mean absolute error. Based on the model reliability assessment results, if the model shows large errors in certain scenarios, adjust the model's hyperparameters (such as the learning rate and the number of hidden layer neurons) for parameter calibration.We used the GridSearchCV method to search within a predefined hyperparameter space, such as setting the learning rate to [0.001, 0.01, 0.1] and the number of hidden layer neurons to [32, 64, 128], to find the optimal parameter combination. We retrained and validated the model to obtain a calibrated PV-imbalance mapping model.

[0073] The present invention can capture the fluctuation characteristics of photovoltaic output from different scales by performing wavelet multi-scale decomposition and fluctuation characteristic identification on photovoltaic output data, and further enhances the ability to characterize the fluctuation law of photovoltaic output by combining time-frequency domain feature conversion. By correlating environmental factor data with photovoltaic fluctuation characteristic data to form photovoltaic-meteorological correlation characteristic data, it helps to understand the impact of environmental conditions on photovoltaic output, thereby more accurately predicting and evaluating photovoltaic output fluctuations. By time-aligning and jointly extracting features of photovoltaic output fluctuation characteristic data and three-phase imbalance characteristic data, a photovoltaic-imbalance mapping relationship model is constructed, which not only reveals the complex intrinsic relationship between photovoltaic output fluctuations and distribution transformer three-phase imbalance, but also provides an effective tool for quantitatively evaluating the impact of photovoltaic output on the three-phase imbalance of the distribution network. The reliability and accuracy of the photovoltaic-imbalance mapping relationship model are ensured through Monte Carlo test verification and parameter calibration.

[0074] Preferably, step S3 includes the following steps:

[0075] Step S31: constructing a distribution transformer technical parameter library; obtaining a distribution transformer structure specification parameter set through the distribution transformer technical parameter library;

[0076] Step S32: extracting material thermal characteristic parameters from the transformer structure specification parameter set to obtain the transformer material thermal characteristic data;

[0077] Step S33: constructing a distribution transformer geometric model based on the distribution transformer structural specification parameter set; meshing the distribution transformer geometric model to obtain a distribution transformer finite element mesh;

[0078] Step S34: performing a three-phase current imbalance distribution simulation based on the photovoltaic-imbalance mapping relationship model to obtain current imbalance load data;

[0079] Step S35: Calculate the winding copper / aluminum loss based on the current unbalanced load data to obtain winding loss distribution data; perform spatial mapping on the winding loss distribution data to obtain distribution transformer heat source spatial distribution data, wherein the spatial mapping has an axial resolution of 50 mm and a radial resolution of 30 mm;

[0080] Step S36: constructing a heat conduction mathematical model of the distribution transformer based on the thermal property data of the distribution transformer material and the spatial distribution data of the distribution transformer heat source;

[0081] Step S37: obtaining an original data set of environmental factors; extracting environmental boundary condition data based on the original data set of environmental factors; and constructing a thermodynamic response model of the distribution transformer based on the heat conduction mathematical model of the distribution transformer and the environmental boundary condition data;

[0082] Step S38: performing finite element thermal field simulation on the thermodynamic response model of the distribution transformer to obtain winding temperature distribution data; performing probability statistical analysis on the winding temperature distribution data to obtain the distribution transformer critical temperature threshold.

[0083] In this embodiment, a database named "Transformer_Parameters" is constructed using Microsoft SQL Server database software. Within this database, a table named "Structural_Specs" is created, with fields such as transformer model, rated capacity (kVA), primary voltage (kV), secondary voltage (kV), winding material type (copper or aluminum), winding dimensions (length, width, height, in millimeters), and cooling method. This data is derived from the transformer manufacturer's technical manual and nameplate information. For example, for a distribution transformer model S11-1000 / 10, its rated capacity is 1000 kVA, primary voltage is 10 kV, secondary voltage is 0.4 kV, winding material is copper, winding dimensions are 500 mm long, 300 mm wide, and 400 mm high, and the cooling method is oil-immersed cooling. The collected transformer parameter data is imported into the database in batches using the Import / Export Wizard in SQL Server Management Studio (SSMS). Using a SQL query such as "SELECT * FROM Structural_Specs WHERE model = 'S11-1000 / 10';," you can retrieve the structural specification parameter set for a specific distribution transformer model. Based on a constructed distribution transformer technical parameter library, MATLAB software is used to extract material thermal properties from the distribution transformer structural specification parameter set. Using the SQL Server Management Studio (SSMS) query tool, information such as the winding material type (such as copper or aluminum) and cooling oil type (such as transformer oil) is extracted from the "Structural_Specs" table in the "Transformer_Parameters" database. A structure array, "material_properties," is created in MATLAB to store the thermal properties of different materials. For example, for copper windings, according to ASTM International standards, the thermal conductivity is set to 385 W / (m·K), the specific heat capacity is 385 J / (kg·K), and the density is 8960 kg / m³. For aluminum windings, the thermal conductivity is set to 205 W / (m·K), the specific heat capacity is 897 J / (kg·K), and the density is 2700 kg / m³. For transformer oil, the thermal conductivity is set to 0.14 W / (m·K), the specific heat capacity is 1670 J / (kg·K), and the density is 880 kg / m³. Based on the extracted winding material type, the corresponding thermal property parameters are selected from the "material_properties" column to obtain the thermal property data for the distribution transformer material and store it in the MATLAB workspace.Based on the obtained distribution transformer structural parameter set, the distribution transformer geometry model was constructed and meshed using ANSYS Workbench software. Based on the winding structural dimensions (e.g., 500 mm length, 300 mm width, and 400 mm height) and the cooling oil tank dimensions from the parameter set, a 3D geometry model was constructed in the ANSYS Design Modeler module using geometric modeling tools such as extrusion and rotation. The model included key components such as the core, windings, tank, and cooling oil flow channels. Meshing parameters were set in the ANSYS Meshing module. Free meshing was performed using tetrahedral elements, with a global mesh size of 10 mm to ensure that the mesh quality met the analysis accuracy requirements. Local mesh refinement was applied to key areas such as the windings and core, with a mesh size of 5 mm. After meshing, the mesh quality was checked to ensure a skewness of less than 0.8 and an aspect ratio of less than 10, resulting in a high-quality distribution transformer finite element mesh. Based on the constructed photovoltaic-imbalance mapping model, a three-phase current imbalance distribution simulation was performed using MATLAB software. Import the PV-imbalance mapping model into MATLAB. This model is a trained neural network whose input is the PV output fluctuation characteristics and whose output is the three-phase current imbalance distribution. Prepare simulation input data, including PV output fluctuation characteristics under different scenarios. This data is derived from a historical operating data set, encompassing various weather conditions and load conditions. Use MATLAB's "sim" function to simulate the neural network model. For example, execute "[y,t,x]=sim(net,input data);", where "net" is the neural network model, "input data" is the PV output fluctuation characteristic matrix, and "y" is the simulated three-phase current imbalance distribution data. The simulated three-phase current imbalance distribution data contains the three-phase current values ​​at each time point. Based on the current imbalance load data, calculate the winding copper / aluminum losses using MATLAB software, and perform spatial mapping using ANSYS Workbench software. In MATLAB, calculate the winding copper / aluminum losses using the formula Ploss=I²×R. Here, I is the unbalanced current data obtained in step S34, and R is the winding resistance, calculated based on the transformer's technical parameters. The calculated loss data was organized into a matrix format, with each row representing the loss distribution at a specific time point. The loss data was then imported into the ANSYS Mechanical module within ANSYS Workbench and associated with the constructed distribution transformer geometry model. Within ANSYS Mechanical, the "Map Data" function was used to map the loss data onto the winding's finite element mesh, setting the spatial resolution to 50 mm axially and 30 mm radially. Using an interpolation algorithm, the loss data was assigned to each mesh element, yielding the spatial distribution data for the distribution transformer's heat source.Based on the thermal property data of the distribution transformer materials and the spatial distribution data of the heat sources, a mathematical model of the distribution transformer heat conduction was constructed using ANSYS Workbench software. In the ANSYS Mechanical module, parameters such as the thermal conductivity, specific heat capacity, and density of the core, windings, oil tank, and cooling oil were set based on the material thermal property data. For example, for the copper winding, the thermal conductivity was set to 385 W / (m·K), the specific heat capacity to 385 J / (kg·K), and the density to 8960 kg / m³. The spatial distribution data of the heat sources was imported and applied as internal heat sources to the mesh elements corresponding to the windings. Boundary conditions were defined, including the convective heat transfer coefficient (e.g., 25 W / (m²·K)) and ambient temperature (e.g., 25°C) on the transformer tank surface, as well as the contact thermal resistance (e.g., 0.05 m²·K / W) between the windings and the core and oil tank. Using the ANSYS thermal conduction analysis module, steady-state and transient heat conduction equations were established. These equations were discretized using the finite element method to form a mathematical model of the distribution transformer heat conduction. Based on the established heat conduction mathematical model of the distribution transformer, ANSYS Workbench software was used to combine it with a raw environmental data set to construct a thermodynamic response model of the distribution transformer. The raw environmental data set, including ambient temperature, wind speed, and humidity, was obtained from the environmental monitoring system and stored in a CSV file. Using ANSYS Workbench's parametric design language (APDL), a script was written to import the ambient temperature data into the simulation model and set it as a time-varying boundary condition. For example, the "ETABLE,TEMP_AMBIENT,TIME,HIST,,ENV_TEMP" command was executed to apply a time-varying curve of the ambient temperature to the outer surface of the fuel tank. Simultaneously, the convective heat transfer coefficient was adjusted based on the wind speed data. Higher wind speeds increase the convective heat transfer coefficient. This coefficient was dynamically updated using the "SF,OUTSURFACE,CONV,COEF,Wind Speed ​​Dependent Heat Transfer Coefficient" command. In ANSYS Mechanical, the heat conduction mathematical model was combined with the updated environmental boundary conditions to resolve the heat conduction equations, resulting in a thermodynamic response model of the distribution transformer under actual environmental conditions, including temperature distribution and heat flux density. For the detailed implementation process of step S38, please refer to the sub-steps of step S38.

