District voltage quality feature classification and adjustment means applicability evaluation method and system based on multi-source heterogeneous data

Abstract

The invention provides a transformer area voltage quality feature classification and adjustment means applicability evaluation method and system based on multi-source heterogeneous data, and belongs to the technical field of power system power distribution network operation control. Comprising the following steps: acquiring multi-source data of a transformer area, constructing an index system reflecting voltage quality influence factors and a sub-punishment index system, and calculating an index value according to the multi-source data; calculating a voltage quality score SV only considering a penalty term according to the index value of the score penalty; according to the index values reflecting the voltage quality influence factors and the index values reflecting the score punishment, establishing an index-SV relation model, and according to the index-SV relation model, obtaining a cause contribution decomposition result of each index and a corresponding transformer area classification; establishing a nonlinear multi-target reactive power regulation and control optimization model, and selecting an optimal regulation and control scheme from the applicable regulation and control schemes corresponding to the transformer area classification result according to the regulation and control scheme applicability scoring matrix R; and outputting the optimal regulation and control scheme.

Description

Technical Field

[0001] This invention belongs to the field of power system distribution network operation and control technology, specifically relating to a method and system for classifying distribution area voltage quality characteristics and evaluating the applicability of adjustment measures based on multi-source heterogeneous data. Background Technology

[0002] Voltage quality in distribution network areas directly affects power supply reliability and user experience. With the large-scale integration of distributed photovoltaic and other renewable energy sources, as well as seasonal load fluctuations and increased single-phase load connections, voltage quality issues in distribution areas are becoming increasingly complex. For example, rural and peri-urban distribution areas have long suffered from low voltage, particularly severe during peak summer load periods; conversely, areas with high photovoltaic output may experience voltage exceeding the upper limit at midday. Furthermore, uneven distribution of single-phase loads leads to three-phase voltage imbalance, resulting in significant differences in voltage levels for users in different locations. Traditional voltage quality assessment and mitigation methods have the following shortcomings:

[0003] Insufficient data utilization and indicator system: Traditional methods often rely on data from a single source (such as voltage monitoring of distribution transformers or user complaint information), which cannot comprehensively characterize the operating characteristics of the distribution area. The lack of a systematic indicator system to quantify the impact of factors such as seasonal load changes, photovoltaic penetration rate, three-phase imbalance, and network structure on voltage leads to incomplete assessment results and makes it difficult to identify the causes of voltage problems in a timely and accurate manner.

[0004] The assessment methods lack comprehensiveness and interpretability: Some existing studies employ machine learning and data fusion techniques to assess power quality in transformer substations, such as predicting multiple indicators like voltage exceedance probability, line loss rate, and harmonic distortion rate through multi-source data tensor fusion and multi-task learning. However, these methods are complex, requiring high-dimensional feature outer products and dimensionality reduction, lacking intuitive physical meaning and interpretability. Each indicator is predicted separately and requires threshold judgment, failing to provide an intuitive comprehensive evaluation to quantify the degree of voltage quality. Furthermore, the training and parameter selection of multi-task models are complex, limiting generalization and accuracy, making it difficult to adapt to changing on-site operating conditions.

[0005] Voltage control decisions are not precise enough: In existing technologies, voltage problems are often addressed by relying on preset rules or a single objective for adjustment decisions. For example, some methods select fixed reactive power or harmonic control strategies by comparing the relationship between over-limit probability, line loss rate, and threshold values. This approach does not make targeted selections based on the main differences in the problems of different transformer substations, nor does it quantitatively compare the effectiveness of various adjustment methods. This may result in either a "one-size-fits-all" approach using a certain general control scheme, or blindly coordinating multiple control methods, which is costly and difficult but does not achieve the best results. In addition, traditional optimization often focuses only on single objectives such as reducing voltage over-limit or line loss, without fully considering the constraints of investment costs, implementation time, and difficulty. Summary of the Invention

[0006] In view of this, the present invention provides a method and system for classifying voltage quality characteristics of distribution transformer areas and evaluating the applicability of regulation methods based on multi-source heterogeneous data. It utilizes data from multiple sources and of different types to comprehensively evaluate the voltage quality of low-voltage distribution transformer areas, including constructing a voltage quality characteristic index system, comprehensive voltage quality scoring, cause analysis and classification, and evaluation and optimization control of the applicability of voltage regulation methods, so as to achieve accurate diagnosis and optimization decision-making for voltage quality problems in distribution transformer areas.

[0007] The technical solution adopted by the embodiments of the present invention to solve its technical problem is as follows:

[0008] A method for classifying transformer substation voltage quality characteristics and evaluating the applicability of regulation measures based on multi-source heterogeneous data includes:

[0009] Step S1: Obtain multi-source data of the transformer area, construct an index system reflecting the factors affecting voltage quality and calculate the index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index, and network structure complexity index; construct a scoring penalty index system and calculate the index values ​​based on multi-source data, including voltage limit violation penalty, voltage fluctuation penalty, and spatial imbalance penalty.

[0010] Step S2: Calculate the voltage quality score SV, which only considers the penalty item, based on the index value of the scoring penalty.

[0011] Step S3: Based on the index values ​​reflecting the factors affecting voltage quality, establish an index-SV relationship model, and obtain the causal contribution decomposition results of each index and the corresponding transformer area classification based on the index-SV relationship model.

[0012] Step S4: Establish a nonlinear multi-objective reactive power control optimization model, and select the optimal control scheme from the applicable control schemes corresponding to the classification results of the transformer area based on the control scheme applicability score matrix R.

[0013] Step S5: Output the optimal control scheme.

[0014] Preferably, step S1 includes:

[0015] Acquire multi-source data, including three-phase voltage, current and active and reactive power data provided by distribution transformer monitoring terminals, load time-series curve data collected by user smart energy meters, photovoltaic output data provided by distributed photovoltaic inverter monitoring systems, and transformer topology and line parameters provided by geographic information systems (GIS) and distribution network asset archives;

[0016] Multi-source data are aligned according to timestamps, and a common time window and sampling period are selected to form a multi-dimensional time series dataset. Missing data are filled in and outliers are removed. Data from different sources and with different dimensions are standardized.

[0017] A set of quantitative indicators was extracted and calculated to characterize the key factors affecting the voltage quality of the transformer substation, including:

[0018] Seasonal load fluctuation index : ;

[0019] In the formula, This represents the maximum value among the average loads of each season. To be the minimum value, This represents the average annual load.

[0020] Three-phase imbalance index : ;

[0021] In the formula, These are the voltages of phases A, B, and C in the three-phase transformer area, respectively.

