Distribution network graph model simplification method for reliability calculation of high-performance distribution network

Abstract

The invention relates to a distribution network graph model simplification method for reliability calculation of a high-performance power distribution network. The method specifically comprises the following steps: realizing key equipment screening by quantifying element reliability sensitivity; generating a hierarchical equivalent topology based on a graph theory community discovery algorithm; adaptive distributed power supply equivalence based on a dynamic aggregation mechanism; power distribution network element relayout and topology reconstruction: completing graph automatic conversion through a regularized drawing space and an intelligent connection rule; according to the method, a'dynamic aggregation-virtual node-automatic check 'three-in-one technical system is innovatively constructed, a ten thousand-level node graph model is compressed to a hundred-level virtual node, the reliability calculation time is compressed to a minute / second level from an hour / day level, and core indexes such as a key load power failure probability and an influence range are undistorted; the method has the advantages of being low in calculation complexity, ensuring integrity, achieving deep fusion, improving power grid toughness and guaranteeing power supply quality.

Description

Technical Field

[0001] This invention belongs to the field of power system reliability assessment and intelligent analysis technology, specifically involving a simplified distribution network diagram model for high-performance distribution network reliability calculation. Background Technology

[0002] In the context of new power system construction, the scale of distribution networks has exploded due to their "three high" characteristics (high proportion of renewable energy, power electronic equipment, and distributed resources). Traditional reliability calculation methods (such as the minimum path set method, fault traversal method, and Monte Carlo simulation) take hours or even days to compute in city-level distribution networks with tens of thousands of nodes, severely restricting online analysis and real-time early warning. Although existing topology simplification techniques have been applied in scenarios such as transmission network equivalence, accelerated electromagnetic transient simulation, or dimensionality reduction of state estimation, they have significant limitations: the simplification objectives are mostly focused on electrical characteristics (such as impedance equivalence and power flow consistency) rather than reliability characteristics (such as load point outage probability, fault impact range, and average system outage frequency / duration), leading to distortion of reliability indicators; simplification strategies rely on human experience. The lack of universal rules makes it difficult to adapt to the diverse structures and changing operating modes of distribution networks; the lack of support for dynamic updates prevents the adaptive adjustment of simplification strategies based on real-time operating status or fault scenarios; and the lack of co-design with high-performance computing architectures (such as parallel computing and GPU acceleration) results in a fragmented "simplification-computation" process, limiting overall efficiency improvement. These contradictions highlight the shortcomings of existing technologies in meeting the requirements of "high-performance + high-fidelity" reliability computing, necessitating the development of new topology simplification methods focused on maintaining reliability indicators, achieving an organic unity of topology compression, indicator fidelity, and computational acceleration. Therefore, it is essential to provide a distribution network diagram simplification method for high-performance distribution network reliability computing that features low computational complexity, ensures integrity, achieves deep integration, enhances grid resilience, and guarantees power supply quality. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a simplified distribution network diagram model for high-performance distribution network reliability calculation that has low computational complexity, ensures integrity, achieves deep integration, improves grid resilience, and guarantees power supply quality.

[0004] The objective of this invention is achieved as follows: a simplified distribution network diagram model for high-performance distribution network reliability calculation, the specific steps of which are as follows:

[0005] Step 1: Screen key equipment by quantifying the reliability sensitivity of components, and establish a three-level quantitative evaluation framework of equipment level, system level, and dynamic indicators;

[0006] Step 2: Generate a hierarchical equivalent topology based on the graph theory community detection algorithm, compressing the original graph model with tens of thousands of nodes to virtual nodes with hundreds of nodes;

[0007] Step 3: Based on the dynamic aggregation mechanism, adapt the distributed power source equivalence to a large number of distributed power sources and convert them into a small number of "virtual power sources";

[0008] Step 4: Achieve coordinated optimization of topology compression, index fidelity, and computational acceleration through distribution network component relocation and topology reconstruction.

[0009] Step 1 specifically includes the following steps:

[0010] Step 1.1: Construction of a three-level sensitivity index system. Based on reliability engineering theory, a quantitative evaluation framework is established from three levels: equipment, system, and dynamics.

[0011] Step 1.2: Multi-dimensional screening algorithm optimization, using an improved random forest algorithm for equipment screening;

[0012] Step 1.3: Identification of equivalent assimilation device type.

[0013] The equipment-level quantitative evaluation framework in step 1.1 specifically includes: establishing a three-dimensional indicator system of equipment failure rate, repair time, and load importance; calculating the marginal contribution of each device to the system reliability index through Monte Carlo simulation; and selecting the top 20% of key devices for retention. The system-level quantitative evaluation framework specifically includes SAIDI (System Average Outage Duration), MAIFI (System Average Failure Frequency), and ENSI (Enhanced Energy Information Reliability Index), which captures the degree of impact of devices on the overall reliability of the power grid from a global perspective.

[0014] The dynamic-level quantitative evaluation framework in step 1.1 specifically involves introducing dynamic indicators into a time-varying failure rate model. The dynamic indicator factor is: In the formula, λ(t) is the failure rate of the system at time t; α1 is the basic failure rate during the period of random failure; α2 and β2 are the shape and degradation parameter of the Weibull distribution, respectively; T i Let θ be the time point of the i-th maintenance; θ(t) is a dynamic index factor that reflects the amplification effect of environmental or load changes on the failure rate; E(t) and L(t) are the real-time environmental index and load index, respectively; E0 and L0 are the historical average values ​​of normal operation; γ and δ are sensitivity coefficients.

[0015] Step 2 specifically includes the following steps:

[0016] Step 2.1: Generate hierarchical equivalent topology based on graph theory community discovery algorithm. A hybrid algorithm framework is adopted, which combines the edge refinement capability of Girvan-Newman algorithm with the fast partitioning advantage of CNM algorithm to achieve a balance between modularity and computational efficiency.

[0017] Step 2.2: Topology Reconstruction Mathematical Modeling. By establishing the matrix representation theory of power grid topology, the complex power grid structure is transformed into a mathematical model.

[0018] Step 2.3: Hierarchical structure engineering practice, constructing a three-level hierarchical structure of "substation-feeder-distribution transformer".

[0019] In this invention, the hierarchical equivalent topology generation process uses an improved Louvain algorithm to divide communities, and determines the optimal community hierarchy structure by maximizing the modularity increment principle. Each community node is equivalent to a virtual node with aggregated electrical parameters.

