Distribution network fault simulation method based on load flow calculation

Abstract

The invention relates to the technical field of distribution network fault simulation, in particular to a distribution network fault simulation method based on load flow calculation, comprising the following steps: first acquisition: acquiring line information of a target distribution network; first construction: constructing a topological data graph of the power distribution network according to the line information, and correcting and simplifying the topological data graph through measurement data; first calculation: carrying out load flow calculation by adopting a forward-backward sweep algorithm based on the corrected and simplified topological data graph to obtain voltage and current distribution of each node; first simulation: simulating fault conditions in the power distribution network according to the voltage and current distribution, including setting fault points, fault types and fault parameters, and forming fault data; and outputting: observing and recording the voltage and current change conditions of each node in the fault simulation process and the overall response of the power distribution network, and outputting a change record report. The method has the effect that the simulation result is more accurate.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network fault simulation, and in particular to a distribution network fault simulation method based on power flow calculation. Background Art

[0002] With the continuous expansion of the distribution network and the high proportion of distributed energy access, the topology of the distribution network is becoming increasingly complex and the operation mode is showing diversified characteristics. In this context, distribution network fault simulation technology has become an important research method to ensure the safe operation of the power grid and optimize the relay protection strategy.

[0003] In the existing technology, some methods use power flow calculation based on the Newton-Raphson method to simulate faults. However, since the distribution network usually has a radial structure and the line impedance ratio (R / X) is high, such algorithms are prone to convergence problems, resulting in low simulation efficiency. In addition, traditional distribution network fault simulation methods are mostly based on simplified circuit models or static power flow calculations, which are difficult to accurately reflect the dynamic response characteristics of the actual power grid. In particular, there are obvious limitations when dealing with complex topologies and multiple types of faults, resulting in inaccurate simulation effects. Summary of the invention

[0004] In order to make the simulation results more accurate, the present application provides a distribution network fault simulation method based on power flow calculation.

[0005] The present application provides a distribution network fault simulation method based on power flow calculation, which adopts the following technical solutions: A distribution network fault simulation method based on power flow calculation includes the following steps: First acquisition: acquiring line information of the target distribution network, wherein the line information includes a line ID, a rated voltage, and a topological structure of each line; First construction: constructing a topological data map of the distribution network according to the line information, and correcting and simplifying the topological data map through the measurement data; First calculation: Based on the corrected and simplified topological data graph, the forward-backward algorithm is used to perform power flow calculation to obtain the voltage and current distribution of each node; First simulation: simulating the fault situation in the distribution network according to the voltage and current distribution, including setting the fault point, fault type and fault parameters to form fault data; Output: Observe and record the voltage and current changes of each node during the fault simulation, as well as the overall response of the distribution network, and output the change record report.

[0006] By adopting the above technical solution and obtaining detailed line information, an accurate distribution network topology data map can be constructed; by using the forward-backward algorithm for power flow calculation, the accurate values ​​of the voltage and current distribution of each node can be obtained, providing accurate basic data for subsequent fault simulation; this method can simulate a variety of fault conditions, including setting different fault points, fault types and fault parameters, so as to meet the diverse needs of distribution network fault analysis; by observing and recording the voltage and current changes of each node during the fault simulation process, as well as the overall response of the distribution network, a strong basis can be provided for the operation analysis, fault diagnosis and decision support of the power system, and this method can also be used to evaluate the stability and reliability of the distribution network.

[0007] This method uses computer simulation technology to quickly simulate the fault conditions of the distribution network, greatly improving the efficiency and accuracy of fault analysis; compared with traditional fault analysis methods, this method can save a lot of manpower, material resources and time costs.

[0008] Optionally, before the first simulation step, a failure risk assessment step is also included: Fault risk assessment: including: construction of fault causal network steps, probability distribution steps and simulation optimization steps; Constructing a fault causal network: defining Bayesian network node variables, including equipment state variables, environmental variables, and fault type variables, and establishing conditional probability relationships between nodes based on historical fault recording data and operation and maintenance records; Generate fault probability distribution: input real-time meteorological data and equipment monitoring data into the fault causal network; calculate the fault probability of each line section ,in is the probability of occurrence of the kth type of failure, is the kth type of fault, for environmental / equipment evidence; Simulation optimization: Generate candidate fault sets by descending order of fault probability ,in represents the i-th line section in the distribution network, represents the possible fault type of the i-th line section, is the probability of the jth fault occurring in the ith line section, is the probability threshold, and ; Will The input parameters are passed to the first simulation step to simulate the fault in combination with the voltage and current distribution.

[0009] By adopting the above technical solution, a fault causal network is constructed using a Bayesian network. By defining node variables and establishing conditional probability relationships, the complex relationship between equipment status, environmental factors and fault types can be more accurately reflected, and line sections and fault types with higher fault risks in the distribution network can be identified; this allows fault simulation to focus more on these high-risk areas, thereby improving the pertinence and efficiency of the simulation.