[0022] Photovoltaic penetration index : ;

[0023] In the formula, This refers to the total installed capacity of photovoltaic power within the distribution area. This represents the maximum load power of the transformer area.

[0024] Network structure complexity index : ;

[0025] In the formula, This refers to the number of branches of the low-voltage lines in the transformer substation area. The total number of power supply users in the transformer area. This refers to the total length of the low-voltage lines in the transformer substation area. This represents the maximum layering depth of the power distribution network. These are the weighting coefficients;

[0026] Voltage over-limit penalty value : ;

[0027] In the formula, This refers to the total time during which the voltage in the transformer substation exceeds the allowable voltage deviation range within the observation period. This is the total duration of the observation period;

[0028] Voltage fluctuation penalty value : ;

[0029] In the formula, and These represent the maximum and minimum voltage values ​​that occurred in the transformer area during the entire observation period, respectively. Nominal voltage;

[0030] Spatial Imbalance Penalty Value : ;

[0031] In the formula, This represents the maximum inter-node voltage difference in the transformer area at a given moment. Nominal voltage; , , The maximum value of the voltage at node i at time t. and minimum value difference.

[0032] Preferably, step S2, calculating the voltage quality score SV, includes: ;

[0033] In the formula, This represents the weighting coefficient of the penalty term.

[0034] Preferably, step S3 includes:

[0035] Establish an indicator-SV relationship model, using all feature indicators calculated in step S1 as input feature vectors. , in order to represent , , , ; build The mapping relationship is defined as a linear sum: ;

[0036] In the formula, For constant terms, ( ) represents the linear coefficients corresponding to each indicator;

[0037] Allocating the smallest amount of power to the distribution area The category label corresponding to the index of the value, min( This indicates the degree of the greatest negative contribution.

[0038] Preferably, step S4 includes:

[0039] M applicable control measures corresponding to predefined classification labels are used. Each applicable control measure is scored from four dimensions to obtain a control measure applicability score matrix. : ;

[0040] In the formula, , , , The evaluation criteria for the applicable control scheme a are: technical effectiveness, economic efficiency, timeliness, and difficulty.

[0041] Using matrix R, a two-step screening process is performed to obtain candidate control schemes:

[0042] First, based on the constraints of economic cost range, schedule range, and difficulty value range, unsuitable solutions are eliminated. The economic cost, schedule, and difficulty values ​​are all calculated using matrix R to determine the reverse penalty cost. Specifically, economic cost = basic cost × (1- / 100), Construction period = Basic construction period × (1- / 100), Difficulty value = Base difficulty × (1- / 100);

[0043] Secondly, solutions that do not meet the constraints of the technical effect range are eliminated; the technical effect is... ;

[0044] Establish a nonlinear multi-objective reactive power regulation optimization model and define the objective function. ,in, To adopt a decision variable vector The function uses 0-1 variables. Indicate whether to adopt the first For each candidate control scheme, a corresponding continuous variable is introduced to represent the degree of adjustment for measures that can be continuously adjusted.

[0045] Define the physical and operational constraints of the power grid, specifically the allowable range of voltage at each node, the limit of equipment output range, the limit of equipment output regulation range, power balance constraints, and branch power flow constraints;

[0046] The optimal decision variables can be obtained by using heuristic algorithms or hybrid intelligent algorithms. .

[0047] Preferably, voltage regulation measures include adjusting transformer tap positions, compensating for parallel capacitors, providing reactive power support for photovoltaic inverters, rebalancing three-phase loads, upgrading and modifying lines, adding on-load tap changers, and adding voltage stabilizing devices.

[0048] Preferably, in step S4, the objective function of the nonlinear multi-objective reactive power regulation optimization model is... The specific components are: ; ;

[0049] In the formula, Indicate decision The corresponding weights of total economic cost C(x), total implementation time T(x), and comprehensive technical difficulty D(x); C(x) and T(x) are obtained by weighting the scores of the selected measures in the R matrix; a is the index of the a-th candidate adjustment measure; Indicate whether method a is selected; For the configuration scale of method a; The fixed investment cost of means a; The cost per unit area of ​​investment; The score for method a in the economic dimension; , is the reverse penalty factor for a; The fixed duration of implementation for method a; The additional construction period per unit size; Rate it for "Timeliness"; The time penalty factor; The fixed implementation difficulty index of method a; To implement difficulty scoring; As a difficulty penalty factor;

[0050] The constraints also include constraints on the range of total investment cost, the range of total construction period, and the range of implementation difficulty.

[0051] A second aspect of the present invention provides a system for classifying transformer substation voltage quality characteristics and evaluating the applicability of adjustment methods based on multi-source heterogeneous data, the system comprising:

[0052] The construction unit is used to acquire multi-source data of the transformer area, construct an index system reflecting the factors affecting voltage quality and calculate index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index and network structure complexity index; construct a scoring penalty index system and calculate index values ​​based on multi-source data, including voltage limit violation penalty, voltage fluctuation penalty and spatial imbalance penalty;

[0053] The calculation unit is used to calculate the voltage quality score SV, which only considers the penalty item, based on the index value of the scoring penalty.

[0054] The classification unit is used to establish an index-SV relationship model based on the index values ​​that reflect the factors affecting voltage quality, and to obtain the causal contribution decomposition results of each index and the corresponding transformer area classification based on the index-SV relationship model.

[0055] Select a unit to establish a nonlinear multi-objective reactive power control optimization model. Based on the applicability score matrix R of the control scheme, select the optimal control scheme from the applicable control schemes corresponding to the classification results of the transformer area.

[0056] The output unit is used to output the optimal control scheme.

[0057] As can be seen from the above technical solution, the method and system for classifying the voltage quality characteristics of distribution areas and evaluating the applicability of regulation methods based on multi-source heterogeneous data provided in this embodiment of the invention first acquires multi-source data of the distribution area, constructs an index system reflecting the influencing factors of voltage quality and calculates the index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index, and network structure complexity index; constructs a scoring penalty index system and calculates the index values ​​based on the multi-source data, including voltage limit violation penalty, voltage fluctuation penalty, and spatial imbalance penalty; calculates the voltage quality score SV considering only the penalty item based on the index values ​​of the scoring penalty; establishes an index-SV relationship model based on the index values ​​reflecting the influencing factors of voltage quality, obtains the causal contribution decomposition results of each index and the corresponding distribution area classification based on the index-SV relationship model; establishes a nonlinear multi-objective reactive power regulation optimization model, selects the optimal regulation scheme from the applicable regulation schemes corresponding to the distribution area classification results based on the regulation scheme applicability scoring matrix R; and outputs the optimal regulation scheme. This invention utilizes data from multiple sources and of different types to comprehensively assess the voltage quality of low-voltage distribution transformer areas. This includes constructing a voltage quality characteristic index system, comprehensive voltage quality scoring, cause analysis and classification, and evaluation and optimization control of the applicability of voltage regulation methods, thereby achieving accurate diagnosis and optimized decision-making for voltage quality problems in transformer areas. Attached Figure Description

[0058] Figure 1 This is a flowchart of the method for classifying and evaluating the applicability of voltage quality characteristics and adjustment measures of transformer substations based on multi-source heterogeneous data, as presented in this invention.