[0020] Step 3 specifically includes the following steps:

[0021] Step 3.1: Generate "typical operating status" daily. At midnight every day, the system automatically pulls three types of measured data from the previous day: power distribution automation, electricity consumption information collection, and distributed power grid connection interface. These data are then overlaid with the 0-24h grid forecasts of photovoltaic irradiance, wind speed, and temperature issued by the meteorological bureau and spliced ​​into a "power-load" time series at 15-minute granularity.

[0022] Step 3.2: Assess the risk level of components in seconds. For each typical operating state generated, the system starts the "Fast N-1 Scan" engine. The scan scope includes four types of components: main feeder, large branches, distributed power grid-connected inverters, and key ring main units.

[0023] Step 3.3: Construct virtual nodes using switches as walls. After obtaining the "aggregable" green area, use the existing sectional switches, ring main units, and branch fuses on site as natural walls to merge all distributed photovoltaics, small wind turbines, energy storage, loads, and branch lines within the green area into a "virtual node".

[0024] Step 3.4: After the equivalent model is generated, immediately start the "verification engine": send the simplified diagram back into the reliability calculation core, perform N-1 fault analysis, output three major indicators - number of users experiencing power outages, power outage duration, and power shortage, and compare them item by item with the results of the original detailed model.

[0025] The “typical operating state” system in step 3.1 retains only key electrical quantities, including: active power and reactive power at the grid connection point, load level, main transformer load rate, and switch status.

[0026] The scanning scope in step 3.2 includes four types of components: main feeder, large branches, distributed power grid-connected inverters, and key ring main units.

[0027] In this invention, the dynamic aggregation mechanism is specifically implemented as follows: a distributed power output time series model is established, and similar power sources are aggregated into equivalent virtual power sources through K-means clustering, while retaining an error band of ±10% of the original power fluctuation range.

[0028] Step 4 specifically includes the following steps:

[0029] Step 4.1: Construct a regularized drawing space: First, logically divide the entire canvas into a uniform grid of X×Y. Each grid cell corresponds to a fixed step size. All plant icons can only be positioned at the grid intersections. Lines can only extend horizontally, vertically, or diagonally at 45° along the grid edges. Lines at arbitrary angles are prohibited.

[0030] Step 4.2: Construct intelligent connection rules: In the grid space, in order to achieve the optimal connection between primitives, eight optional directions are provided for each line step length, and real-time scoring is performed based on multiple association factors; through a weighted scoring mechanism, the direction with the highest score is selected as the forward path at each step, thereby guiding the line to approach the target in the shortest, smoothest, and most orderly way;

[0031] Step 4.3: Path Construction - Regular Connection Generation: After determining the optimal direction for each step, start from the edge of the starting primitive and gradually construct the path along the grid; after moving one grid, recalculate the direction score from the current position to the target and decide the next step, until the edge of the target primitive is reached;

[0032] Step 4.4: Eliminate intersection and inflection point redundancy: After all connections are completed, randomly select some line segments to try to translate or shorten them, and evaluate whether the number of intersections and inflection points has been reduced or the neatness of the line bundle has been improved; if the optimization is effective, retain the adjustment results; if there is no improvement, revert to the original state.

[0033] The weighted scoring mechanism in step 4.2 is as follows: Score(n) = w1*Distance(n,target) + w2*DirectionAlignment(n) + w3*Smoothness(n), where n is the candidate node or direction to be evaluated; Distance(n,target) is the distance from node n to the target; DirectionAlignment(n) is the angle between the current movement direction and the target direction; Smoothness(n) measures the smoothness of the path; and w1, w2, and w3 are weight coefficients.

[0034] In this invention, the topology reconstruction includes: a) a shortest path optimization unit based on Dijkstra's algorithm; b) a network loss optimization unit based on power flow calculation; and c) a topology self-repair unit based on cellular automata.

[0035] The beneficial effects of this invention are as follows: This invention provides a distribution network diagram simplification method for high-performance distribution network reliability calculation. Based on graph theory community discovery algorithms and dynamic aggregation mechanisms, this method is suitable for rapid reliability calculation and weak link analysis of large-scale distribution networks containing distributed generation sources. In use, this invention significantly reduces computational complexity while ensuring the integrity of key reliability information. It can also be deeply integrated with high-performance computing platforms, resolving the significant contradiction between diagram size, computational efficiency, and fidelity in existing distribution network reliability calculations. It breaks through traditional simplification approaches, meeting the real-time, accurate, and efficient reliability assessment needs of modern distribution networks. Solving this problem has significant theoretical and engineering value for improving grid resilience, ensuring power supply quality for users, and promoting the construction of new power systems. This invention has the advantages of low computational complexity, ensuring integrity, achieving deep integration, improving grid resilience, and ensuring power supply quality. Attached Figure Description

[0036] Figure 1 This is a flowchart of the present invention.

[0037] Figure 2 This is a schematic diagram of a single-line network diagram according to the present invention. Detailed Implementation

[0038] With the continuous advancement of the construction of new power systems, the distribution network, as a key link connecting the main grid and end users, is becoming increasingly complex in structure and flexible in operation, exhibiting the "three highs" characteristics of "high proportion of renewable energy access, high proportion of power electronic equipment grid connection, and high proportion of distributed resource interaction." Against this backdrop, the amount of topology calculations involved in the reliability assessment of the distribution network has increased significantly, affecting the calculation performance results. This paper addresses this issue by using graph simplification technology to summarize and simplify equipment nodes, and reconstructing the network structure graph based on the geographical map of the distribution network, greatly improving calculation efficiency and providing core support for reliability calculation and smart grid construction.

[0039] The present invention will be further described below with reference to the embodiments and / or accompanying drawings.

[0040] Example 1

[0041] like Figure 1-2 As shown, a method for simplifying distribution network diagrams for high-performance distribution network reliability calculations is presented. The specific steps of the method are as follows:

[0042] Step 1: Screen key equipment by quantifying the reliability sensitivity of components, and establish a three-level quantitative evaluation framework of equipment level, system level, and dynamic indicators;

[0043] In this embodiment, specifically, by quantifying the sensitivity of different equipment types to power grid reliability indicators, the accurate selection and equivalent replacement of key equipment can be achieved. The specific implementation path is divided into three dimensions:

[0044] ① A three-level sensitivity index system is constructed, based on reliability engineering theory, and a quantitative evaluation framework is established from three levels: equipment, system, and dynamics.

[0045] 1) Equipment-level indicators focus on the characteristics of a single piece of equipment, including failure frequency (failure rate λ), failure repair efficiency (repair time r), and aging degradation rate (aging coefficient β). These indicators directly reflect the reliability shortcomings of the equipment itself.