[0084] The present invention provides an accurate structural and material property basis for the subsequent construction of a heat conduction mathematical model by constructing a technical parameter library of distribution transformers and extracting material thermal characteristic parameters, combined with the grid division of the distribution transformer geometric model. Based on the photovoltaic-unbalance mapping relationship model, the three-phase current imbalance distribution is simulated, and then the winding loss distribution is calculated and spatial mapping is performed, which can accurately determine the spatial distribution of the distribution transformer heat source. Furthermore, the heat conduction mathematical model is combined with the environmental boundary condition data to construct a distribution transformer thermodynamic response model, and the winding temperature distribution data is obtained through finite element thermal field simulation, which realizes the accurate simulation of the thermal state of the distribution transformer under the actual operating environment. Finally, the distribution transformer critical temperature threshold is obtained through probabilistic statistical analysis of the winding temperature distribution data, providing a key temperature indicator for the safe operation of the distribution transformer, effectively solving the problem of hidden overload risk caused by the traditional method ignoring the difference in the access capacity of each phase in the self-generation and self-use mode of photovoltaic users, improving the accuracy and reliability of the distribution network carrying capacity assessment, and helping to prevent problems such as accelerated insulation aging caused by overheating of the distribution transformer winding, and ensuring the safe and stable operation of the distribution network.

[0085] Preferably, step S34 includes the following steps:

[0086] Step S341: generating a typical photovoltaic fluctuation scenario set according to preset photovoltaic penetration gradient parameters;

[0087] Step S342: normalizing the typical photovoltaic fluctuation scene set to obtain a standard photovoltaic scene feature vector set;

[0088] Step S343: performing forward propagation reasoning on the standard photovoltaic scenario feature vector set based on the photovoltaic-imbalance mapping relationship model to obtain initial three-phase imbalance prediction data;

[0089] Step S344: performing physical constraint correction on the initial three-phase imbalance prediction data to obtain physically compliant imbalance data;

[0090] Step S345: Performing grid-based environmental monitoring of the low-voltage substation at the meteorological station to obtain substation ambient temperature data; performing line load characteristic monitoring on the low-voltage substation to obtain a substation load characteristic data set; performing load cycle characteristic analysis on the low-voltage substation based on the substation load characteristic data set to obtain substation load cycle characteristic data; dynamically correcting the physical compliance imbalance data based on the substation ambient temperature data and the substation load cycle characteristic data to obtain current compensation imbalance data;

[0091] Step S346: performing three-phase load distribution mapping on the current compensation imbalance data to obtain phase current spatial distribution data;

[0092] Step S347: Obtain historical measured data of the electric meter; correct the phase current spatial distribution data according to the historical measured data of the electric meter to obtain current unbalanced load data.

[0093] In this embodiment, Python and its numpy and pandas libraries are used to generate a set of typical photovoltaic fluctuation scenarios based on the preset photovoltaic penetration gradient parameters. The photovoltaic penetration gradient parameter range is defined as 0.1 to 0.9, with a step size of 0.1, which represents the ratio of photovoltaic installed capacity to the total load capacity of the substation. The penetration rate sequence is generated using numpy's linspace function: penetration_rates=np.linspace(0.1,0.9,9). Combined with historical photovoltaic output data and load data, photovoltaic output fluctuation scenarios under different penetration rates are simulated. Historical data comes from photovoltaic inverters and smart meters in the substation. Use pandas to read data: historical_data=pd.read_csv('historical_pv_load_data.csv'), and extract photovoltaic output and load data columns. By adjusting the proportion of photovoltaic output data, photovoltaic fluctuation scenarios under different penetration rates are generated. For example, for a penetration rate of 0.5, execute pv_output = historical_data['pv_output'] * 0.5 to obtain the corresponding PV fluctuation scenario. This ultimately generates a set of typical PV fluctuation scenarios for various penetration rates, each including a timestamp and PV output value. Based on this set of typical PV fluctuation scenarios, Python and the sklearn library perform normalization to obtain a set of standard PV scenario feature vectors. The data for each PV fluctuation scenario is organized into a feature matrix, with each row representing a time point, containing the PV output value and the corresponding environmental factor (such as temperature and light intensity). The sklearn.preprocessing.StandardScaler class is used to standardize the data. For example, create a standardization object: scaler = StandardScaler() , and fit and transform the data for each scenario: standardized_data = scaler.fit_transform(scene_data) The standardized data has a mean of 0 and a standard deviation of 1. The processed data is stored as a set of standard PV scenario feature vectors. Each vector contains normalized PV output and environmental factor characteristics, providing uniformly scaled input data for subsequent PV-imbalance mapping model inference. Based on the constructed PV-imbalance mapping model, forward propagation inference is performed on the standard PV scenario feature vector set using TensorFlow. The standard PV scenario feature vector set is loaded into TensorFlow as a two-dimensional tensor with a shape of (number of samples, number of features).Load the trained PV-imbalance mapping model. Assume the model is saved as the pv_imbalance_model.h5 file and use tf.keras.models.load_model('pv_imbalance_model.h5') to load the model. Perform forward propagation on the feature vector set and execute initial_predictions = model.predict(feature_vectors) to obtain the initial three-phase imbalance prediction data. The prediction results contain the three-phase current imbalance values ​​at each time point. For example, the output shape is (number of samples, 3), corresponding to the imbalance current prediction values ​​for phases A, B, and C. Based on the initial three-phase imbalance prediction data, use MATLAB software and Kirchhoff's current law to perform physical constraint correction. Import the initial prediction data into MATLAB. The data contains the three-phase current imbalance values ​​at each time point. According to Kirchhoff's current law, the vector sum of the three-phase currents should be zero, that is, Ia + Ib + Ic = 0. Use MATLAB's matrix operations to correct the prediction data. For example, calculate the error term: error = Ia + Ib + Ic, and adjust the three-phase current values ​​to satisfy the current conservation law. A proportional allocation method can be used to proportionally distribute the error to the three phases: Ia = Ia - error / 3; Ib = Ib - error / 3; Ic = Ic - error / 3. The corrected data is stored as physically compliant imbalance data. A LoRa gateway combined with a data collection station performs grid-based environmental monitoring of the low-voltage substation area, obtaining area ambient temperature data. Five monitoring points are evenly distributed within the substation area, located at the four corners and the center, approximately 2 meters above the ground. A HOBO UX120 data collection station is installed at each monitoring point, configured to collect ambient temperature data every 10 minutes. The collection station transmits the data to the LoRa gateway via the LoRa wireless communication protocol, which connects to a local server. The server runs customized environmental data collection software and stores the data as a raw environmental factor dataset, including timestamps and temperature values. Simultaneously, smart meters installed in the substation area monitor line load characteristics, collecting three-phase current and voltage data every 15 minutes. Hooyman software is used to analyze the periodic characteristics of the collected load data and extract the peak, valley, average and daily load curves of the load. The physical compliance imbalance data is dynamically corrected by combining the ambient temperature data and the load cycle characteristic data. For example, during high temperature periods, the load increases, and the amplitude of the imbalance data is adjusted accordingly. The corrected data is stored as current compensation imbalance data, which reflects the three-phase imbalance under actual operating conditions. Based on the current compensation imbalance data, QGIS software is used to map the three-phase load distribution. The current compensation imbalance data is imported into QGIS, and the data contains the timestamp, station number and three-phase current value.Geographic Information System (GIS) data for the distribution network was loaded. This data, stored in shapefile format, contains the geographic coordinates of the lines, transformers, and meters within the substation. In QGIS, the "Join Attributes by Location" tool was used to spatially associate the current data with the geographic data. For example, the three-phase current values ​​were matched to the corresponding meter locations, generating a new layer containing the meter's geographic coordinates and the three-phase current values. The "Interpolation" plugin was used to spatially interpolate the three-phase current values ​​to generate the spatial distribution data of the phase currents. The inverse distance weighted (IDW) interpolation method was selected, and a search radius of 50 meters was set to ensure smoothness and accuracy of the interpolation results. Finally, the spatial distribution of the phase currents was displayed in the QGIS map view, visually reflecting the three-phase load distribution in each area within the substation. Based on the spatial distribution data of the phase currents, corrections were performed using Python and its pandas and numpy libraries combined with historical meter data. Historical meter data, including timestamps, meter numbers, and three-phase current values, was obtained from a database. Use the merge function of pandas to time-align the spatial distribution data of phase current with the actual measured data of the electric meter. For example, execute:

[0094] merged_data=pd.merge(phase_current_data,meter_data,on='timestamp');

[0095] Calculate the error between the two and use the NumPy function np.abs to calculate the absolute error matrix. Based on the error distribution, use the weighted average method to correct the spatial distribution of phase current data. Set the weight of the meter measured data to 0.7 and the weight of the spatially interpolated data to 0.3. For example, execute corrected_current = 0.7 * meter_current + 0.3 * interpolated_current to obtain the corrected current unbalanced load data.

[0096] The present invention generates a set of typical photovoltaic fluctuation scenarios and performs normalization processing to obtain a set of standard photovoltaic scenario feature vectors, providing comprehensive and standardized input data for subsequent three-phase imbalance prediction. By performing forward propagation reasoning based on the photovoltaic-imbalance mapping relationship model and combining Kirchhoff's current law to perform physical constraint correction on the initial prediction data, the physical rationality and accuracy of the prediction results are improved. By performing grid-based environmental monitoring of meteorological stations and line load characteristic monitoring on low-voltage substations, ambient temperature data and load cycle characteristic data are obtained, and the physical compliance imbalance data is further dynamically corrected, so that the current imbalance load data is more consistent with the actual operating conditions. By using the historical measured data of the electric meter to correct the phase current spatial distribution data, the accuracy of the current imbalance load data is further improved.

[0097] Preferably, step S38 includes the following steps:

[0098] Step S381: Acquire historical operating data of the distribution transformer; perform parameter calibration on the distribution transformer thermodynamic response model according to the historical operating data of the distribution transformer to obtain a calibrated distribution transformer thermodynamic response model;

[0099] Step S382: performing a steady-state thermal field simulation on the calibration distribution transformer thermodynamic response model to obtain steady-state temperature distribution data of the distribution transformer; performing a transient thermal field simulation on the calibration distribution transformer thermodynamic response model to obtain transient temperature evolution data of the distribution transformer;

[0100] Step S383: performing thermal field characteristic fusion based on the distribution transformer steady-state temperature distribution data and the distribution transformer transient temperature evolution data to obtain comprehensive thermal characteristic data of the distribution transformer;

[0101] Step S384: Optimizing the accuracy of the calibrated distribution transformer thermodynamic response model based on the comprehensive thermal characteristic data of the distribution transformer to obtain an optimized distribution transformer thermodynamic response model;

[0102] Step S385: performing Monte Carlo simulation based on a typical photovoltaic fluctuation scenario set and an optimized distribution transformer thermodynamic response model to obtain distribution transformer winding temperature distribution data;

[0103] Step S386: performing generalized extreme value distribution fitting on the temperature distribution data of the distribution transformer winding to obtain a winding temperature probability distribution model;

[0104] Step S387: Obtain the insulation material aging characteristic curve by querying the preset material aging database; calculate the critical temperature threshold according to the winding temperature probability distribution model and the insulation material aging characteristic curve to obtain the distribution transformer critical temperature threshold.