[0059] Figure 2 This is a schematic diagram of the structure of the system for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of transformer substations based on multi-source heterogeneous data, as per the present invention. Detailed Implementation

[0060] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0061] This invention provides a method and system for classifying distribution transformer voltage quality characteristics and evaluating the applicability of regulation measures based on multi-source heterogeneous data. It comprehensively considers various factors affecting distribution transformer operation, establishing a feature system including indicators such as seasonal load changes, three-phase imbalance, distributed photovoltaic penetration rate, and network structure complexity, and calculates the comprehensive voltage quality score (SV) based on this system. An interpretive model decomposes the contribution rate of each factor to SV, enabling quantitative analysis and classification of the causes of voltage quality problems in the distribution transformer area. Furthermore, it combines multiple preset voltage regulation measures, scoring each measure from four dimensions: technical effectiveness, economic cost, implementation cycle, and technical difficulty, constructing a regulation measure applicability score matrix R. Finally, the evaluation results of this matrix are embedded into a nonlinear multi-objective reactive power regulation optimization model, with maximizing the voltage quality score as the primary objective while balancing constraints such as cost, timeliness, and difficulty, to obtain an optimized regulation scheme for the distribution transformer area. This invention's method is comprehensive and interpretable, and the system covers four stages: "monitoring and evaluation—cause diagnosis—measure evaluation—optimization decision-making," significantly improving the accuracy of voltage quality problem diagnosis and the targeting of regulation schemes.

[0062] refer to Figure 1 As shown, the method for classifying and evaluating the applicability of voltage quality characteristics and adjustment measures of transformer substations based on multi-source heterogeneous data according to the present invention includes:

[0063] Step S1: Obtain multi-source data of the transformer area, construct an index system reflecting the factors affecting voltage quality and calculate the index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index, and network structure complexity index; construct a scoring penalty index system and calculate the index values ​​based on multi-source data, including voltage limit violation penalty, voltage fluctuation penalty, and spatial imbalance penalty.

[0064] Step S2: Calculate the voltage quality score SV, which only considers the penalty item, based on the index value of the scoring penalty.

[0065] Step S3: Based on the index values ​​reflecting the factors affecting voltage quality, establish an index-SV relationship model, and obtain the causal contribution decomposition results of each index and the corresponding transformer area classification based on the index-SV relationship model.

[0066] Step S4: Establish a nonlinear multi-objective reactive power control optimization model, and select the optimal control scheme from the applicable control schemes corresponding to the classification results of the transformer area based on the control scheme applicability score matrix R.

[0067] Step S5: Output the optimal control scheme.

[0068] Step S1 is used for data preprocessing and indicator construction, aiming to clean, align, and extract features from multi-source heterogeneous data related to the transformer substation area, and establish an indicator system reflecting the factors affecting the voltage quality of the substation area. Different data sources have varying time granularities and completeness, requiring unified processing. The specific steps are as follows:

[0069] Acquire multi-source data, including but not limited to: three-phase voltage, current and active and reactive power data provided by distribution transformer monitoring terminals, load time-series curve data collected by user smart energy meters, photovoltaic output data provided by distributed photovoltaic inverter monitoring systems, and multi-source data such as transformer topology and line parameters provided by geographic information systems (GIS) and distribution network asset archives;

[0070] First, align the multi-source data according to timestamps, selecting a common time window and sampling period (e.g., 15 minutes or 1 hour as the step size) to form a multidimensional time series dataset. For missing data points, use appropriate methods for interpolation or imputation (e.g., linear interpolation, mean imputation, or imputation based on adjacent day patterns) to ensure data integrity. For obviously anomalous data (e.g., spikes exceeding the physically reasonable range), identify them according to a set threshold or through statistical methods, and remove or correct them to reduce the impact of noise on subsequent analysis. After alignment and cleaning, standardize the data from different sources and with different dimensions, for example, using Z-score standardization or interval scaling, to convert each dimension of data into zero mean, unit variance, or a uniform numerical range. This step can eliminate differences in the dimensions of different indicators and prevent one dimension of data from dominating the results in subsequent calculations.

[0071] A set of quantitative indicators was extracted and calculated to characterize the key factors affecting the voltage quality of the transformer substation, including:

[0072] Seasonal load fluctuation index : (1)

[0073] In the formula, This represents the maximum value among the average loads of each season. To be the minimum value, This represents the average annual load. It reflects the magnitude of load variation in the transformer area with the seasons. A year can be divided into several typical seasons (such as spring, summer, autumn, and winter), and the average load level or peak load for each season can be calculated. A larger value indicates that the load in the distribution area has more obvious seasonal fluctuations. Significant differences in load levels may lead to significant differences in voltage levels in different seasons.

[0074] Three-phase imbalance index : (2)

[0075] In the formula, These are the voltages of phases A, B, and C in the three-phase distribution area (each taken as the steady-state effective value at a representative moment, such as the value measured during the daily peak load period), and the denominator is the average value of the three-phase voltages; This reflects the degree of imbalance in the three-phase voltage / current distribution within the transformer area. A balance of 0 indicates that the three-phase voltage amplitudes are completely equal, while a higher value indicates a greater deviation in the three-phase voltage. Uneven connection of a single-phase load can cause the voltage of that phase to be too low or even exceed its limit. The larger the value, the more serious the three-phase load imbalance problem is usually.

[0076] Photovoltaic penetration index : (3)

[0077] In the formula, This refers to the total installed photovoltaic capacity within the distribution area (which can be obtained from installation registration or the sum of inverter rated power). The maximum load power of the transformer area (e.g., the highest annual load power selected based on historical data); This measures the penetration of distributed photovoltaic (PV) systems into a transformer substation, representing the relative ratio of PV installations to load capacity. When... When the load is significantly higher than 100% (e.g., the photovoltaic scale exceeds the local maximum load), during periods of low load and high sunlight, a large number of photovoltaic cells may be connected to the grid, causing voltage rises or even exceeding the upper limit; conversely, A very low voltage indicates that the area is mainly supplied by traditional loads, and the impact of photovoltaics on voltage can be ignored.