[0046] 2) System-level indicators focus on overall operational performance, such as SAIDI (System Average Outage Duration), MAIFI (System Average Failure Frequency), and ENSI (Enhanced Energy Index), which capture the degree of impact of equipment on the overall reliability of the power grid from a global perspective.

[0047] 3) Introducing dynamic indicators into the time-varying failure rate model: The dynamic indicator factor is: In the formula, λ(t) is the failure rate of the system at time t; α1 is the basic failure rate during the period of random failure; α2 and β2 are the shape and degradation parameter of the Weibull distribution, respectively; T i Let θ be the time point of the i-th maintenance; θ(t) is a dynamic index factor that reflects the amplification effect of environmental or load changes on the failure rate; E(t) and L(t) are real-time environmental indicators (such as temperature, humidity, and wind speed) and load indicators, respectively; E0 and L0 are the historical average values ​​of normal operation; γ and δ are sensitivity coefficients that can be set through data fitting or expert experience.

[0048] As a concrete example of an achievable implementation, such as the transformer failure rate increasing exponentially with rising temperature (driven by the temperature sensitivity coefficient γ), this dynamic model can more realistically reflect the real-time impact of environmental factors on equipment reliability.

[0049] ② Multi-dimensional screening algorithm optimization: An improved random forest algorithm is adopted for equipment screening. Where R represents the unbalance coefficient of the parent node (number of majority class samples / number of minority class samples); R1′ and R2′ represent the unbalance coefficients of the left and right child nodes after splitting, respectively; by incorporating topological location weights (such as the node location of the device in the power grid and connection complexity) through feature engineering, the algorithm not only considers the device's own parameters, but also incorporates the influence of its location in the power grid.

[0050] As a concrete example of an achievable implementation, an empirical analysis of 20,000 devices in a provincial power grid shows that the top 10% of highly sensitive devices (such as transformers) contribute 85% of the fluctuations in system reliability indicators, while low-sensitivity devices (such as capacitors) account for only 2% of the impact. This difference is addressed by determining the indicator weights using the entropy weight method (based on the objective distribution of data rather than subjective experience), thus avoiding the biases of traditional weighting methods.

[0051] ③ Identification of equivalent assimilation equipment types: Through the above sensitivity quantification and algorithm screening, it can be determined which equipment types have a minimal impact on system reliability (such as capacitors). These equipment types can be processed through equivalent assimilation (such as merging similar equipment and simplifying model parameters) without significantly affecting the overall reliability calculation results. Meanwhile, highly sensitive equipment (such as transformers) needs to be monitored and optimized. This "focus on the big and let go of the small" screening logic reduces the complexity of system reliability analysis and ensures precise control of key equipment, ultimately achieving equivalent substitution of equipment types and efficient resource allocation.

[0052] Step 2: Generate a hierarchical equivalent topology based on the graph theory community detection algorithm, compressing the original graph model with tens of thousands of nodes to virtual nodes with hundreds of nodes;

[0053] In this embodiment, the distribution network is topologically equivalent through three core steps, which effectively reduces the number of equipment nodes and the complexity of the network structure. The specific implementation path is as follows:

[0054] ① Based on the graph theory community discovery algorithm, a hierarchical equivalent topology is generated. By comparing the applicability of six mainstream graph theory algorithms in the power grid, the algorithm combination most suitable for the characteristics of the distribution network is selected. For example, the Louvain algorithm is suitable for handling large-scale power grid nodes due to its high modularity of 0.78; the multi-scale characteristics of the Infomap algorithm can adapt to the natural hierarchical structure of the distribution network; and the fast convergence characteristics of the LPA algorithm can meet the real-time computing requirements. An innovative hybrid algorithm framework is adopted, combining the edge refinement capability of the Girvan-Newman algorithm with the fast partitioning advantage of the CNM algorithm, achieving a balance between modularity and computational efficiency. This algorithm selection ensures that the topology equivalence not only conforms to the actual structure of the power grid, but also efficiently completes node aggregation.

[0055] ② Topology reconstruction mathematical modeling: By establishing the matrix representation theory of power grid topology (such as the grid base matrix, adjacency matrix, and Laplace matrix), the complex power grid structure is transformed into a mathematical model. In the empirical study of a 1000-node distribution network, the complexity of the reconstructed graph model is reduced by 65%, while the visualization effect meets the requirements of engineering aesthetics, ensuring that the topology equivalence is both simplified and intuitive.

[0056] As a specific implementation example, 1) a schematic diagram of a single-line network diagram is shown below. Figure 2As shown, the branch numbers are: e1 = (1, 2), e2 = (3, 2), e3 = (2, 4); the node numbers are: 1 = slack; 2, 3, 4 = PQ; the direction of the branches is always "smaller node → larger node".

[0057] 2) Bus-Incidence Matrix (A, |V|×|E|)

[0058] Numerical results (node ​​order 1…4, branch order e1, e2, e3),

[0059] 3) Adjacency matrix (unweighted, |V|×||V|)

[0060] E(i,j) = 1, if nodes i and j are directly connected.

[0061] 4) Laplace matrix (combined Laplace, L = DE)

[0062] First, calculate the degree matrix D = diag(deg(i)). then,

[0063] ③ Hierarchical structure engineering practice: Construct a three-level hierarchical structure of "substation-feeder-distribution transformer", divide multiple power supply areas through community discovery algorithm, with the equipment aggregation degree within each area reaching 90% and the number of connections between areas reduced by 40%; optimize the connection layout to reduce the number of cross points by 55% and form a visual topology that conforms to human factors engineering. This hierarchical structure not only reduces the complexity of the distribution network, but also improves the efficiency and maintainability of operation monitoring.

[0064] Through the synergistic effect of the above three steps, this module achieves an equivalent simplification of the distribution network topology. While retaining key operating characteristics, it significantly reduces the number of equipment nodes and the complexity of the network structure, providing strong support for the efficient management and optimized operation of the distribution network.

[0065] Step 3: Based on the dynamic aggregation mechanism, adapt the distributed power source equivalence to a large number of distributed power sources and convert them into a small number of "virtual power sources";

[0066] In this embodiment, the core objective is to transform a large number of distributed power sources into a small number of "virtual power sources" without losing critical reliability information, thereby compressing the topology scale and transforming reliability calculations from "finding a needle in a haystack" to "precise location." The specific process includes the following four steps:

[0067] ① The system generates a "typical operating state" on a rolling basis every day. At midnight every day, the system automatically pulls three types of measured data from the previous day: power distribution automation, electricity consumption information collection, and distributed power grid connection interface. These data are then superimposed with the 0-24h grid forecasts of photovoltaic irradiance, wind speed, and temperature issued by the meteorological bureau and spliced ​​into a "power-load" time series at 15-minute granularity.