[0105] In this embodiment, the historical operating data of the transformer is first obtained. This data is stored in a SQL Server database and includes information such as timestamp, three-phase current, three-phase voltage, and winding temperature. The data is imported into ANSYS Workbench software using the import / export wizard of SQL Server Management Studio (SSMS). In ANSYS Workbench, the constructed transformer thermodynamic response model is loaded, which includes geometric structure, material properties, and boundary conditions. Using the ANSYS Parameter Manager tool, the temperature values ​​in the historical operating data are compared with the temperature values ​​predicted by the model, and the error percentage is calculated. For example, execute error = (predicted_temp - historical_temp) / historical_temp * 100. According to the error distribution, parameters such as the material thermal conductivity and convective heat transfer coefficient in the model are adjusted. For copper windings, the initial thermal conductivity is set to 385W / (m·K) and adjusted to 375W / (m·K) based on the error. After multiple iterative adjustments, the average error between the model predicted temperature and the historical data is less than 5%, and a calibrated transformer thermodynamic response model is obtained. ANSYS Workbench was used to perform steady-state and transient thermal field simulations on the calibrated distribution transformer thermodynamic response model. Steady-state simulation conditions were set, with an ambient temperature of 25°C and a convective heat transfer coefficient of 25 W / (m²·K). The spatial distribution data of the heat source obtained in step S35 was loaded as the internal heat source. A steady-state simulation was run to obtain steady-state temperature distribution data for the distribution transformer, showing the temperature distribution of the windings, core, and oil tank. Transient simulation conditions were set, with a time step of 10 minutes and a total simulation time of 24 hours. In addition to steady-state boundary conditions, the time-dependent rate of temperature change was also considered in the transient simulation. A transient simulation was run to obtain transient temperature evolution data for the distribution transformer, showing the temperature trend over time. For example, during peak load periods, the winding temperature gradually increased from 70°C to 85°C, and then slowly decreased to 65°C during low load periods. The steady-state and transient simulation results were stored in separate ANSYS result files. Based on the steady-state and transient thermal field simulation data, thermal field characteristics were fused using Python and its numpy and pandas libraries. Import the steady-state temperature distribution data and transient temperature evolution data into Python as CSV files. Use the pandas read_csv function to read the data and merge them into a single data frame. Calculate the composite temperature value for each time point and component using a weighted average, with a weight of 0.6 for the steady-state data and 0.4 for the transient data. For example, execute composite_temp = 0.6 * steady_temp + 0.4 * transient_temp.Through the above processing, the fused data reflects both the steady-state temperature distribution and the transient temperature trends. The resulting comprehensive thermal characteristic data for the distribution transformer is stored in a new CSV file. Based on the comprehensive thermal characteristic data, the calibration model is optimized for accuracy using ANSYS Workbench. The comprehensive thermal characteristic data is imported into ANSYS and compared with the temperature data predicted by the model. Using ANSYS's parametric study tool, the variable ranges of the model parameters are defined, such as a ±10% variation range for material thermal conductivity and a ±20% variation range for convective heat transfer coefficient. The optimization objective is to minimize the root mean square error (RMSE) between the predicted and actual temperatures. The optimization algorithm is run, using ANSYS's built-in genetic algorithm to search for parameters. After multiple iterations, the parameter combination that minimizes the RMSE is found. For example, the optimized copper winding thermal conductivity is 380 W / (m·K) and the convective heat transfer coefficient is 28 W / (m²·K). The optimized model is verified to ensure that its prediction accuracy meets engineering requirements, resulting in an optimized distribution transformer thermodynamic response model. Monte Carlo simulations were performed using ANSYS Workbench based on a set of typical PV fluctuation scenarios and an optimized distribution transformer thermodynamic response model. The PV fluctuation scenario set was imported into ANSYS, each containing different PV output fluctuation characteristics. Monte Carlo simulation parameters were set, with a sampling frequency of 100 and a randomly selected PV fluctuation scenario for each sampling. In each simulation, the intensity and distribution of the heat source were updated based on the PV output fluctuation characteristics of the scenario. The simulations were run to obtain 100 sets of distribution transformer winding temperature distribution data, each containing temperature values ​​at different locations along the winding. For example, in one scenario, the mean value of the winding hotspot temperature was 90°C with a standard deviation of 5°C. All simulation results were stored in the ANSYS result database. Based on the distribution transformer winding temperature distribution data, a generalized extreme value distribution was fitted using Python and the Scipy and Matplotlib libraries. The temperature distribution data, containing 100 sets of winding temperature values ​​from different scenarios, was imported into Python. Fitting was performed using the genextreme function in scipy.stats. For example, executing `params=genextreme.fit(temperature_data)` yielded the shape, location, and scale parameters of the distribution. Plot the fitted probability density function (PDF) and cumulative distribution function (CDF), using Matplotlib's plot function to visually display the temperature distribution. For example, the fitting results show that the generalized extreme value distribution parameters for the winding temperature are shape parameter 0.2, location parameter 80°C, and scale parameter 10°C. Statistical analysis verifies that the R² value of the fitting result is greater than 0.95, indicating a good fit. The resulting winding temperature probability distribution model is stored as a Python model object.Based on the winding temperature probability distribution model and a pre-set material aging database, MATLAB is used to calculate the critical temperature threshold. The aging characteristic curve of insulating materials (such as transformer insulation paper) is retrieved from the material aging database, which is stored in a .mat file in MATLAB. The curve describes the relationship between the insulation material's lifespan and temperature. For example, at 100°C, the lifespan is 20 years, while at 120°C, the lifespan is halved. In MATLAB, the temperature probability distribution model and aging characteristic curve data are loaded. Using numerical integration, the material aging rate at different temperatures is calculated and integrated to obtain the total aging. For example, execute aging_rate=integral(@(T)aging_function(T,curve),T_min,T_max), where aging_function is a function defined based on the aging curve. Based on the design lifespan requirement of the insulation material (e.g., 25 years), the critical temperature threshold is derived by inverse calculation. For example, calculations show that when the temperature exceeds 110°C, the material lifespan will be less than 25 years, so 110°C is set as the critical temperature threshold.

[0106] The present invention calibrates the parameters of the model by obtaining the historical operating data of the distribution transformer, combines steady-state and transient thermal field simulations, comprehensively captures the thermal characteristics of the distribution transformer under different operating conditions, and obtains comprehensive thermal characteristic data through thermal field characteristic fusion, providing a solid basis for model accuracy optimization. By using Monte Carlo simulation to simulate the temperature distribution of the distribution transformer winding under typical photovoltaic fluctuation scenarios, combining the generalized extreme value distribution fitting to construct a winding temperature probability distribution model, and combining it with the insulation material aging characteristic curve, the distribution transformer critical temperature threshold is accurately calculated. This not only improves the accuracy and reliability of the distribution transformer thermodynamic response model, but also provides a scientific basis for evaluating the thermal state and insulation life of the distribution transformer under complex operating conditions, effectively preventing equipment failures and accelerated insulation aging caused by temperature exceeding the limit, thereby enhancing the operational safety and stability of the distribution network.

[0107] Preferably, step S4 includes the following steps:

[0108] Step S41: obtaining low-voltage area line topology data;

[0109] Step S42: collecting user profiles for the low-voltage area to obtain basic user information data; correlating and matching the low-voltage area line topology data with the basic user information data to obtain user-topology mapping data;

[0110] Step S43: extracting user power load data from the reliable distribution network operation data set to obtain user load characteristic data; extracting time series features from the user load characteristic data to obtain user load time series feature data;

[0111] Step S44: obtaining a photovoltaic power generation original data set; identifying users with photovoltaic devices based on the photovoltaic power generation original data set to obtain photovoltaic user identification data;

[0112] Step S45: extracting photovoltaic output characteristics of the user in the photovoltaic user identification data to obtain photovoltaic output characteristic data;

[0113] Step S46: Calculate the photovoltaic self-generation and self-consumption rate based on the user load characteristic data and the photovoltaic output characteristic data to obtain the user energy interaction mode data;

[0114] Step S47: constructing a digital twin model framework based on the user-topology mapping data, the user load time series characteristic data, and the user energy interaction mode data;

[0115] Step S48: Perform physical electrical characteristic modeling on the digital twin infrastructure data to obtain a low-voltage area electrical characteristic model; construct an area topology digital twin model based on the user-topology mapping data and the low-voltage area electrical characteristic model;

[0116] Step S49: Grouping users' electricity consumption characteristics based on the digital twin model of the substation topology to obtain user complementary characteristic grouping data; performing phase reconstruction based on the user complementary characteristic grouping data to obtain a user phase adjustment plan.

[0117] It is particularly important that step S49 further includes the following steps:

[0118] Step S491: clustering the user load time series feature data based on the substation topology digital twin model to obtain user load cluster feature data;

[0119] Step S492: clustering the photovoltaic output characteristic data into output patterns based on the substation topology digital twin model to obtain photovoltaic output pattern clustering data;

[0120] Step S493: quantifying the spatiotemporal complementarity based on the user load clustering feature data and the photovoltaic output pattern clustering data to obtain the spatiotemporal complementarity characteristic data of the substation area;

[0121] Step S494: grouping users by complementary characteristics based on the spatial-temporal complementary characteristic data of the substations to obtain user complementary characteristic grouping data;

[0122] Step S495: Reconstruct the phase of the low-voltage substation according to the user complementary characteristic grouping data and the substation topology digital twin model to obtain a user phase adjustment plan.