[0078] Network structure complexity index This characterizes the impact of the low-voltage distribution network topology of a transformer substation on voltage distribution. Network complexity considers factors such as the number of line branches, power supply radius, and user distribution within the substation. For example, it can be defined using the following weighted combination form: (4)

[0079] In the formula, This refers to the number of branches of the low-voltage lines in the transformer substation area. The total number of power supply users in the transformer area. Total length of low-voltage lines in the transformer area (unit: kilometers); This refers to the maximum hierarchical depth of the power distribution network (e.g., the number of vertical levels of the main line and its downstream branches). This is a weighting coefficient used to balance the influence of the dimensions of each component. This index comprehensively reflects the power supply range and complexity of the network: when the number of users is constant, a large number of branches, long buses, and deep hierarchies indicate that the network has a tree-like radial structure and a large power supply radius, with significant voltage drops at remote nodes and uneven voltage spatial distribution; at this time... A larger value is obtained. Conversely, for simple and compact transformer substations (such as those with concentrated users, short lines, and few branches), a larger value is obtained. The value is relatively small.

[0080] The aforementioned indicator system comprehensively quantifies the main heterogeneous factors affecting voltage quality in distribution areas. In practical applications, the indicator set can be expanded or adjusted based on data availability and the characteristics of the distribution area. For example, it can also include transformer load rate indicators (the ratio of actual transformer load to rated capacity) and reactive power compensation adequacy indicators (the proportion of installed compensation capacity to theoretical demand). This invention is not limited to specific indicators, but requires that the selected indicators can clearly measure specific factors and are easy to calculate. After all indicators are calculated, the indicator values ​​can be normalized or standardized as needed to make their value ranges similar, providing good input for subsequent model analysis.

[0081] Voltage over-limit penalty value This reflects the severity of the voltage deviation in the distribution area exceeding the allowable range. Assume the allowable voltage deviation range specified for the distribution network is... (e.g., nominal value) ,Right now , During the monitoring period, the duration or frequency of voltage violations is recorded, and the degree of violation is expressed as the percentage of duration exceeding the limit. (5)

[0082] In the formula, The total time (or equivalent time length, such as the cumulative number of minutes per day) during the observation period when the voltage in the transformer area exceeds the allowable voltage deviation range. This is the total duration of the observation period; The physical meaning is the percentage of time the voltage is not up to standard. For example... This indicates that the voltage exceeds the standard (too high or too low) approximately 5% of the time. This indicates that the voltage is consistently within the acceptable range. A larger penalty value indicates more frequent or longer-lasting voltage exceedances, and poorer voltage quality. For transformer substations with multiple measuring points, the most severe exceedance among all nodes can be used for calculation. The total time that the voltage at any node is out of specification is recorded. If necessary, the degree of overvoltage and undervoltage can be further subdivided, but this invention prefers to use a simple overall over-limit penalty value.

[0083] Voltage fluctuation penalty value This is used to quantify the voltage fluctuation amplitude in the time domain of a distribution area. Even if the voltage does not exceed the limit, significant fluctuations in the voltage value can still affect electrical equipment and power supply stability. The ratio of the peak-to-valley range of the voltage in the monitoring area to the nominal voltage can be taken: (6)

[0084] In the formula, and These represent the maximum and minimum voltage values ​​that occurred in the transformer area during the entire observation period (which can be understood as the global voltage peaks and troughs). This refers to the nominal voltage (e.g., 220V corresponds to 1.0 pu); such as This indicates that during the monitoring period, the voltage value reached a maximum of 8% higher than the nominal value at a certain moment, and a minimum of 8% lower than the nominal value at another moment (these two events may occur at different times). It should be noted that, to eliminate the influence of extreme anomalies, calculations can be performed... and A few outliers are ignored, or the average maximum / minimum value within a certain time window is used as an approximation. Voltage fluctuation penalty reflects the stability of the distribution area's voltage supply: the larger the value, the more drastic the voltage change over time. For example, after photovoltaic installation, the voltage rises at noon and falls during nighttime when the load is high; such drastic intraday fluctuations will result in higher voltage. .

[0085] Spatial Imbalance Penalty Value This is used to reflect the degree of voltage level difference between different locations within a transformer area at the same time, i.e., the spatial uniformity of voltage distribution. Due to line voltage drops, the voltage of users closer to the transformer is often higher than that of users farther away; three-phase load imbalance can also cause asynchronous voltage drops or rises between phases. This invention uses the ratio of the instantaneous maximum voltage difference within the network to its nominal value to measure spatial imbalance:

[0086] First, for each sampling time Find the difference between the maximum and minimum values ​​of the voltage at all nodes in the area. This represents the voltage distribution range at that moment. Then, the maximum value of this difference over the entire observation period is taken. Define a space imbalance penalty by comparing it to the nominal voltage: (7)

[0087] in This represents the maximum inter-node voltage difference in the transformer area at a certain moment (this moment is usually when the voltage is lowest at the far end and still higher at the near end due to full load or other operating conditions). This is the nominal voltage. A higher value indicates a more inconsistent voltage distribution within the transformer substation, with larger voltage level differences between different users. This is especially true for substations with long power supply radii and numerous branches, where significant voltage differences may occur between the beginning and end of the substation, and between the phase with the highest and lowest voltage, leading to high voltage spikes. Conversely, if the transformer substation has a simple structure or a mid-course voltage regulating device, the voltage at each point will be relatively close. Smaller.

[0088] Step S2 sets the baseline full score to 100 points (representing the score under ideal conditions with no voltage problems). The calculated voltage quality score SV for the transformer area can then be calculated according to the following relationship: (8)

[0089] In the formula, The weighting coefficients for penalty items can be set according to actual needs. These weights reflect the relative proportion of the impact of the three types of voltage problems on the overall score. For example, weights can be allocated based on the power sector's emphasis on voltage compliance rate (exceeding limits), fluctuations, three-phase imbalance, etc., or simply by taking equal weights (making the weighting coefficients equal to the weighting coefficients). And each is 1 / 3). If If all values ​​are 0 (ideally, there are no violations, the voltage is constant, and the values ​​at all points are equal), then... If problems exist, these penalty values ​​will make The corresponding points will be deducted from 100. Note that in the above formula... These are percentage values; to maintain consistency of dimensions, It should be calibrated according to the appropriate proportion. For example, if equal weights are used... If we use a percentage system, then Then when a certain area hour, Points. Generally speaking, The score ranges from 0 to 100, with lower scores indicating more severe voltage problems. Using a comprehensive score to measure voltage quality allows for convenient horizontal comparisons between different transformer substations or longitudinal tracking of the voltage status of the same substation at different times.