[0068] As a concrete and feasible implementation, by selecting typical times such as the hour, the 96 points throughout the day are compressed into 20 to 50 "typical operating states". Each state represents a most likely combination of "weather-load-maintenance", such as "high solar power generation at noon + industrial load trough", "no wind in the evening + residential evening peak", "N-1 maintenance + cloudy day", etc.

[0069] For each typical scenario, the system retains only key electrical quantities: active power, reactive power, load level, main transformer load rate, and switch status at the grid connection point. The file size is less than 1% of the complete power flow, but it covers more than 95% of the probability space. If the weather changes suddenly or the maintenance ticket is rescheduled the next day, the scenario library generated the previous day will be automatically refreshed before the morning rush hour to ensure that it is "counted on the latest map".

[0070] ②Assess the risk level of components in seconds. For each typical operating state generated in the previous step, the system starts the "fast N-1 scan" engine: only considers the three-step logic of "component exit - load transfer - islanding", does not iteratively solve voltage and current, and focuses on answering "who is losing power".

[0071] The scanning scope includes four types of components: main feeders, large branches, distributed power grid-connected inverters, and key ring main units.

[0072] After scanning each element, two numbers are immediately output: the number of newly added power outage users and the newly added power shortage. If both numbers are lower than the preset threshold (e.g., 5 users, 10kWh), the element is marked as "aggregable"; otherwise, it is marked as "must be retained".

[0073] The entire process uses parallel computing. 200 components in a feeder can be scored within 3 seconds, generating a "red-green" risk map: the green area represents "even if the component disappears, it will not cause a big problem", providing a basis for the next step of packaging; the red area represents "must remain on the map as is", ensuring that key reliability information is not lost.

[0074] ③ After constructing virtual nodes using switches as walls and obtaining “aggregable” green areas, the system begins to “claim territory”: using the actual sectional switches, ring main units, and branch fuses on site as natural walls, all distributed photovoltaics, small wind turbines, energy storage, loads, and branch lines within the green area are merged into a “virtual node”.

[0075] ④ After the equivalent model is generated, the system immediately starts the "verification engine": the simplified diagram is sent back into the reliability calculation core, the N-1 fault analysis is run through completely, and three major indicators are output - number of users without power, power outage duration, and power shortage, and compared with the results of the original detailed model item by item; as long as the error of any indicator exceeds 1%, the system will automatically trigger "rollback": the virtual node where the error source is located is split into two smaller packages according to the "binary" principle, and the equivalent model is re-equivalent and re-verified; if it still exceeds the tolerance, the splitting continues, and convergence can be achieved in a maximum of three rounds.

[0076] The entire verification process requires no manual intervention, takes an average of 2 to 3 minutes, and the splitting depth does not exceed 3 levels, which ensures accuracy while avoiding "over-refinement" that could cause a rebound in the number of nodes. After the verification is passed, the equivalent model is labeled with a "timestamp + check code" and officially entered into the database, becoming the "sole legitimate source" for all reliability assessments on that day.

[0077] Step 4: Achieve coordinated optimization of topology compression, index fidelity, and computational acceleration through distribution network component relocation and topology reconstruction.

[0078] In this embodiment, specifically: based on the simplified diagram data of the distribution network and distributed power sources obtained in steps 2-3, a complete automated link from the "original topology" to the "regular and aesthetically pleasing diagram" is constructed through the following four steps, which ensures both diagram quality and computational efficiency, as detailed below:

[0079] ① Constructing a regularized drawing space: In order to transform the originally freely laid-out power grid topology into a regular and controllable drawing space, the entire canvas is first logically divided into a uniform X×Y grid. Each grid cell corresponds to a fixed step size (such as 12 pixels), forming an invisible "chessboard". All plant icons can only be positioned at the grid intersections, and lines can only extend horizontally, vertically, or diagonally at 45° along the grid edges, prohibiting arbitrary angles. This gridding method not only unifies the layout basis of the primitives, but also provides a discrete and computable spatial structure for subsequent path planning. The step size can be dynamically adjusted according to the primitive size, canvas resolution, and computing resources, which ensures both graphic accuracy and algorithm complexity, laying the foundation for subsequent automatic connection, obstacle avoidance, alignment, and other operations.

[0080] ② Construct intelligent connection rules. In the grid space, to achieve optimal connections between primitives, the system provides eight selectable directions (up, down, left, right, and four diagonal directions) for each step length and scores them in real time based on multiple correlation factors. The main considerations include: whether to move towards the target primitive (shorten the path), whether to continue the previous direction (reduce inflection points), whether the path ahead is open (avoid overlap), and whether to maintain the same direction as lines of the same voltage level (form a bundle). Through a weighted scoring mechanism, the system selects the direction with the highest score as the path of progress at each step, thereby guiding the line to approach the target in the shortest, smoothest, and neatest way. This rule-driven direction selection strategy replaces the inefficiency and uncertainty of traditional random search, ensuring that the connection process is both efficient and conforms to visual and topological logic.

[0081] As a concrete implementation method, the weighted scoring mechanism is as follows: Score(n) = w1*Distance(n,target) + w2*DirectionAlignment(n) + w3*Smoothness(n), where n is the candidate node or direction to be evaluated; Distance(n,target) is the distance from node n to the target, used to measure "shortest"; DirectionAlignment(n) is the angle (or cosine similarity) between the current movement direction and the target direction, used to measure "smoothest"; Smoothness(n) measures the smoothness of the path, such as the angle of deviation from the previous direction, the smaller the deviation, the higher the score; w1, w2, and w3 are weight coefficients used to adjust the importance of each factor, satisfying w1 + w2 + w3 = 1.

[0082] ③ Path Construction - Achieving Regularized Connection Generation: After determining the optimal direction for each step, the system starts from the edge of the starting point and gradually constructs the path along the grid. Each time the system moves one grid cell, it recalculates the direction score from the current position to the target and determines the next step, until the target point is reached. The entire path consists of a series of continuous grid points, naturally possessing a regular shape with horizontal and vertical lines or 45° diagonal lines, avoiding irregular jagged edges and redundant turns. If an obstacle is encountered at a certain step (such as other lines or elements), the system allows one "backtrack + lane change" operation, simulating checkers logic, ensuring the path can eventually be successfully connected. This process is based entirely on local optimal decisions while also considering the global path trend.