[0123] In this example, the power grid company's Geographic Information System (GIS) platform is used to obtain low-voltage substation line topology data. This platform stores the geographic coordinates and connectivity of equipment within the substation, such as lines, transformers, and meters. The GIS platform's export function is used to export the line topology data into a shapefile format. The shapefile is loaded using QGIS software to visualize the substation's line layout. Equipment attribute data, such as transformer rated capacity and line conductor type, is exported from the power company's asset management system and stored as an Excel file. The Excel file is read using the Python pandas library and linked to the GIS data to generate complete low-voltage substation line topology data, including the equipment's geographic and attribute information. Customer profile data, including customer ID, name, address, and electricity usage category, is exported from the power company's marketing system. The Python pandas library is used to read the customer profile data and perform preprocessing, such as removing duplicate and missing values. The obtained line topology data, which includes the geographic coordinates of the meters, is loaded. The pandas merge function is used to match the customer profile data with the meter's geographic coordinates, using the customer ID as the key, to generate customer-topology mapping data. Through the above processing, each user's basic information is associated with the corresponding meter location and connection relationship. User load data, including user ID, timestamp, active power, reactive power, and other information, is extracted from the power company's historical database and stored as a CSV file. The Python pandas library is used to read the data and perform data cleaning, such as handling missing values ​​and outliers. The groupby function in pandas is used to group the data by user ID and extract each user's load time series. The numpy library is used to calculate each user's load time series features, such as maximum load, minimum load, average load, load standard deviation, and peak occurrence time. These features are organized into user load time series feature data. The original photovoltaic power generation dataset contains photovoltaic inverter output data, such as timestamp, inverter ID, and output power. The Python pandas library is used to read the data and perform preprocessing, such as removing duplicates and handling missing values. PV user registration information, including user ID, inverter ID, and grid connection time, is exported from the power company's photovoltaic grid-connected system. Using the Pandas merge function, using the inverter number as the key, we matched the PV power generation data with the user registration information to obtain PV user identification data. This process identified users with PV installations and associated them with their corresponding PV output data. We extracted historical output data for each PV user from the original PV power generation dataset, including timestamps and output power. We used the Python pandas library to read the data and perform data cleaning, such as addressing missing values ​​and outliers.The data was aggregated by hour using the resample function in Pandas to calculate the average hourly output power. Key output characteristics for each PV user were calculated, including maximum output power, minimum output power, average output power, output fluctuation rate, and peak occurrence time. These characteristics were organized into PV output characteristic data. User load data and PV output data were aligned to ensure consistent timestamps. Data was merged using the Python pandas library. The self-consumption rate was calculated at each time point using the formula: self-consumption rate = (PV output power - grid-connected power) / PV output power. The apply function in Pandas was used to perform calculations on each row of data. The results were organized into user energy interaction pattern data, including a time series of self-consumption rates for each user. A digital twin model framework was constructed based on the user-topology mapping data, user load time series characteristic data, and user energy interaction pattern data. The three types of data were merged using the Python pandas library, using user ID as the key. The merged data was serialized using JSON format to generate the initial framework for the digital twin model. A 3D model of the substation was created using Blender. Geographic coordinates and equipment layout information were imported. The model was then linked to JSON data, and corresponding attributes and behaviors were assigned to each device and user. Ultimately, a digital twin model framework was constructed, encompassing basic user information, load characteristics, and energy interaction patterns. MATLAB Simulink was used to create an electrical model of the low-voltage substation. Based on the device attribute data in the digital twin model framework, such as the transformer's rated capacity and line impedance parameters, a circuit model was constructed in Simulink. Using Simulink's power system blockset, components such as lines, loads, and photovoltaic inverters were added to construct a complete electrical model of the low-voltage substation. Parameters and connections were set to ensure the model accurately reflected the electrical characteristics of the substation, ultimately resulting in a low-voltage substation electrical characteristic model. The user-topology mapping data was also linked to the electrical model. User load time series data was imported into a pandas Data Frame, containing features such as user number, maximum load, minimum load, average load, load standard deviation, and peak occurrence time. The data was normalized using scikit-learn's StandardScaler to eliminate dimensionality differences. Users were clustered using the KMeans clustering algorithm, with the number of cluster centers set to 5. The optimal number of clusters was determined using the elbow rule. This yielded user load clustering data, with each user assigned a cluster label, such as "industrial high load" or "residential low load." Based on PV output characteristic data, output pattern clustering was performed using Python and the scikit-learn library.PV output characteristic data was imported into a pandas DataFrame. The data contained features such as PV user ID, maximum output power, minimum output power, average output power, output fluctuation rate, and peak occurrence time. StandardScaler was also used to normalize the data. PV users were clustered using the KMeans clustering algorithm, with the number of cluster centers set to 4. The clustering effect was evaluated using the silhouette coefficient. This yielded PV output pattern cluster data, with each PV user assigned a cluster label, such as "peak afternoon type" or "stable output type." Based on the user load clustering feature data and PV output pattern clustering data, spatiotemporal complementarity was quantified using Python and the pandas and numpy libraries. The user load clustering data and PV output pattern clustering data were imported into a pandas DataFrame, containing user ID and cluster label, and PV user ID and cluster label, respectively. The user load data and PV output data were merged into a single time series dataframe by aligning timestamps. The correlation coefficient between load and PV output at each time point was calculated using the numpy function corrcoef to generate the correlation coefficient matrix. Based on the correlation coefficient matrix, spatiotemporal complementarity is quantified. For example, a correlation coefficient less than 0.3 indicates strong complementarity, while a coefficient greater than 0.7 indicates weak complementarity. This ultimately yields data on the spatiotemporal complementarity characteristics of the substation area. Based on this data, users are grouped using Python and the scikit-learn library. This spatiotemporal complementarity data is imported into a pandas Data Frame. The data contains features such as user ID, PV user ID, and the correlation coefficient between load and PV output. Users are grouped using the DBSCAN density clustering algorithm, with a neighborhood radius of eps = 0.5 and a minimum number of samples min_samples = 5, to identify user groups with similar complementary characteristics. This yields user complementarity grouping data, each with similar spatiotemporal complementarity. This user complementarity grouping data and substation area topology data are imported into a NetworkX graph structure, where nodes represent users and PV installations, and edges represent electrical connections. Based on the complementary characteristics of the user groups, phase adjustment objective functions are defined, such as minimizing three-phase imbalance and maximizing PV consumption efficiency. Phase allocation is optimized using the NetworkX minimum spanning tree algorithm. According to the minimum spanning tree result, the user's phase connection is adjusted to obtain the user phase adjustment plan, and the final plan allocates users to different phases.

[0124] The present invention obtains low-voltage substation line topology data and user profile information to construct user-topology mapping data, providing spatial and user basic information support for subsequent analysis of user electricity consumption characteristics. By extracting time series features from user electricity load data, identifying photovoltaic users and their output characteristics in combination with photovoltaic power generation data, and further calculating the photovoltaic self-generation and self-use rate, the user's electricity consumption behavior and energy production characteristics are fully characterized. Based on these data, a digital twin model of the substation topology is constructed to achieve digital mapping and real-time monitoring of the electrical characteristics of the low-voltage substation. By grouping and phase reconstruction of user electricity consumption characteristics, a scientific basis is provided for optimizing the three-phase balance state of the distribution network, which helps to improve the power quality and operating efficiency of the distribution network, reduce equipment loss and energy waste caused by three-phase imbalance, and enhance the stability and reliability of the distribution network.

[0125] Preferably, step S5 includes the following steps:

[0126] Step S51: normalizing the distribution transformer critical temperature threshold to obtain a standard critical temperature index;

[0127] Step S52: performing three-phase power dynamic balancing calculation and evaluation based on the user phase adjustment scheme to obtain user phase optimization evaluation data;

[0128] Step S53: Performing a multi-dimensional risk-carrying capacity visualization mapping of the power grid anti-disturbance impedance based on the substation topology digital twin model to obtain power grid anti-disturbance impedance characteristic data;

[0129] Step S54: performing a thermal-electrical coupling risk assessment based on the photovoltaic-imbalance mapping relationship model and the distribution transformer thermodynamic response model to obtain a thermal-electrical coupling risk index;

[0130] Step S55: Perform fuzzy membership graded evaluation based on the standard critical temperature index to obtain temperature risk level evaluation data;

[0131] Step S56: performing a benefit quantitative evaluation on the user phase optimization evaluation data to obtain a user phase adjustment benefit index;

[0132] Step S57: Obtain a distribution network historical fault record data set; perform grid fault prediction on the low-voltage substation based on the grid anti-disturbance impedance characteristic data and the distribution network historical fault record data set to obtain grid fault probability prediction data;

[0133] Step S58: Perform risk fusion assessment on the low-voltage substation based on the thermal-electrical coupling risk index and the temperature risk level assessment data to obtain comprehensive risk grading data for the substation;

[0134] Step S59: Perform risk-carrying capacity visualization mapping on the low-voltage substation according to the comprehensive risk grading data of the substation to obtain a visual warning result of the substation's carrying capacity.