[0010] The fault risk assessment step combines real-time meteorological data, equipment monitoring data, and historical fault data to provide a more comprehensive and accurate information basis for fault simulation; this helps to more realistically reflect the fault behavior of the distribution network under actual conditions and improve the accuracy of the simulation.

[0011] The candidate fault set generated by the simulation optimization step provides operation and maintenance personnel with a priority reference for fault prevention and handling. Combined with the distribution of voltage and current for fault simulation, the impact of different faults on the distribution network can be further evaluated, providing a scientific basis for formulating fault response strategies.

[0012] Optionally, in the step of constructing the fault causal network, the establishment of the conditional probability relationship adopts an EM algorithm and a domain knowledge fusion method, specifically including: For nodes with complete historical data, the CPT table is directly calculated by maximum likelihood estimation; For nodes with missing data, the Dirichlet prior distribution constrained by expert experience is introduced, and the probability parameters are completed through Bayesian reasoning.

[0013] By adopting the above technical solution, for nodes with complete historical data, the conditional probability table (CPT) can be directly calculated through maximum likelihood estimation, which can make full use of existing data and ensure the accuracy of probability parameters.

[0014] For nodes with missing data, the Dirichlet prior distribution constrained by expert experience is introduced, and the probability parameters are supplemented through Bayesian reasoning, which effectively solves the problems caused by missing data and improves the overall utilization of data. The EM algorithm is an iterative algorithm that can find the parameter estimates of the model through continuous trial and error and maximizing the likelihood function, making the model's description of the data generation process more accurate. Combined with domain knowledge, especially expert experience, it can provide additional constraints and guidance for the model, further enhancing the robustness and reliability of the model.

[0015] Accurate conditional probability relationship is the basis for constructing fault causal network, which directly affects the accuracy of fault prediction. The conditional probability relationship established by integrating EM algorithm with domain knowledge is more accurate, which can improve the accuracy of fault prediction and provide more reliable decision support for the operation and maintenance of power system.

[0016] Optionally, in the simulation optimization step, the probability adjustment threshold in the simulation optimization is dynamically adjusted, and the dynamic adjustment method is: ,in is the baseline threshold, is the meteorological influence coefficient, , It is the real-time meteorological disaster warning level. It is the maximum value of the meteorological disaster warning level.

[0017] By adopting the above technical solution and dynamically adjusting the probability adjustment threshold, it is possible to more accurately reflect the impact of real-time meteorological conditions on the probability of power system failure, thereby improving the accuracy of the simulation; it is possible to better simulate the impact of severe weather conditions on the power system, and provide a more reliable basis for power system fault prevention and emergency handling.

[0018] The dynamic adjustment method takes into account the real-time changes in meteorological conditions, enabling the model to be adaptively adjusted according to different meteorological conditions, thereby enhancing the adaptability and flexibility of the model; by dynamically adjusting the probability adjustment threshold, the failure probability of the power system can be more accurately predicted, thereby optimizing resource allocation, improving the operating efficiency of the power system, and at the same time improving the accuracy of the simulation results.

[0019] Optionally, after the first simulation step, a model building step is also included; Model construction: including: a first model construction step, a prediction step, a second model construction step, a training step and a second simulation step; First model construction: establish the first Bayesian network model, the first-level nodes of the first Bayesian network model are the historical distribution network operation data, and the directed edges of the Bayesian network model are the conversion probabilities between the historical operation data of each node; Prediction: Input the real-time operation data of the distribution network into the first Bayesian network model to obtain the operation data at the next moment, which is recorded as simulation data; Second model construction: Establish a full-node dynamic model of the distribution network, integrating the real-time measurement data of the SCADA system and the synchronous phasor information of the PMU; Training: Input simulation data and historical operation data into the full-node dynamic model of the distribution network to train the model; Second simulation: The fault data is input into the trained full-node dynamic model of the distribution network, and then the output step is performed.

[0020] By adopting the above technical solution, the establishment of the first Bayesian network model can use the historical distribution network operation data to predict the operation data at the next moment; this probability-based prediction method can take into account the uncertainty between the data, thereby improving the accuracy of the prediction; by inputting simulation data and historical operation data into the full-node dynamic model of the distribution network for training, the model can learn the dynamic behavior characteristics of the distribution network, and further improve the ability to predict future fault conditions; the Bayesian network model can flexibly handle the dependencies between different nodes, as well as the changes of these relationships over time; this enables the model to adapt to various changes in the distribution network structure, equipment status and operating environment.

[0021] The full-node dynamic model of the distribution network integrates the real-time measurement data of the SCADA system and the synchronous phasor information of the PMU, which can reflect the actual operating status of the distribution network in real time and enhance the real-time and adaptability of the model.