[0083] ④ Eliminate redundancy in intersections and inflection points. The initial connection rules may still have problems such as too many intersections, inflection points, or messy wire bundles in densely populated areas. To address this, after all connections are completed, randomly select some line segments to try to shift or shorten them, and evaluate whether this reduces the number of intersections and inflection points or improves the neatness of the wire bundles. If the optimization is effective, retain the adjustment results; if there is no improvement, revert to the original state.

[0084] This process is repeated dozens to hundreds of times, gradually approaching the visually optimal solution. Since each adjustment only involves local path changes, the computational load is small and it is usually completed within 1 second. The final output topology map is not only logically correct, but also has a uniform distribution of lines, clear hierarchy, and minimal intersections, satisfying the requirements of large-screen display, drawing printing, and on-site inspection. These four steps constitute a complete automated link from "original topology" to "regular and beautiful graphics", which ensures both graphic quality and computational efficiency.

[0085] In summary, the method of this invention achieves the following improvements in computational performance: 1) Computational complexity is reduced from 0 (n... 2 1) The time required for reliability calculation is reduced to 0(nlogn); 2) The time required for reliability calculation is reduced to 1 / 50 of the original method; 3) The calculation error for the probability of power outage of critical loads is ≤2%, and the calculation error for the scope of impact is ≤5%.

[0086] This invention presents a simplified distribution network diagram model for high-performance distribution network reliability calculation. In its application, this invention addresses the core contradictions in distribution network reliability assessment, namely, the explosion in diagram model size, low computational efficiency, and insufficient reliability fidelity. It innovatively constructs a three-in-one technical system of "dynamic aggregation - virtual nodes - automatic verification": through daily rolling generation of typical operating states, second-level N-1 risk scanning, automatic verification with an error ≤1%, and a dynamic response mechanism for sudden weather / maintenance changes, it compresses the diagram model of tens of thousands of nodes to hundreds of virtual nodes, reducing reliability calculation time from hours / days to minutes / seconds. Key indicators such as the probability of outages of critical loads and the scope of impact remain undistorted, effectively supporting the construction of minute-level reliability assessment capabilities under new power systems and empowering proactive emergency response. This invention, through its focus on repair, resilience enhancement, and source-grid-load-storage synergistic optimization, holds significant theoretical and engineering value for promoting the construction of digital and intelligent distribution networks under the "dual carbon" objective. Breaking through the limitations of traditional topology simplification that focuses on electrical characteristics, this invention pioneers a new simplification paradigm oriented towards maintaining reliability indicators, forming a collaborative theoretical system of "topology compression - indicator fidelity - computational acceleration." This system supports the construction of minute-level reliability assessment capabilities under the new power system, empowering proactive repair, resilience enhancement, and source-grid-load-storage synergistic optimization, thus promoting the construction of digital and intelligent distribution networks under the "dual carbon" objective. It possesses significant theoretical innovation value, engineering application significance, and industry-leading role. Furthermore, this invention offers advantages such as low computational complexity, ensuring integrity, achieving deep integration, enhancing grid resilience, and guaranteeing power supply quality.

[0087] Example 2

[0088] like Figure 1-2 As shown, a simplified distribution network diagram modeling method for high-performance distribution network reliability calculation is presented, with the following specific steps:

[0089] Step 1: Screening key equipment by quantifying component reliability sensitivity: Accurately identify key equipment using a three-level sensitivity index system and an improved random forest algorithm;

[0090] Step 2: Generate hierarchical equivalent topology based on graph theory community discovery algorithm: Use hybrid algorithm framework (such as Louvain, Infomap) to realize hierarchical aggregation of nodes, and combine matrix representation theory to reduce the number of device nodes;

[0091] Step 3: Adaptive Distributed Power Equivalence Based on Dynamic Aggregation Mechanism: Achieve equivalent aggregation of virtual nodes of distributed power sources through rolling generation of "typical operating states" and second-level risk assessment;

[0092] Step 4: Distribution network component relocation and topology reconstruction: Automatic graphical conversion is completed through a regularized drawing space and intelligent connection rules.

[0093] In this embodiment, as Figure 1 As shown, the method of the present invention forms a closed-loop simplification process through four vertical swimlanes: "key equipment screening → distribution network equivalent topology → distributed power source equivalence → component relocation and topology reconstruction". Each swimlane can be preprocessed in parallel or connected serially according to data dependencies, and finally outputs a set of lightweight graph models with significantly reduced node size and controllable reliability error.

[0094] Step 1: Key Equipment for Selecting Swimming Lanes

[0095] Responsibility: Identify the few devices that have the most significant impact on system reliability from the original model, form a "must-retain" whitelist, and provide simplified boundary constraints for subsequent three-lane tests.

[0096] Inputs: Power grid CIM / SVG model, SCADA / AMI operation data, historical fault records, and environmental meteorological data.

[0097] Outputs: List of critical equipment types (approximately 10% of the total), comprehensive sensitivity ranking file, and SHAP explanation report.

[0098] 1) Input power grid model

[0099] ① Parse the CIM / XML topology and generate a node-branch association table;

[0100] ② Verify electrical connectivity and eliminate isolated data;

[0101] ③ Establish a three-level hash index of "device-ID-parameters" to support millisecond-level queries.

[0102] 2) Running data

[0103] ① The time window is uniformly set to 1 minute resolution, and missing values ​​are interpolated using bidirectional linear interpolation;

[0104] ② Detect abnormal changes in the four measurements of active power, reactive power, voltage, and current (using the 3σ+Grubbs joint criterion);

[0105] ③ Output a "clean" multidimensional time series for 1.3 to calculate the expected load level and fluctuation variance.

[0106] 3) Equipment reliability calculation parameters

[0107] ① Failure rate λ: Weibull regression is used, taking into account operating years, load rate, temperature, and lightning strike density;

[0108] ② Repair time r: Log-normal distribution fitting to distinguish between on-site repair and waiting for spare parts;

[0109] ③ Planned maintenance rate: Automatically calculated based on the monthly maintenance plan;

[0110] ④ Create a "device-λ-r" parameter library and write it to the Redis cache for real-time use by the subsequent matrix model.