[0090] The SV scoring model of this invention is intuitive and scalable. If other power quality factors (such as harmonics, voltage sags, etc.) need to be considered, corresponding penalty terms and weights can be added to the scoring model to obtain an extended comprehensive score. In this embodiment, the focus is on issues related to steady-state voltage deviation; therefore, the above three main penalty terms are selected to constitute the SV.

[0091] Step S3 involves causal contribution decomposition and transformer area feature classification. After obtaining the transformer area feature indicators and voltage quality score (SV), this step performs correlation analysis to quantify the contribution of each indicator to voltage quality, and then classifies the transformer areas according to the dominant factors. The core idea is to use an interpretable mathematical model to fit SV as the dependent variable and the feature indicators as the independent variables to obtain the coefficients or importance measures of each indicator, thereby determining which factors are the main causes of voltage quality problems in the transformer area. Unlike the implicit relationship mining of black-box complex models (such as deep neural networks, tensor high-order features, etc.), this invention uses models with clear causal structures such as linear regression and decision trees to ensure that the output results are easy to understand and apply. The specific implementation process is as follows:

[0092] Establish an indicator-SV relationship model, using all feature indicators calculated in step S1 as input feature vectors. , in order to represent , , , Other parameters; construction In a preferred embodiment, the mapping relationship is fitted using a multiple linear regression model, i.e., assuming... It exhibits a linear summation relationship with each indicator: (9)

[0093] In the formula, For constant terms, ( The coefficients (regression coefficients) are the linear coefficients corresponding to each indicator. These coefficients can be estimated using a large amount of historical sample data (observations from different time periods or different monitoring areas) and methods such as least squares. The advantage of the linear model is that each coefficient directly reflects the direction and relative magnitude of the corresponding indicator's influence on SV. For regular expressions Increasing the value will improve the SV score; a negative value indicates... Increasing the value leads to a decrease in SV; The larger the value, the more likely it is to indicate The greater the impact on SV, the more likely it is to cause problems. In many scenarios, voltage quality exhibits an approximately linear relationship with these physical factors. For example, high photovoltaic penetration often reduces voltage quality scores (negative correlation), as does high three-phase imbalance (negative correlation). A certain degree of network structure improvement (such as...) can also negatively affect SV. Decreasing the number of features will increase the SV score (negative correlation). The linear assumption can well characterize and explain these trends. Of course, if nonlinear effects need to be considered, methods such as decision tree regression and random forest can also be used. These methods can characterize the influence of factors by splitting nodes or feature importance. However, in general, this module does not introduce product or higher-order terms between features, avoiding complex feature interactions in the form of outer product tensors, and ensuring the interpretability of the model results.

[0094] Allocating the smallest amount of power to the distribution area The category label corresponding to the index of the value, min( This indicates the degree of the greatest negative contribution.

[0095] Step S4 is used for nonlinear multi-objective reactive power regulation optimization:

[0096] First, construct a scoring matrix for the applicability of the control scheme. This module addresses different types of voltage problems in transformer substations, predefines a series of optional voltage regulation measures, and quantitatively evaluates the applicability of each scheme from multiple dimensions, constructing a rating matrix for the applicability of regulation methods. Common voltage regulation methods include, but are not limited to: transformer tap changer adjustment, parallel capacitor compensation, reactive power support from photovoltaic inverters, three-phase load rebalancing, line upgrades (such as increasing conductor cross-section), and the addition of on-load tap changers or voltage stabilizers. For each alternative method, this invention evaluates it from four aspects: voltage improvement effect, economic cost, implementation period, and technical difficulty. Each dimension's score can use an appropriate quantitative scale (e.g., a 0-10 or 0-100 point system) to evaluate the method's merits in the corresponding aspect. Taking this type of regulatory measure as an example, its scores in terms of technical effectiveness, economy, timeliness, and difficulty are as follows: , , , Then the scoring results of all alternative methods can be summarized into a matrix: (10)

[0097] matrix This invention intuitively presents the scoring of different measures across various aspects, serving as a multi-dimensional quantitative evaluation tool for the applicability of each measure. Existing technologies often rely on expert experience to directly select a measure or rank options based solely on a single indicator (such as investment cost). However, this invention constructs... The matrix quantifies and compares the advantages and disadvantages of each adjustment measure, making the performance of different schemes clear at a glance and helping decision-makers make comprehensive trade-offs. In practical applications, especially for distribution areas with high photovoltaic penetration or complex network structures, this matrix allows for in-depth analysis of... The matrix can clearly identify which control measures are more advantageous in terms of technical effectiveness and which measures are more acceptable in terms of economy or implementation difficulty, thus providing a basis for subsequent optimization decisions.

[0098] It should be noted that the construction The matrix scoring process can be completed based on empirical data or expert scoring. For example, in the technical effectiveness dimension, the degree of improvement of a measure on voltage compliance rate or fluctuation range can be evaluated based on simulation or historical implementation results; the economic cost dimension is scored based on investment and operation and maintenance costs; the implementation cycle dimension considers the time required from scheme formulation to operation; and the technical difficulty dimension evaluates the magnitude of technical challenges encountered during the implementation of the scheme (such as construction complexity, impact on existing systems, etc.). A unified scoring standard ensures the rationality and consistency of horizontal comparisons between different methods within the matrix. Once the matrix is ​​constructed, it provides a quantitative evaluation basis for the next step of optimization decision-making.

[0099] Using matrix R, a two-step screening process is performed to obtain candidate control schemes:

[0100] First, based on the constraints of economic cost range, schedule range, and difficulty value range, unsuitable solutions are eliminated. The economic cost, schedule, and difficulty values ​​are all calculated using matrix R to determine the reverse penalty cost. Specifically, economic cost = basic cost × (1- / 100), Construction period = Basic construction period × (1- / 100), Difficulty value = Base difficulty × (1- / 100);

[0101] Secondly, solutions that do not meet the constraints of the technical effect range are eliminated; the technical effect is... ;

[0102] Obtaining voltage quality scores for transformer substations and the applicability matrix of each adjustment method Subsequently, this module embeds the evaluation results into the optimization decision model to solve for the optimal control scheme for the transformer area. Specifically, based on the aforementioned transformer area type and available control measures, a multi-objective optimization model is established, with maximizing the improvement of voltage quality score as the primary objective. Simultaneously, factors such as economic cost, implementation cycle, and technical difficulty are considered as constraints or secondary objectives to seek the optimal balance between technical benefits and cost-effectiveness. The decision variables of the optimization model may include the selection and configuration of each candidate control measure. For example, 0-1 variables can be used. Indicate whether to adopt the first Various regulatory measures ( For selection, (If not adopted), for continuously adjustable measures, corresponding continuous variables can be introduced to represent their degree of adjustment (such as the size of reactive power compensation capacity, the tap adjustment position of transformers, etc.). Based on this, the objective function and constraints are constructed.