[0022] By first establishing a Bayesian network model for preliminary prediction and then establishing a full-node dynamic model of the distribution network for training and simulation, the complete process from data input to result output is optimized; it can improve the efficiency of fault simulation and prediction while ensuring the accuracy of the results.

[0023] Optionally, before the second model building step, a modeling step is also included; Modeling: including data screening step and third model building step; Data screening: Screen historical operation data to obtain historical fault data; The third model construction: establish the second Bayesian network model, the first-level nodes of the second Bayesian network model are historical fault data, the second-level nodes are the constraints of the distribution network in each period, and establish directed edges from the first-level nodes to the second-level nodes; Data processing: input the fault data into the second Bayesian network model to obtain the processed fault data, and then perform the second simulation step.

[0024] By adopting the above technical solution, historical fault data are extracted from the historical operation data through the data screening step. These data more directly reflect the fault characteristics and laws of the distribution network; the screened data has higher quality, reduces noise and redundant information, and provides a more reliable data basis for subsequent model construction and fault simulation.

[0025] The second Bayesian network model uses historical fault data as the first-level nodes and the constraints of the distribution network in each time period as the second-level nodes to construct a relationship network with directed edges; this model structure can more comprehensively consider the complex relationship between fault data and distribution network constraints, and improve the depth and breadth of fault analysis; by inputting fault data into the second Bayesian network model, processed fault data can be obtained, which are more detailed and accurate, and contribute to the accuracy of subsequent fault simulation; the fault data after data screening and processing by the second Bayesian network model is closer to the actual operation of the distribution network.

[0026] In the second simulation step, these processed fault data are input into the trained full-node dynamic model of the distribution network, which can more accurately simulate and predict the fault conditions of the distribution network; this not only improves the accuracy of fault simulation and prediction, but also enhances its practicality, providing more powerful support for the operation and maintenance and fault handling of the power system.

[0027] Optionally, after the data screening step, a distributed power source determination step is also included; Distributed power source judgment: judge whether there is a distributed power source in the distribution network, if yes, perform the third model optimization step, otherwise, perform the third model construction step; Optimization of the third model: intervene in the construction steps of the third model, add three-level nodes, the three-level nodes are the distributed power supply fault ride-through control strategy, and establish directed edges between the first-level nodes and the third-level nodes, and directed edges between the second-level nodes and the third-level nodes respectively.

[0028] By adopting the above technical solution, through the distributed power judgment step, it is possible to identify whether there is a distributed power supply in the distribution network. When there is a distributed power supply in the distribution network, the third model optimization step will add a third-level node, that is, a distributed power supply fault ride-through control strategy; this strategy can take into account the behavioral characteristics of the distributed power supply under fault conditions, thereby improving the accuracy of fault analysis; by establishing directed edges between the first-level node, the second-level node and the third-level node, the model can more comprehensively describe the complex relationship between the fault data and the distributed power supply fault ride-through control strategy; the optimized third model can more accurately simulate and predict the behavior of the distribution network under fault conditions.

[0029] Optionally, a fault construction step is further included before the first simulation step; Fault construction: Construct an extended admittance matrix with fault impedance parameters, solve the network power flow distribution under fault boundary conditions, and intervene in the first simulation step.

[0030] By adopting the above technical solution and solving the network power flow distribution under fault boundary conditions, the impact range and degree of the fault on the power system can be analyzed, providing strong support for subsequent fault handling and recovery.

[0031] Optionally, the expanded admittance matrix in the fault construction step satisfies: ; in, is the basic admittance matrix, is the time-varying impedance of the nth fault point, is the fault correlation matrix.

[0032] By adopting the above technical solution, the extended admittance matrix constructed in the fault construction step takes into account the influence of the fault impedance parameters, which is closer to the actual situation when the distribution network fails; by solving the network flow distribution under the fault boundary conditions, it can more accurately reflect the impact of the fault on the distribution network, thereby improving the accuracy of fault simulation.

[0033] The extended admittance matrix The time-varying impedance of multiple fault points This enables the model to flexibly adapt to situations with different fault points and fault types; by adjusting the fault impedance parameters and the fault correlation matrix, different fault scenarios can be simulated and predicted, which enhances the adaptability and versatility of the model.

[0034] Due to the time-varying impedance The extended admittance matrix can reflect the change of the impedance of the fault point over time, so it can support real-time fault analysis and location; this helps operation and maintenance personnel to quickly and accurately identify the fault location, take appropriate treatment measures, and reduce the impact of the fault on the distribution network; through accurate fault simulation and prediction, operation and maintenance personnel can formulate fault prevention and emergency treatment plans in advance.