[0135] It is particularly important that step S59 further includes the following steps:

[0136] Step S591: performing a reliability balance quantitative evaluation on the user phase adjustment benefit index and the power grid fault probability prediction data to obtain reliability balance evaluation data;

[0137] Step S592: constructing a quantitative indicator of the area's carrying capacity based on the area's comprehensive risk classification data and reliability balance assessment data;

[0138] Step S593: performing time series decomposition on the quantitative index of the substation carrying capacity to obtain time series decomposition characteristic data of the carrying capacity;

[0139] Step S594: performing bidirectional gated recurrent network prediction on the bearing capacity time series decomposition feature data to obtain bearing capacity prediction sequence data;

[0140] Step S595: performing kernel density estimation threshold segmentation based on the carrying capacity prediction sequence data to obtain load-bearing grade warning probability data; performing reliability calibration on the load-bearing grade warning probability data to obtain calibrated warning probability data;

[0141] Step S596: Based on the calibrated warning probability data, the low-voltage substation is divided into warning levels to obtain the three-level warning classification data of the substation; based on the three-level warning classification data of the substation, spatiotemporal visualization rendering is performed to obtain the visual warning results of the substation's carrying capacity.

[0142] In this example, the distribution transformer critical temperature threshold data is imported into a pandas Data Frame, which contains the distribution transformer number and the corresponding critical temperature value. The temperature data is normalized using scikit-learn's MinMaxScaler and mapped to the range [0, 1]. Based on the user phase adjustment scheme, MATLAB is used to calculate and evaluate the dynamic balance of three-phase power. The phase adjustment scheme, including the user number and the adjusted phase information, is imported into MATLAB. The three-phase electrical model of the substation is constructed using MATLAB's Power System Toolbox, and simulation parameters are set, such as a 15-minute simulation time step and a 24-hour simulation cycle. The three-phase power imbalance before and after the adjustment is calculated through simulation, and the results show that the imbalance is significantly reduced after the phase adjustment. Based on the digital twin model of the substation topology, Python and its matplotlib and networkx libraries are used to perform a multidimensional risk-carrying capacity visualization mapping. The digital twin model data is imported into a networkx graph structure, where nodes represent devices and users and edges represent electrical connections. Matplotlib is used to draw a network diagram, with node size scaled according to the device impedance value or user load size and color filled according to the risk level. Simultaneously, grid anti-disturbance impedance characteristic data, such as network connectivity, key node identification, and path redundancy, are calculated. Based on the constructed PV-imbalance mapping model and the optimized distribution transformer thermodynamic response model, coupling risk assessment is performed using Python and the pandas and scikit-learn libraries. The output data of the relevant models, including timestamps, three-phase imbalance indicators, and distribution transformer temperature predictions, are imported into a pandas Data Frame. The coupling risk assessment model is constructed using scikit-learn's RandomForestClassifier, with the three-phase imbalance indicator and temperature as input features and the target variable, i.e., the risk level, as the output. By training the model and performing predictions, a thermal-electric coupling risk index is obtained. Based on the standard critical temperature indicator, a fuzzy membership rating assessment is performed using Python and the scikit-fuzzy library. A fuzzy set of temperature risk levels is defined, including "low risk," "medium risk," and "high risk," and corresponding membership functions are set. The standard critical temperature indicator is input into the membership function, and the risk membership corresponding to each temperature value is calculated, ultimately generating temperature risk level assessment data. Based on user phase optimization evaluation data, we used Python and the pandas and matplotlib libraries to quantify the benefits. We imported the three-phase power imbalance data before and after phase optimization into a pandas Data Frame. We calculated optimization benefit metrics, such as the percentage reduction in imbalance and the estimated annual energy loss reduction. We used matplotlib to plot a comparison chart before and after optimization, visually demonstrating the benefits of phase adjustment.Based on the power grid's anti-disturbance impedance characteristic data and a dataset of historical distribution network fault records, Python and the scikit-learn library were used to predict fault probabilities. The historical fault record data was imported into a pandas Data Frame. The data was divided into training and test sets using the scikit-learn function train_test_split. A fault probability prediction model was constructed using LogisticRegression, with the anti-disturbance impedance characteristic data and time features as input features and the target variable, fault probability, as output. By training the model and performing predictions, grid fault probability prediction data was obtained. Risk fusion assessment was performed using Python, the pandas library, and the scikit-learn library based on the thermal-electric coupling risk index and temperature risk level assessment data. The thermal-electric coupling risk index and temperature risk level data were imported into a pandas Data Frame. Risk fusion rules, such as the weighted summation method, were defined. Risk values ​​were classified based on the fused risk values, with risk thresholds set, such as low risk (0-0.3), medium risk (0.3-0.6), and high risk (0.6-1.0). The fusion model's accuracy was evaluated using the scikit-learn confusion_matrix function, ultimately generating comprehensive risk grading data for the substation area. Data integration and calculations were performed using Python, along with the pandas and numpy libraries. User phase adjustment benefit indicators (e.g., the percentage reduction in three-phase power imbalance) and grid fault probability prediction data (e.g., the probability of a fault occurring within the next 24 hours) were imported into a pandas Data Frame. A reliability balance assessment formula was defined, for example, reliability balance assessment value = 0.7 * benefit indicator + 0.3 * (1 - fault probability), with weights set based on historical data. Through data processing and calculations, a column of reliability balance assessment data was generated for each time point. Based on the comprehensive risk grading data and reliability balance assessment data, a quantitative indicator of the substation's carrying capacity was constructed. Data fusion and indicator construction were performed using Python, along with the pandas and scikit-learn libraries. The comprehensive risk grading data (high, medium, and low risk values ​​are 1, 2, and 3, respectively) and the reliability balance assessment data were imported into a pandas Data Frame. Define a formula for quantitative indicators of carrying capacity. For example, the carrying capacity indicator = 0.6 * (1 / comprehensive risk value) + 0.4 * reliability balance assessment value. Weights are set based on the impact of risk and reliability on carrying capacity. Through data processing and calculation, a column of quantitative indicators of substation carrying capacity is obtained. Time series decomposition of the quantitative indicators of substation carrying capacity is performed. Time series analysis is performed using Python and the statsmodels library. Import the quantitative indicator data into a pandasSeries, setting the timestamp as the index.Using the StatsModels STL (Seasonal-Trend decomposition using LOESS) decomposition method, the trend, seasonal, and residual components of the data were separated. This method yielded time-series decomposition feature data for carrying capacity, including trend, seasonal, and residual components. Based on this time-series decomposition feature data, a Bidirectional Gated Recurrent Network (BiLSTM) prediction model was implemented using Python and the TensorFlow library. The time-series decomposition feature data (trend, seasonal, and residual) were organized as the input and output of a supervised learning problem. The BiLSTM model was constructed using TensorFlow's Keras API. The model architecture consists of two BiLSTM layers, each containing 50 units, followed by a Dense layer to output the predicted values. Mean Squared Error was selected as the loss function and Adam as the optimizer when compiling the model. Through model training and prediction, carrying capacity prediction series data was obtained and stored as a NumPy array. Based on this carrying capacity prediction series data, kernel density estimation threshold segmentation and reliability calibration were performed using Python and the Scipy and Sklearn libraries. The carrying capacity prediction series data was imported into a NumPy array. Kernel density estimation was performed using the scipy function gaussian_kde to obtain load-bearing capacity classification warning probability data. Thresholds were set based on the warning level (e.g., low, medium, and high), with a low threshold of 0.3 and a medium threshold of 0.6. The probability data was calibrated for reliability using the calibration.CalibratedClassifierCV function in sklearn to obtain calibrated warning probability data. Based on the calibrated warning probability data, warning level classification and spatiotemporal visualization were performed using Python's matplotlib and folium libraries. The calibrated warning probability data was imported into a pandas Data Frame and classified according to the set thresholds (low load < 0.3, 0.3 ≤ medium load < 0.6, high load ≥ 0.6). An interactive map was created using the folium library to visualize the substation's geographic information and warning level data, generating a visual warning result for the substation's load capacity. The map then visually displays the warning level for each substation.