[0103] Establish a nonlinear multi-objective reactive power regulation optimization model and define the objective function. Based on the aforementioned model, It is a function of voltage over-limit, fluctuation, and imbalance in the distribution area, which in turn depends on the voltage regulation measures taken and their extent. It can be represented as a vector of decision variables. function Use 0-1 variables Indicate whether to adopt the first For each candidate control scheme, a corresponding continuous variable is introduced to represent the degree of adjustment for measures that can be continuously adjusted.

[0104] The optimization model also needs to include power grid physical and operational constraints, such as ensuring that the voltage at each node is within allowable limits (through...). The optimization problem implicitly pursues this goal, and is constrained by limitations on equipment output or adjustment range (such as the upper limit of capacitor compensation capacity, inverter reactive power margin, transformer tap adjustment range, etc.), as well as electrical network equations involving power balance and branch power flow constraints. These constraints typically exhibit nonlinear characteristics, making the entire optimization problem a mixed-integer nonlinear programming (MINLP) problem. For this optimization model, heuristic algorithms (such as genetic algorithms, particle swarm optimization, etc.) or hybrid intelligent algorithms can be used to solve it, obtaining the optimal decision variables. The optimization solution provides the optimal control scheme for the target transformer area, including the combination of control measures to be taken and their specific parameter settings.

[0105] As a preferred approach, constraints are introduced regarding the cost, timeliness, and difficulty of each adjustment measure. For example, constraints can be set such as total investment cost not exceeding the budget limit, total construction period not exceeding the predetermined duration, and implementation difficulty being controlled within an acceptable range. Alternatively, these factors can be treated as secondary objectives that need to be optimized simultaneously, and combined with the primary objective into a single objective function using weighted coefficients. For example, using... If we consider the weights that decision-makers place on cost, time, and difficulty, then we can optimize the following comprehensive objective: (11) (12)

[0106] In the formula, Indicate decision The corresponding weights of total economic cost C(x), total implementation time T(x), and comprehensive technical difficulty D(x); C(x) and T(x) are obtained by weighting the scores of the selected measures in the R matrix; a is the index of the a-th candidate adjustment measure; Indicate whether method a is selected; The configuration scale / adjustment range of method a (such as kvar, number of gear steps, number of units, etc.); The fixed investment cost of means a; The investment cost per unit size (ten thousand yuan / unit); The score for method a in terms of economic efficiency (the higher the score, the more money is saved). , is the reverse penalty factor for a; The fixed implementation duration (in days) for method a; The additional construction period per unit (days / unit); Rate the "timeliness" (a higher score indicates faster implementation); The time penalty factor; The fixed implementation difficulty index (0–10) for method a. The increment of difficulty per unit size ( / unit); The difficulty of implementation is scored (the higher the score, the easier it is to implement); This serves as a difficulty penalty factor; the above processing is equivalent to transforming multi-objective optimization into weighted single-objective optimization. In this embodiment of the invention, to simplify calculations, a method is preferentially adopted that treats cost, cycle, and difficulty as constraints, that is, maximizing while satisfying the constraints. This is to avoid bias caused by improper weight selection.

[0107] The constraints also include constraints on the range of total investment cost, the range of total construction period, and the range of implementation difficulty.

[0108] A second aspect of the present invention provides a system for classifying transformer substation voltage quality characteristics and evaluating the applicability of adjustment methods based on multi-source heterogeneous data, the system comprising:

[0109] The construction unit is used to acquire multi-source data of the transformer area, construct an index system reflecting the factors affecting voltage quality and calculate index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index and network structure complexity index; construct a scoring penalty index system and calculate index values ​​based on multi-source data, including voltage limit violation penalty, voltage fluctuation penalty and spatial imbalance penalty;

[0110] The calculation unit is used to calculate the voltage quality score SV, which only considers the penalty item, based on the index value of the scoring penalty.

[0111] The classification unit is used to establish an index-SV relationship model based on the index values ​​that reflect the factors affecting voltage quality, and to obtain the causal contribution decomposition results of each index and the corresponding transformer area classification based on the index-SV relationship model.

[0112] Select a unit to establish a nonlinear multi-objective reactive power control optimization model. Based on the applicability score matrix R of the control scheme, select the optimal control scheme from the applicable control schemes corresponding to the classification results of the transformer area.

[0113] The output unit is used to output the optimal control scheme.

[0114] The specific implementation methods for each of the above units are as described in the preceding steps.

[0115] The actual effects of the present invention will be described below based on specific embodiments.

[0116] Engineering verification was conducted in a city's power distribution network, selecting 62 low-voltage distribution transformer areas as samples, including 41 high-penetration distributed photovoltaic transformer areas and 21 ordinary residential transformer areas. Based on one month of historical multi-source measurement data (distribution transformer monitoring terminals, electricity meters, photovoltaic inverters, GIS topology, and asset archives), an indicator system was constructed using data from 45 of these transformer areas, and a causal analysis model was trained. The remaining 17 transformer areas served as an independent test set to verify the evaluation and control effects of the invention.

[0117] First, voltage quality diagnosis and control scheme formulation were carried out for the test set of transformer areas using both the "traditional method" and the "method of this invention". The traditional method uses only single statistical indicators such as voltage exceedance rate and pass rate, allowing operators to select a primary control measure (such as uniform capacitor switching or general OLTC adjustment) based on experience, without distinguishing the cause types of transformer area problems or quantitatively comparing the economics and implementation difficulty of different measures. The method of this invention, however, follows the steps in the instruction manual: in stages S1-S4, it establishes multi-source voltage quality characteristic indicators, calculates the comprehensive score SV, performs cause contribution decomposition and transformer area classification, constructs the applicability matrix R of control measures, and in stage S5, it uses a nonlinear multi-objective optimization model to jointly consider "improving SV, controlling costs, shortening the construction period, and reducing implementation difficulty," automatically providing a combined control scheme.