[0111] 4) Obtain system reliability indicators

[0112] ①Based on all component parameters, a minimum path set-fault propagation two-layer algorithm is used to quickly calculate the baselines of SAIFI, SAIDI, and ENS;

[0113] System reliability based on minimum path sets: Let all minimum path sets be P = {P1, P2, P3, P4, ..., P...} m}, then the system reliability is: Using the principle of inclusion-exclusion: in, Correction after fault propagation: If we consider the failure of fault propagation along the path, we define path P. i The effective working probability is: If there is an external disturbance, the injection rate ε j Along the path:

[0114] ② Write the index into shared memory as the benchmark value for “simplification error limit”. Any subsequent simplification must not cause the SAIDI deviation to be greater than 5%.

[0115] 5) Calculate the Sobol index and sensitivity.

[0116] ① Establish Sobol sequence sampling for four sets of input variables: λ, r, load level, and DG output (N = 4096);

[0117] ② Run the alternative model (XGBoost) to predict reliability indicators and calculate the first-order and total effect indices;

[0118] ③ Output the "Component-Sensitivity" ranking to form the first coarse screening.

[0119] 6) Construct 2D feature output SHAP score

[0120] ① 2D characteristics = (Electrical dielectric constant, load importance);

[0121] ② Train the SHAP interpretation model to obtain the marginal contribution of each element to SAIDI;

[0122] ③ Mark components with an absolute SHAP value ≥ 0.01 as "high interpretation".

[0123] 7) Generate a comprehensive sensitivity ranking

[0124] ①Weighted fusion Sobol total effect index (weight 0.6) and SHAP absolute value (weight 0.4);

[0125] ② Sort by score in descending order, generate a CSV list, write it to the Kafka message queue, and provide real-time subscription for the equivalent topology swimlane of the distribution network.

[0126] 8) The top 10% are considered key equipment types.

[0127] ① Group the equipment by type (transformer, circuit breaker, line segment), and then select the top 10% from each group to prevent a monopoly of a particular type of equipment;

[0128] ② The final list of critical devices is written to PostgreSQL and flag=1 is set. Devices with flag=1 are forcibly retained in subsequent three-lane swimming tests.

[0129] Step 2: Equivalent Topology Swimlane of Distribution Network

[0130] Responsibility: Under the boundary constraints of "critical equipment", perform matrix-level equivalent merging of non-critical areas to generate a model with the smallest scale but conserved electrical characteristics.

[0131] Input: Whitelist of critical equipment, original CIM, and merge criteria configuration table.

[0132] Outputs: Simplified CIM / SVG, node-branch mapping table, and equivalent parameter library.

[0133] 1) Design of a multi-collection merging hybrid algorithm

[0134] ① "Multiple loads" refers to multiple voltage levels, multiple regions, and multiple feeder lines;

[0135] ② Adopt a hybrid strategy: Norton equivalent is used at the end of the feeder, star-delta transformation is used for the bus group, and Thevenin equivalent is used for the ring network;

[0136] ③ Generate an "equivalent error budget" curve before merging to ensure that the SAIDI increment is <0.5% at each step.

[0137] 2) Constructing matrix modeling

[0138] ① The node admittance Ybus, reliability connection matrix R, and load transfer matrix T are constructed synchronously;

[0139] ② CSR compressed storage is used, reducing memory usage by 60%;

[0140] ③ Add a "lock mark" to the row / column containing critical equipment to prevent it from being merged.

[0141] 3) Formulate merger criteria

[0142] ① Electrical distance < 0.01 pu;

[0143] ② Load capacity ratio < 5%;

[0144] ③ No distributed power injection;

[0145] ④ Fault consequence domain overlap > 80%;

[0146] A merger is allowed only if all four conditions are met.

[0147] 4) Obtain distribution network diagram data

[0148] ① Subscribe to the "critical devices" topic in Kafka in real time and dynamically refresh the lock list;

[0149] ② After each round of merging, update the "node old-ID→new-ID" mapping table for querying the equivalent swimlane of distributed power sources.

[0150] 5) Output the simplified distribution network diagram model

[0151] ① Generate CIM / XML conforming to IEC 61970-552 standard;

[0152] ② Automatically generate SVG, using colors to distinguish between "existing key equipment", "equivalent equipment", and "virtual nodes";

[0153] ③ Output JSON equivalent parameters, including equivalent impedance, equivalent load, and equivalent failure rate, for direct use by the reliability assessment module.

[0154] Step 3: Equivalent swimlane for distributed power sources

[0155] Responsibilities: To aggregate massive amounts of small-capacity DGs into "virtual nodes" to reduce model dimensionality while ensuring the spatiotemporal relevance of output and the accuracy of reliability contributions.

[0156] Input: DG measurement timing sequence, typical scenario library, component reliability deviation threshold.

[0157] Output: Virtual node model, aggregation parameters, and tagged database file.

[0158] 1) Data splicing of distributed power source measurement

[0159] ① Read the active power, reactive power, and voltage at the grid connection point from the AMI / SCADA system;

[0160] ② Missing data were filled using "similar dates + KNN";

[0161] ③ A unified resolution of 15 minutes is used to form a three-dimensional array of “DG-ID-time”.

[0162] 2) Generate typical running states

[0163] ① K-shape clustering was used to cluster the annual power output curves into 8 categories;

[0164] ② Extract "expected output + covariance matrix" for each category to form a typical day;

[0165] ③ Output typical scenario CSV files for dynamic updates to the scenario library.

[0166] 3) Scene library dynamically refreshed

[0167] ① The latest measurements are automatically retrieved and re-clustered monthly;

[0168] ② If the distance between the center of gravity of the new scene and the old scene is greater than 0.05 pu, the version number will be incremented by 1, and a message will be pushed to the component layout swimlane, prompting a re-evaluation of the equivalent range.

[0169] 4) Assess component reliability calculation deviations

[0170] ① For each typical scenario, calculate the SAIDI difference between the "complete DG model" and the "aggregated virtual node model";

[0171] ②If the difference is greater than 1%, then reduce the aggregation range and increase the number of virtual nodes;

[0172] ③ Create a "deviation-range" lookup table for use in 3.6.

[0173] 5) Define the virtual area

[0174] ① Use Voronoi + electrical distance dual constraints to divide the service domain of virtual nodes;

[0175] ② Ensure that the DG capacity aggregated by each virtual node is ≤2MVA and the electrical distance is ≤0.02pu;

[0176] ③ Output the mapping table of "virtual node-ID→DG group".