[0143] The present invention can quantify the benefits of phase adjustment and provide data support for the optimized operation of the distribution network by normalizing the critical temperature threshold of the distribution transformer and combining it with the optimization evaluation of the user's phase adjustment scheme. By using the digital twin model of the substation topology to visualize the anti-disturbance impedance of the power grid, the anti-disturbance capability and risk-carrying capacity relationship of the power grid are intuitively presented, which helps to identify the weak links of the power grid in advance. Through thermal-electric coupling risk assessment and fuzzy membership graded assessment, the impact of various factors such as temperature and power load on power grid risks is comprehensively considered, which improves the comprehensiveness and accuracy of risk assessment. Through the prediction of power grid faults based on historical fault data, the ability to predict potential faults is further enhanced. By performing risk fusion assessment and risk-carrying capacity visualization mapping on low-voltage substations, intuitive visual warning results of carrying capacity are generated.

[0144] Preferably, the present invention further provides a system for evaluating the carrying capacity of a distribution network, which is used to execute the method for evaluating the carrying capacity of a distribution network as described above. The system for evaluating the carrying capacity of a distribution network comprises:

[0145] Data acquisition module, used to obtain noise reduction distribution network operation data set;

[0146] A modeling module is used to calculate the three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; perform wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; and associate the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model;

[0147] The electrothermal coupling simulation module is used to perform electrothermal coupling modeling of the distribution transformer winding based on the photovoltaic-imbalance mapping relationship model to obtain the distribution transformer thermodynamic response model; perform finite element thermal field simulation on the distribution transformer thermodynamic response model to obtain winding temperature distribution data; and perform probabilistic statistical analysis on the winding temperature distribution data to obtain the distribution transformer critical temperature threshold;

[0148] The optimization module is used to build a digital twin model of the substation topology; group users' electricity consumption characteristics based on the digital twin model to obtain user complementary characteristic grouping data; and reconstruct phases based on the user complementary characteristic grouping data to obtain user phase adjustment solutions.

[0149] The risk warning module is used to conduct risk fusion assessment of low-voltage substations based on the distribution transformer critical temperature threshold and the user phase adjustment plan to obtain comprehensive risk grading data of the substations; based on the comprehensive risk grading data of the substations, risk-carrying capacity visualization mapping of the low-voltage substations is performed to obtain visual warning results of the substation carrying capacity.

[0150] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0151] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating the carrying capacity of a distribution network, characterized in that: The following steps are involved: Step S1: Obtain the noise reduction distribution network operation data set; Step S2: Calculate the three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; perform wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; associate the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model; wherein step S2 includes the following steps: Step S21: performing three-phase current separation and extraction on the noise reduction distribution network operation data set to obtain three-phase current characteristic data; Step S22: Calculating the three-phase current imbalance degree according to the three-phase current characteristic data to obtain a current imbalance index; Step S23: performing three-phase voltage separation and extraction on the noise reduction distribution network operation data set to obtain three-phase voltage characteristic data; Step S24: Calculating the three-phase voltage imbalance according to the three-phase voltage characteristic data to obtain a voltage imbalance index; Step S25: extracting zero-sequence current data from the noise reduction distribution network operation data set to obtain zero-sequence current characteristic data; Step S26: Calculate the zero-sequence current ratio based on the zero-sequence current characteristic data and the three-phase current characteristic data to obtain a neutral line load index; Step S27: performing time series aggregation on the current imbalance index, the voltage imbalance index, and the neutral line load index to obtain three-phase imbalance characteristic data; Step S28: performing wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; correlating the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model; wherein step S28 includes the following steps: Step S281: performing wavelet multi-scale decomposition on the photovoltaic output data in the noise reduction distribution network operation data set to obtain a photovoltaic output multi-scale component set; Step S282: Identify the fluctuation characteristics of the photovoltaic output multi-scale component set to obtain photovoltaic output fluctuation characteristic data; Step S283: performing time-frequency domain feature conversion on the photovoltaic output fluctuation characteristic data to obtain photovoltaic fluctuation time-frequency characteristic data; Step S284: Acquire an original data set of environmental factors; associate the photovoltaic fluctuation time-frequency characteristic data with the original data set of environmental factors to obtain photovoltaic-meteorological correlation characteristic data; Step S285: Time-aligning the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain photovoltaic-imbalance time series correlation data; Step S286: performing hybrid mutual information entropy and cross spectral density joint feature extraction on the photovoltaic-imbalance time series correlation data to obtain photovoltaic-imbalance feature data; and constructing a photovoltaic-imbalance mapping relationship model based on the photovoltaic-imbalance feature data; Step S287: performing a Monte Carlo test on the photovoltaic-imbalance mapping relationship model to obtain a model reliability assessment result, wherein each set of scenarios is repeated 100 times; performing parameter calibration on the photovoltaic-imbalance mapping relationship model based on the model reliability assessment result to obtain a calibrated photovoltaic-imbalance mapping relationship model; Step S3: Based on the photovoltaic-unbalance mapping relationship model, the distribution transformer winding is subjected to electro-thermal coupling modeling to obtain the distribution transformer thermodynamic response model; the distribution transformer thermodynamic response model is subjected to finite element thermal field simulation to obtain the winding temperature distribution data; the winding temperature distribution data is subjected to probability statistical analysis to obtain the distribution transformer critical temperature threshold; wherein, Step S4: constructing a digital twin model of the substation topology; grouping users' electricity consumption characteristics based on the digital twin model of the substation topology to obtain user complementary characteristic grouping data; reconstructing phases based on the user complementary characteristic grouping data to obtain a user phase adjustment plan; Step S5: Based on the distribution transformer critical temperature threshold and the user phase adjustment plan, a risk fusion assessment is performed on the low-voltage substation to obtain the comprehensive risk grading data of the substation; based on the comprehensive risk grading data of the substation, a risk-carrying capacity visualization mapping is performed on the low-voltage substation to obtain the visual warning result of the substation carrying capacity.

2. The method for evaluating the carrying capacity of a distribution network according to claim 1, wherein: Step S1 includes the following steps: Step S11: collecting three-phase current and voltage data of the low-voltage substation through smart meters to obtain the original data set of the distribution transformer operation; Step S12: Collecting photovoltaic inverter output data for distributed photovoltaics to obtain a photovoltaic power generation original data set; Step S13: Collect environmental factor data for the low-voltage area to obtain an original data set of environmental factors; Step S14: performing time stamp calibration on the distribution transformer operation original data set, the photovoltaic power generation original data set, and the environmental factor original data set to obtain a unified time reference data set; Step S15: classify and label the unified time reference dataset to obtain a multi-source heterogeneous label dataset; Step S16: acquiring distribution network topology data; performing spatial correlation mapping on the multi-source heterogeneous label dataset and the distribution network topology data to obtain a spatiotemporal correlation power grid operation dataset; Step S17: filtering outliers on the spatiotemporally correlated power grid operation dataset to obtain a noise-reduced distribution network operation dataset.

3. The method for evaluating the carrying capacity of a distribution network according to claim 1, wherein: Step S3 includes the following steps: Step S31: constructing a distribution transformer technical parameter library; obtaining a distribution transformer structure specification parameter set through the distribution transformer technical parameter library; Step S32: extracting material thermal characteristic parameters from the transformer structure specification parameter set to obtain the transformer material thermal characteristic data; Step S33: constructing a distribution transformer geometric model based on the distribution transformer structural specification parameter set; meshing the distribution transformer geometric model to obtain a distribution transformer finite element mesh; Step S34: performing a three-phase current imbalance distribution simulation based on the photovoltaic-imbalance mapping relationship model to obtain current imbalance load data; Step S35: Calculate the winding copper / aluminum loss based on the current unbalanced load data to obtain winding loss distribution data; perform spatial mapping on the winding loss distribution data to obtain distribution transformer heat source spatial distribution data, wherein the spatial mapping has an axial resolution of 50 mm and a radial resolution of 30 mm; Step S36: constructing a heat conduction mathematical model of the distribution transformer based on the thermal property data of the distribution transformer material and the spatial distribution data of the distribution transformer heat source; Step S37: obtaining an original data set of environmental factors; extracting environmental boundary condition data based on the original data set of environmental factors; and constructing a thermodynamic response model of the distribution transformer based on the heat conduction mathematical model of the distribution transformer and the environmental boundary condition data; Step S38: performing finite element thermal field simulation on the thermodynamic response model of the distribution transformer to obtain winding temperature distribution data; performing probability statistical analysis on the winding temperature distribution data to obtain the distribution transformer critical temperature threshold.