[0118]

[0119] Experimental results show that the method of this invention is significantly superior to traditional empirical methods in terms of accuracy in identifying problematic transformer areas, improvement in voltage quality after treatment, and control of treatment costs and schedule. Comparing voltage quality score (SV) and key operational indicators: Under the traditional method, 3 out of 17 transformer areas in the test set still exhibited significant deviations from the upper or lower limits, with the overall average SV increasing from 83.4 to 88.5, and the average duration of voltage exceedances decreasing from 7.1% to 4.0%. With the method of this invention, all problematic transformer areas were correctly identified and differentiated control solutions were provided, with the average SV increasing to 93.6, and the duration of voltage exceedances further decreasing to 1.9%. Compared to the traditional method, the present invention increased the voltage quality score improvement on the test set by approximately 9.7% (from +5.1 to +10.2), and the reduction in exceedance duration increased from 43.7% to 73.2%, demonstrating the significant advantages of this invention in identifying the root cause and providing precise solutions.

[0120] Compared with the prior art, the present invention has the following beneficial effects:

[0121] This invention utilizes multi-source data to construct explicit indicators, improving the comprehensiveness and transparency of assessments. It integrates heterogeneous data from multiple sources, including electricity consumption information collection systems, distribution transformer monitoring, distributed power source monitoring, and GIS topology. After preprocessing, explicit indicators reflecting different aspects of transformer substation characteristics are extracted, such as seasonal load fluctuation rate, three-phase imbalance, photovoltaic penetration rate, and network structure complexity. Clear calculation formulas and physical meanings are provided. Compared to existing methods that use neural networks for black-box feature extraction or tensor fusion of multi-dimensional data, the indicator system constructed in this invention is more intuitive and understandable, facilitating maintenance personnel's understanding of transformer substation operating characteristics and providing a transparent basis for subsequent decision-making.

[0122] This invention introduces a comprehensive voltage quality score (SV) to replace the fragmented evaluation of multiple indicators. Addressing the traditional method's need to separately predict multiple indicators such as voltage exceedance, line loss, and harmonics, this invention innovatively proposes the comprehensive SV score. It synthesizes a single score by weighting factors such as voltage exceedance penalties, fluctuation penalties, and spatial imbalance penalties. The SV score intuitively depicts the overall quality of voltage in a distribution area, avoiding the inconvenience of evaluating and weighing multiple indicators separately, and providing a unified benchmark for setting optimization targets. Especially when evaluating distribution areas with complex scenarios including high-penetration photovoltaics, the SV score can comprehensively reflect the severity of various voltage problems, making the evaluation results more holistic.

[0123] This invention employs interpretable models for causal contribution analysis and transformer substation classification. It uses interpretable models such as linear regression or decision trees to correlate the aforementioned feature indicators with SV scores, calculating the contribution rate of each indicator to SV and clarifying the degree of influence of each factor in voltage quality issues. Unlike existing methods that require indirect capture of factor influence through high-order interactive features and multi-task architectures, this invention avoids dimensional outer products and high-order coupling of feature vectors, and does not introduce complex operations such as tensor fusion, preserving model transparency through linear summation. Through contribution rate analysis, transformer substations can be classified into different feature types (such as seasonal load-dominated, photovoltaic overvoltage, three-phase imbalance, weak network structure, or multi-factor mixed types), enabling maintenance personnel to clearly identify the "cause" and thus laying the foundation for differentiated control strategies.

[0124] This invention constructs an applicability matrix for regulation methods to achieve quantitative evaluation of optimal measures. For different types of voltage problems in distribution transformer areas, this invention predefines a series of selectable voltage regulation methods (including but not limited to transformer tap adjustment, parallel capacitor compensation, reactive power support from photovoltaic inverters, three-phase load rebalancing, overhead line modification, or the addition of voltage regulating equipment, etc.), and scores each method from four aspects: voltage improvement effect, economy, timeliness, and implementation difficulty, forming an applicability scoring matrix R. This matrix-based evaluation tool is unprecedented in existing technologies: traditionally, methods often rely on experience to directly select a measure or use a single dimension (such as cost) for ranking. This invention, however, uses multi-dimensional quantitative scoring to make the advantages and disadvantages of different measures readily apparent. Especially for distribution transformer areas with high photovoltaic penetration or complex networks, the R matrix can clearly identify which regulation method is technically most effective, economically most cost-effective, and quickest and easiest to implement, providing a scientific basis for optimization decisions.

[0125] This invention proposes a multi-objective optimization decision-making approach that integrates applicability evaluation to improve the pertinence and overall effectiveness of control schemes. The invention embeds the aforementioned R-matrix evaluation results into the optimization model for reactive power and voltage control. It sets maximizing the SV score as the primary objective and uses cost, timeliness, and difficulty as additional objectives to be optimized. By determining reasonable weighting coefficients, the multi-objective problem is transformed into a single comprehensive objective function for nonlinear optimization. Compared to existing practices that select fixed "cooperative control" strategies based on threshold conditions, the optimization model of this invention can dynamically adjust the strategy combination according to different transformer substation types, prioritizing measures with high applicability scores to achieve the optimal balance between technical and economic benefits. For example, for low-voltage substations dominated by seasonal loads, the model tends to choose measures to increase the base voltage (such as adjusting transformer taps or adding local compensation), weighing the investment cost and construction period. For photovoltaic overvoltage substations, it prioritizes measures to absorb excess reactive power (such as adjusting photovoltaic inverters or adding reactive power-consuming equipment) to eliminate voltage exceedances at the lowest cost. While ensuring voltage quality meets standards, we should minimize investment and maintenance costs to improve the feasibility and effectiveness of the solution.

[0126] In summary, this invention provides a complete solution for voltage quality assessment and optimization in distribution areas, with innovative designs at the data layer, model layer and decision layer. It can more accurately identify the causes of voltage problems and formulate optimization measures according to local conditions, thereby improving the management level of voltage quality at the end of the distribution network.

[0127] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the above-described embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for classifying transformer substation voltage quality characteristics and evaluating the applicability of adjustment measures based on multi-source heterogeneous data, characterized in that, include: Step S1: Obtain multi-source data of the transformer area, construct an index system reflecting the factors affecting voltage quality and calculate the index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index, and network structure complexity index; construct a scoring penalty index system and calculate the index values ​​based on multi-source data, including voltage limit violation penalty, voltage fluctuation penalty, and spatial imbalance penalty. Step S2: Calculate the voltage quality score SV, which only considers the penalty item, based on the index value of the scoring penalty. Step S3: Based on the index values ​​reflecting the factors affecting voltage quality, establish an index-SV relationship model, and obtain the causal contribution decomposition results of each index and the corresponding transformer area classification based on the index-SV relationship model. Step S4: Establish a nonlinear multi-objective reactive power control optimization model, and select the optimal control scheme from the applicable control schemes corresponding to the classification results of the transformer area based on the control scheme applicability score matrix R. Step S5: Output the optimal control scheme.