[0177] 6) Generate distributed power virtual nodes

[0178] ① Aggregation parameter: Equivalent capacity = ∑P i The equivalent λ = weighted average, and the equivalent r = weighted average;

[0179] ② Output curve: The Copula function is used to maintain spatiotemporal correlation;

[0180] ③ Write the standard CIM class "SynchronousMachineEquivalent" to read the equivalent topology swimlane of the power distribution network.

[0181] 7) Distributed power source model tagging and database entry

[0182] ①The tags include: equivalent capacity, equivalent λ, equivalent r, typical scenario ID, and version number;

[0183] ② Write to MongoDB, supporting geographic indexes and version backtracking;

[0184] ③ Provide RESTful interfaces to facilitate the reuse of multiple services such as planning, scheduling, and transactions.

[0185] Step 4: Component Relocation and Topology Reconfiguration Swimlanes

[0186] 1) Grid division based on equipment size

[0187] ① GeoHash+LZW compression is used to reduce the dimensionality of 10m-level coordinates;

[0188] ② Dynamic calculation of grid side length: 200m in urban areas, 500m in suburban areas, and 1000m in rural areas;

[0189] ③ Output the hash table of “mesh-ID→device group”.

[0190] 2) Locate power grid elements to grid points

[0191] ①The graphic elements include: poles and towers, switches, ring main units, and junction boxes;

[0192] ② Nearest neighbor matching is used to map latitude and longitude to the center point of the grid;

[0193] ③ Generate a “Element-Grid” visualization layer for manual verification.

[0194] 3) Set line extension rules

[0195] ① Rule 1: Do not cross the grid containing critical equipment;

[0196] ② Rule 2: Prioritize extending along road vector data to reduce turning angles;

[0197] ③ Rule 3: The length of a single line segment must be ≤500m; otherwise, add a virtual node.

[0198] 4) Determine the direction selection factor

[0199] ①Factor = α × Load moment + β × Investment cost + γ × Reliability sensitivity;

[0200] ②α+β+γ=1, default 0.5 / 0.3 / 0.2, adjustable via user interface;

[0201] ③ The A* algorithm is used to search for the optimal direction, and the heuristic function includes the factors mentioned above.

[0202] 5) Calculate the optimal direction for each step.

[0203] ① Generate an eight-neighbor extension tree with the current node as the root;

[0204] ② For each candidate direction, call the fast reliability assessment kernel to provide the SAIDI increment in seconds; ③ Select the direction with the smallest increment that satisfies rule 4.4 as the optimal one.

[0205] 6) Dynamically construct connection paths

[0206] ① The path table stores "starting grid-ID → ending grid-ID → line segment parameters";

[0207] ② Supports dynamic updates; the path is automatically recalculated when the virtual node position changes.

[0208] ③ Output GeoJSON for direct rendering by the GIS platform.

[0209] 7) Obstacle detection and rollback processing

[0210] ① Obstacles include: critical equipment being mistakenly deleted, SAIDI increment > 5%, and virtual node electrical distance exceeding the limit;

[0211] ② Once triggered, initiate a three-layer rollback:

[0212] a. Single-step rollback – Undo the most recent merge;

[0213] b. Local rollback – rollback to the previous stable sub-region;

[0214] c. Global rollback – Restore to the initial topology and reset a stricter merging threshold.

[0215] ③ Rollback logs are written to Elasticsearch to support fault recovery.

[0216] 8) Output the simplified distribution network topology diagram

[0217] ① Format: SVG + JSON dual output;

[0218] ②SVG uses different line widths to distinguish between the original key segments, equivalent segments, and virtual node connections;

[0219] ③The JSON contains node coordinates, branch parameters, equivalent identifiers, and version numbers, which can be reused by multiple services such as scheduling, planning, and reliability assessment.

[0220] In summary, this invention constructs a three-in-one technical system of "dynamic aggregation-virtual nodes-automatic verification", which compresses tens of thousands of node graph models to hundreds of virtual nodes, significantly reduces reliability calculation time, and transmits core indicators without distortion. It supports the construction of minute-level reliability assessment capabilities for new power systems and promotes the development of digital smart distribution networks.

[0221] This invention has the following advantages: ① Dynamic aggregation technology: Core function: Real-time compression of large-scale (tens of thousands) physical nodes into a small number (hundreds) of virtual nodes; Beneficial effect: This is the core of achieving a leap in computing efficiency; It directly solves the computing bottleneck caused by the large number of nodes in traditional methods, enabling reliability assessment to leap from "hours" to "minutes", while ensuring the accuracy (without distortion) of core indicators (such as reliability, voltage, power flow), providing the possibility for real-time or near-real-time decision-making.

[0222] ② Virtual Node Equivalent Modeling: Core Function: Replacing the aggregated physical node group with high-fidelity "virtual nodes" to equivalently reflect its electrical characteristics and reliability behavior; Beneficial Effects: This is the basis for the establishment of "dynamic aggregation"; It ensures that while significantly reducing topology complexity, key system information is not lost, resolving the contradiction between "compression" and "accuracy", and making subsequent analysis results based on the simplified model still reliable.

[0223] ③ Automatic verification mechanism: Core function: After aggregation and equivalence processing, it automatically verifies the consistency between the calculation results of the simplified model (virtual node topology) and the original detailed model; Beneficial effect: This is the guarantee of the "reliability" and "practicality" of the entire technical system; It ensures in a closed loop that no unacceptable errors are introduced in the aggregation and equivalence process, and achieves "lossless transmission of core indicators", allowing operators to trust and rely on this high-speed but simplified model for critical decision-making.

[0224] This invention presents a simplified distribution network diagram model for high-performance distribution network reliability calculation. In practice, this invention achieves a closed-loop simplification process involving "precise identification of key equipment—topology matrix-level compression—distributed power source aggregation—geographic topology secondary optimization," supporting real-time reliability assessment of distribution networks with a high proportion of distributed power sources. It has significant theoretical and engineering value for improving grid resilience, ensuring power supply quality for users, and promoting the construction of new power systems. This invention boasts advantages such as low computational complexity, ensuring completeness, achieving deep integration, improving grid resilience, and guaranteeing power supply quality.

Claims

1. A method for simplifying distribution network diagrams for high-performance distribution network reliability calculation, characterized in that: The specific steps of the method are as follows: Step 1: Screen key equipment by quantifying the reliability sensitivity of components, and establish a three-level quantitative evaluation framework of equipment level, system level, and dynamic indicators; Step 2: Generate a hierarchical equivalent topology based on the graph theory community detection algorithm, compressing the original graph model with tens of thousands of nodes to virtual nodes with hundreds of nodes; Step 3: Based on the dynamic aggregation mechanism, adapt the distributed power source equivalence to a large number of distributed power sources and convert them into a small number of "virtual power sources"; Step 4: Achieve coordinated optimization of topology compression, index fidelity, and computational acceleration through distribution network component relocation and topology reconstruction.