4. The method for evaluating the carrying capacity of a distribution network according to claim 3, wherein: Step S34 includes the following steps: Step S341: generating a typical photovoltaic fluctuation scenario set according to preset photovoltaic penetration gradient parameters; Step S342: normalizing the typical photovoltaic fluctuation scene set to obtain a standard photovoltaic scene feature vector set; Step S343: performing forward propagation reasoning on the standard photovoltaic scenario feature vector set based on the photovoltaic-imbalance mapping relationship model to obtain initial three-phase imbalance prediction data; Step S344: performing physical constraint correction on the initial three-phase imbalance prediction data to obtain physically compliant imbalance data; Step S345: Performing grid-based environmental monitoring of the low-voltage substation at the meteorological station to obtain substation ambient temperature data; performing line load characteristic monitoring on the low-voltage substation to obtain a substation load characteristic data set; performing load cycle characteristic analysis on the low-voltage substation based on the substation load characteristic data set to obtain substation load cycle characteristic data; dynamically correcting the physical compliance imbalance data based on the substation ambient temperature data and the substation load cycle characteristic data to obtain current compensation imbalance data; Step S346: performing three-phase load distribution mapping on the current compensation imbalance data to obtain phase current spatial distribution data; Step S347: Obtain historical measured data of the electric meter; correct the phase current spatial distribution data according to the historical measured data of the electric meter to obtain current unbalanced load data.

5. The method for evaluating the carrying capacity of a distribution network according to claim 4, characterized in that: Step S38 includes the following steps: Step S381: Acquire historical operating data of the distribution transformer; perform parameter calibration on the distribution transformer thermodynamic response model according to the historical operating data of the distribution transformer to obtain a calibrated distribution transformer thermodynamic response model; Step S382: performing a steady-state thermal field simulation on the calibration distribution transformer thermodynamic response model to obtain steady-state temperature distribution data of the distribution transformer; performing a transient thermal field simulation on the calibration distribution transformer thermodynamic response model to obtain transient temperature evolution data of the distribution transformer; Step S383: performing thermal field characteristic fusion based on the distribution transformer steady-state temperature distribution data and the distribution transformer transient temperature evolution data to obtain comprehensive thermal characteristic data of the distribution transformer; Step S384: Optimizing the accuracy of the calibrated distribution transformer thermodynamic response model based on the comprehensive thermal characteristic data of the distribution transformer to obtain an optimized distribution transformer thermodynamic response model; Step S385: performing Monte Carlo simulation based on a typical photovoltaic fluctuation scenario set and an optimized distribution transformer thermodynamic response model to obtain distribution transformer winding temperature distribution data; Step S386: performing generalized extreme value distribution fitting on the temperature distribution data of the distribution transformer winding to obtain a winding temperature probability distribution model; Step S387: Obtain the insulation material aging characteristic curve by querying the preset material aging database; calculate the critical temperature threshold according to the winding temperature probability distribution model and the insulation material aging characteristic curve to obtain the distribution transformer critical temperature threshold.

6. The method for evaluating the carrying capacity of a distribution network according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: obtaining low-voltage area line topology data; Step S42: collecting user profiles for the low-voltage area to obtain basic user information data; correlating and matching the low-voltage area line topology data with the basic user information data to obtain user-topology mapping data; Step S43: extracting user power load data from the reliable distribution network operation data set to obtain user load characteristic data; extracting time series features from the user load characteristic data to obtain user load time series feature data; Step S44: obtaining a photovoltaic power generation original data set; identifying users with photovoltaic devices based on the photovoltaic power generation original data set to obtain photovoltaic user identification data; Step S45: extracting photovoltaic output characteristics of the user in the photovoltaic user identification data to obtain photovoltaic output characteristic data; Step S46: Calculate the photovoltaic self-generation and self-consumption rate based on the user load characteristic data and the photovoltaic output characteristic data to obtain the user energy interaction mode data; Step S47: constructing a digital twin model framework based on the user-topology mapping data, the user load time series characteristic data, and the user energy interaction mode data; Step S48: Perform physical electrical characteristic modeling on the digital twin infrastructure data to obtain a low-voltage area electrical characteristic model; construct an area topology digital twin model based on the user-topology mapping data and the low-voltage area electrical characteristic model; Step S49: Grouping users' electricity consumption characteristics based on the digital twin model of the substation topology to obtain user complementary characteristic grouping data; performing phase reconstruction based on the user complementary characteristic grouping data to obtain a user phase adjustment plan.

7. The method for evaluating the carrying capacity of a distribution network according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: normalizing the distribution transformer critical temperature threshold to obtain a standard critical temperature index; Step S52: performing three-phase power dynamic balancing calculation and evaluation based on the user phase adjustment scheme to obtain user phase optimization evaluation data; Step S53: Performing a multi-dimensional risk-carrying capacity visualization mapping of the power grid anti-disturbance impedance based on the substation topology digital twin model to obtain power grid anti-disturbance impedance characteristic data; Step S54: performing a thermal-electrical coupling risk assessment based on the photovoltaic-imbalance mapping relationship model and the distribution transformer thermodynamic response model to obtain a thermal-electrical coupling risk index; Step S55: Perform fuzzy membership graded evaluation based on the standard critical temperature index to obtain temperature risk level evaluation data; Step S56: performing a benefit quantitative evaluation on the user phase optimization evaluation data to obtain a user phase adjustment benefit index; Step S57: Obtain a distribution network historical fault record data set; perform grid fault prediction on the low-voltage substation based on the grid anti-disturbance impedance characteristic data and the distribution network historical fault record data set to obtain grid fault probability prediction data; Step S58: Perform risk fusion assessment on the low-voltage substation based on the thermal-electrical coupling risk index and the temperature risk level assessment data to obtain comprehensive risk grading data for the substation; Step S59: Perform risk-carrying capacity visualization mapping on the low-voltage substation according to the comprehensive risk grading data of the substation to obtain a visual warning result of the substation's carrying capacity.

8. A system for evaluating the carrying capacity of a distribution network, characterized in that: For executing the method for evaluating the carrying capacity of a distribution network according to claim 1, the system for evaluating the carrying capacity of a distribution network comprises: Data acquisition module, used to obtain noise reduction distribution network operation data set; A modeling module is used to calculate the three-phase imbalance index based on the noise reduction distribution network operation data set to obtain three-phase imbalance characteristic data; perform wavelet transform decomposition on the noise reduction distribution network operation data set to obtain photovoltaic output fluctuation characteristic data; and associate the photovoltaic output fluctuation characteristic data with the three-phase imbalance characteristic data to obtain a photovoltaic-imbalance mapping relationship model; The electrothermal coupling simulation module is used to perform electrothermal coupling modeling of the distribution transformer winding based on the photovoltaic-imbalance mapping relationship model to obtain the distribution transformer thermodynamic response model; perform finite element thermal field simulation on the distribution transformer thermodynamic response model to obtain winding temperature distribution data; and perform probabilistic statistical analysis on the winding temperature distribution data to obtain the distribution transformer critical temperature threshold; The optimization module is used to build a digital twin model of the substation topology; group users' electricity consumption characteristics based on the digital twin model to obtain user complementary characteristic grouping data; and reconstruct phases based on the user complementary characteristic grouping data to obtain user phase adjustment solutions. The risk warning module is used to conduct risk fusion assessment of low-voltage substations based on the distribution transformer critical temperature threshold and the user phase adjustment plan to obtain comprehensive risk grading data of the substations; based on the comprehensive risk grading data of the substations, risk-carrying capacity visualization mapping of the low-voltage substations is performed to obtain visual warning results of the substation carrying capacity.

Images