2. The method for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of distribution areas based on multi-source heterogeneous data as described in claim 1, characterized in that, Step S1 includes: Acquire multi-source data, including three-phase voltage, current and active and reactive power data provided by distribution transformer monitoring terminals, load time-series curve data collected by user smart energy meters, photovoltaic output data provided by distributed photovoltaic inverter monitoring systems, and transformer area topology and line parameters provided by geographic information systems and distribution network asset archives. Multi-source data are aligned according to timestamps, and a common time window and sampling period are selected to form a multi-dimensional time series dataset. Missing data are filled in and outliers are removed. Data from different sources and with different dimensions are standardized. A set of quantitative indicators was extracted and calculated to characterize the key factors affecting the voltage quality of the transformer substation, including: Seasonal load fluctuation index : ; In the formula, This represents the maximum value among the average loads of each season. To be the minimum value, This represents the average annual load. Three-phase imbalance index : ; In the formula, These are the voltages of phases A, B, and C in the three-phase transformer area, respectively. Photovoltaic penetration index : ; In the formula, This refers to the total installed capacity of photovoltaic power within the distribution area. This represents the maximum load power of the transformer area. Network structure complexity index : ; In the formula, This refers to the number of branches of the low-voltage lines in the transformer substation area. The total number of power supply users in the transformer area. This refers to the total length of the low-voltage lines in the transformer substation area. This represents the maximum layering depth of the power distribution network. These are the weighting coefficients; Voltage over-limit penalty value : ; In the formula, This refers to the total time during which the voltage in the transformer substation exceeds the allowable voltage deviation range within the observation period. This is the total duration of the observation period; Voltage fluctuation penalty value : ; In the formula, and These represent the maximum and minimum voltage values ​​that occurred in the transformer area during the entire observation period, respectively. Nominal voltage; Spatial Imbalance Penalty Value : ; In the formula, This represents the maximum inter-node voltage difference in the transformer area at a given moment. Nominal voltage; , , The maximum value of the voltage at node i at time t. and minimum value difference.

3. The method for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of distribution areas based on multi-source heterogeneous data as described in claim 2, characterized in that, Step S2, which calculates the voltage quality score SV, includes: ; In the formula, This represents the weighting coefficient of the penalty term.

4. The method for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of distribution areas based on multi-source heterogeneous data as described in claim 3, characterized in that, Step S3 includes: Establish an indicator-SV relationship model, using all feature indicators calculated in step S1 as input feature vectors. , in order to represent , , , ; build The mapping relationship is defined as a linear sum: ; In the formula, For constant terms, ( ) represents the linear coefficients corresponding to each indicator; Allocating the smallest amount of power to the distribution area The category label corresponding to the index of the value, min( This indicates the degree of the greatest negative contribution.

5. The method for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of distribution areas based on multi-source heterogeneous data as described in claim 4, characterized in that, Step S4 includes: M applicable control measures corresponding to predefined classification labels are used. Each applicable control measure is scored from four dimensions to obtain a control measure applicability score matrix. : ; In the formula, , , , The evaluation criteria for the applicable control scheme a are: technical effectiveness, economic efficiency, timeliness, and difficulty. Using matrix R, a two-step screening process is performed to obtain candidate control schemes: First, based on the constraints of economic cost range, schedule range, and difficulty value range, unsuitable solutions are eliminated. The economic cost, schedule, and difficulty values ​​are all calculated using matrix R to determine the reverse penalty cost. Specifically, economic cost = basic cost × (1- / 100), Construction period = Basic construction period × (1- / 100), Difficulty value = Base difficulty × (1- / 100); Secondly, solutions that do not meet the constraints of the technical effect range are eliminated; the technical effect is... ; Establish a nonlinear multi-objective reactive power regulation optimization model and define the objective function. ,in, To adopt a decision variable vector The function uses 0-1 variables. Indicate whether to adopt the first For each candidate control scheme, a corresponding continuous variable is introduced to represent the degree of adjustment for measures that can be continuously adjusted. Define the physical and operational constraints of the power grid, specifically the allowable range of voltage at each node, the limit of equipment output range, the limit of equipment output regulation range, power balance constraints, and branch power flow constraints; The optimal decision variables can be obtained by using heuristic algorithms or hybrid intelligent algorithms. .

6. The method for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of distribution areas based on multi-source heterogeneous data as described in claim 5, characterized in that, Voltage regulation measures include adjusting transformer tap positions, compensating for parallel capacitors, providing reactive power support for photovoltaic inverters, rebalancing three-phase loads, upgrading and modifying lines, adding on-load tap changers, and adding voltage stabilizing devices.

7. The method for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods of distribution areas based on multi-source heterogeneous data as described in claim 6, characterized in that, In step S4, the objective function of the nonlinear multi-objective reactive power regulation optimization model is... The specific components are: ; ; In the formula, Indicate decision The corresponding weights of total economic cost C(x), total implementation time T(x), and comprehensive technical difficulty D(x); C(x) and T(x) are obtained by weighting the scores of the selected measures in the R matrix; a is the index of the a-th candidate adjustment measure; Indicate whether to choose method a; For the configuration scale of method a; The fixed investment cost of means a; The cost per unit area of ​​investment; The score for method a in the economic dimension; , is the reverse penalty factor for a; The fixed duration of implementation for method a; The additional construction period per unit size; Rate it for "Timeliness"; The time penalty factor; The fixed implementation difficulty index of method a; The increment in difficulty per unit size; To implement difficulty scoring; As a difficulty penalty factor; The constraints also include constraints on the range of total investment cost, the range of total construction period, and the range of implementation difficulty.

8. A system for classifying and evaluating the applicability of voltage quality characteristics and adjustment methods for transformer substations based on multi-source heterogeneous data, characterized in that, The system for implementing the method according to any one of claims 1-7 comprises: The construction unit is used to acquire multi-source data of the transformer area, construct an index system reflecting the factors affecting voltage quality and calculate index values, including seasonal fluctuation index, three-phase imbalance index, photovoltaic penetration index and network structure complexity index; construct a scoring penalty index system and calculate index values ​​based on multi-source data, including voltage limit violation penalty, voltage fluctuation penalty and spatial imbalance penalty; The calculation unit is used to calculate the voltage quality score SV, which only considers the penalty item, based on the index value of the scoring penalty. The classification unit is used to establish an index-SV relationship model based on the index values ​​that reflect the factors affecting voltage quality, and to obtain the causal contribution decomposition results of each index and the corresponding transformer area classification based on the index-SV relationship model. Select a unit to establish a nonlinear multi-objective reactive power control optimization model. Based on the applicability score matrix R of the control scheme, select the optimal control scheme from the applicable control schemes corresponding to the classification results of the transformer area. The output unit is used to output the optimal control scheme.

Images