2. The distribution network diagram simplification method for high-performance distribution network reliability calculation as described in claim 1, characterized in that: Step 1 specifically includes the following steps: Step 1.1: Construction of a three-level sensitivity index system. Based on reliability engineering theory, a quantitative evaluation framework is established from three levels: equipment, system, and dynamics. Step 1.2: Multi-dimensional screening algorithm optimization, using an improved random forest algorithm for equipment screening; Step 1.3: Identification of equivalent assimilation device type.

3. The method for simplifying distribution network diagrams for high-performance distribution network reliability calculation as described in claim 1, characterized in that: The equipment-level quantitative evaluation framework in step 1.1 specifically includes: establishing a three-dimensional indicator system of equipment failure rate, repair time, and load importance; calculating the marginal contribution of each device to the system reliability index through Monte Carlo simulation; and selecting the top 20% of key devices for retention. The system-level quantitative evaluation framework specifically includes SAIDI, MAIFI, and ENSI, which capture the degree of impact of devices on the overall reliability of the power grid from a global perspective.

4. The distribution network diagram simplification method for high-performance distribution network reliability calculation as described in claim 3, characterized in that: The dynamic-level quantitative evaluation framework in step 1.1 specifically involves introducing dynamic indicators into a time-varying failure rate model. The dynamic indicator factor is: In the formula, λ(t) is the failure rate of the system at time t; α1 is the basic failure rate during the period of random failure; α2 and β2 are the shape and degradation parameter of the Weibull distribution, respectively; T i Let θ be the time point of the i-th maintenance; θ(t) is a dynamic index factor that reflects the amplification effect of environmental or load changes on the failure rate; E(t) and L(t) are the real-time environmental index and load index, respectively; E0 and L0 are the historical average values ​​of normal operation; γ and δ are sensitivity coefficients.

5. The method for simplifying distribution network diagrams for high-performance distribution network reliability calculation as described in claim 1, characterized in that: Step 2 specifically includes the following steps: Step 2.1: Generate hierarchical equivalent topology based on graph theory community discovery algorithm. A hybrid algorithm framework is adopted, which combines the edge refinement capability of Girvan-Newman algorithm with the fast partitioning advantage of CNM algorithm to achieve a balance between modularity and computational efficiency. Step 2.2: Topology Reconstruction Mathematical Modeling. By establishing the matrix representation theory of power grid topology, the complex power grid structure is transformed into a mathematical model. Step 2.3: Hierarchical structure engineering practice, constructing a three-level hierarchical structure of "substation-feeder-distribution transformer".

6. The distribution network diagram simplification method for high-performance distribution network reliability calculation as described in claim 1, characterized in that: Step 3 specifically includes the following steps: Step 3.1: Generate "typical operating status" daily. At midnight every day, the system automatically pulls three types of measured data from the previous day: power distribution automation, electricity consumption information collection, and distributed power grid connection interface. These data are then overlaid with the 0-24h grid forecasts of photovoltaic irradiance, wind speed, and temperature issued by the meteorological bureau and spliced ​​into a "power-load" time series at 15-minute granularity. Step 3.2: Assess the risk level of components in seconds. For each typical operating state generated, the system starts the "Fast N-1 Scan" engine. The scan scope includes four types of components: main feeder, large branches, distributed power grid-connected inverters, and key ring main units. Step 3.3: Construct virtual nodes using switches as walls. After obtaining the "aggregable" green area, use the existing sectional switches, ring main units, and branch fuses on site as natural walls to merge all distributed photovoltaics, small wind turbines, energy storage, loads, and branch lines within the green area into a "virtual node". Step 3.4: After the equivalent model is generated, immediately start the "verification engine": send the simplified diagram back into the reliability calculation core, perform N-1 fault analysis, output three major indicators - number of users experiencing power outages, power outage duration, and power shortage, and compare them item by item with the results of the original detailed model.

7. The distribution network diagram simplification method for high-performance distribution network reliability calculation as described in claim 6, characterized in that: The "typical operating state" system in step 3.1 retains only key electrical quantities, including: active power and reactive power at the grid connection point, load level, main transformer load rate, and switch status.

8. The method for simplifying distribution network diagrams for high-performance distribution network reliability calculation as described in claim 6, characterized in that: The scanning scope in step 3.2 includes four types of components: main feeder, large branches, distributed power grid-connected inverters, and key ring main units.

9. The method for simplifying distribution network diagrams for high-performance distribution network reliability calculation as described in claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4.1: Construct a regularized drawing space: First, logically divide the entire canvas into a uniform grid of X×Y. Each grid cell corresponds to a fixed step size. All plant icons can only be positioned at the grid intersections. Lines can only extend horizontally, vertically, or diagonally at 45° along the grid edge. Lines at arbitrary angles are prohibited. Step 4.2: Construct intelligent connection rules: In the grid space, in order to achieve the optimal connection between primitives, eight optional directions are provided for each line step length, and real-time scoring is performed based on multiple association factors; through a weighted scoring mechanism, the direction with the highest score is selected as the forward path at each step, thereby guiding the line to approach the target in the shortest, smoothest, and most orderly way; Step 4.3: Path Construction - Regularized Connection Generation: After determining the optimal direction for each step, start from the edge of the starting primitive and gradually construct the path along the grid; after moving one grid, recalculate the direction score from the current position to the target and decide the next step, until the edge of the target primitive is reached; Step 4.4: Eliminate intersection and inflection point redundancy: After all connections are completed, randomly select some line segments to try to translate or shorten them, and evaluate whether the number of intersections and inflection points has been reduced or the neatness of the line bundle has been improved; if the optimization is effective, retain the adjustment results; if there is no improvement, revert to the original state.

10. The method for simplifying distribution network diagrams for high-performance distribution network reliability calculation as described in claim 9, characterized in that: The weighted scoring mechanism in step 4.2 is as follows: Score(n) = w1*Distance(n,target) + w2*DirectionAlignment(n) + w3*Smoothness(n), where n is the candidate node or direction to be evaluated; Distance(n,target) is the distance from node n to the target; DirectionAlignment(n) is the angle between the current movement direction and the target direction; Smoothness(n) measures the smoothness of the path; and w1, w2, and w3 are weight coefficients.

Images