[0035] Optionally, the fault-crossing control strategy in the third model optimization step includes: LVRT voltage support curve of photovoltaic inverter; Virtual synchronous machine inertia response of energy storage converter; Action logic sequence of the crowbar circuit of a doubly-fed wind turbine.

[0036] By adopting the above technical solution, the existing technology often only focuses on a single fault ride-through control strategy, such as only considering the LVRT technology of the photovoltaic inverter or the inertia response of the energy storage converter, while the present technical solution comprehensively considers multiple fault ride-through control strategies to form a more comprehensive and systematic fault response plan.

[0037] By introducing actual fault ride-through control strategies (such as LVRT voltage support curve, virtual synchronous machine inertia response and crowbar circuit action logic timing), this technical solution makes the model closer to the actual situation when simulating and predicting distribution network faults, thereby improving the practicality and accuracy of the model.

[0038] In summary, the present application includes at least one of the following beneficial technical effects: 1. This method can simulate a variety of fault conditions, including setting different fault points, fault types and fault parameters, so as to meet the diverse needs of distribution network fault analysis; users can flexibly adjust fault parameters during the simulation process according to actual needs to observe the response of the distribution network under different fault conditions; by observing and recording the voltage and current changes of each node during the fault simulation process, as well as the overall response of the distribution network, it can provide a strong basis for the operation analysis, fault diagnosis and decision support of the power system; this method can also be used to evaluate the stability and reliability of the distribution network, and provide a reference for the planning and design of the power system; 2. By adjusting the fault impedance parameters and fault correlation matrix, different fault scenarios can be simulated and predicted, which enhances the adaptability and versatility of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flowchart of the fault simulation method in Example 1 of the present application; Figure 2 is a detailed step flow chart of the fault risk assessment step in Example 2 of the present application; Figure 3 is a detailed step flow chart of the model building step in Example 2 of the present application; Figure 4 is a detailed step flow chart of the modeling step in Example 2 of the present application; Figure 5 It is a flowchart of the fault construction in Example 2 of the present application. DETAILED DESCRIPTION

[0040] The following combination Figures 1 to 5 This application is described in further detail.

[0041] This embodiment discloses a distribution network fault simulation method based on power flow calculation.

[0042] Example 1: Reference Figure 1 , a distribution network fault simulation method based on power flow calculation includes the following steps: First acquisition: acquiring line information of the target distribution network, wherein the line information includes a line ID, a rated voltage, and a topological structure of each line; Specifically, line information is extracted from the database or design drawings of the distribution network. The line information should include the unique identifier (line ID) of each line, the rated voltage, and the topological connection relationship of the line. This information can be automatically extracted using data import tools or scripts.

[0043] First construction: constructing a topological data map of the distribution network according to the line information, and correcting and simplifying the topological data map through the measurement data; Specifically, the line information is converted into a structure of nodes and edges using graph theory methods. The nodes represent the connection points in the power grid, and the edges represent the lines connecting these nodes. According to the rated voltage and topological structure of the lines, corresponding attributes are assigned to the edges and nodes in the graph. The topological data graph is verified using measurement data (such as voltage and current measurements) to ensure the accuracy of the model. The topological data graph is simplified and unnecessary details are removed to improve computational efficiency.

[0044] First calculation: Based on the corrected and simplified topological data graph, the forward-backward algorithm is used to perform power flow calculation to obtain the voltage and current distribution of each node; Specifically, the forward-backward algorithm is a power flow calculation method suitable for radial distribution networks; starting from the power source point, the power distribution of each node is calculated forward until the leaf node, and then starting from the leaf node, the voltage distribution of each node is calculated backward until the root node. Through iteration, the power and voltage distribution are continuously adjusted until the convergence conditions are met, and finally the voltage and current distribution of each node are obtained.

[0045] First simulation: simulating the fault situation in the distribution network according to the voltage and current distribution, including setting the fault point, fault type and fault parameters to form fault data; Output: Observe and record the voltage and current changes of each node during the fault simulation, as well as the overall response of the distribution network, and output the change record report.

[0046] Specifically, on the basis of power flow calculation, set the fault point, fault type (such as three-phase short circuit, single-phase grounding, etc.) and fault parameters (such as fault resistance, fault duration, etc.); modify the attributes of the corresponding nodes and edges in the topological data graph according to the fault type and parameters; re-calculate the power flow to observe the impact of the fault on the distribution network; record the voltage and current changes of each node during the fault simulation; use visualization tools (such as charts, animations, etc.) to display the voltage and current changes of each node during the fault simulation; record the overall response of the distribution network, including the fault propagation range, protection action, etc.; generate a change record report, including the comparison before and after the fault, changes in key indicators, etc.

[0047] For example, suppose there is a simple radial distribution network consisting of three nodes (A, B, C) and two lines (AB, BC), where A is the power supply point with a rated voltage of 10kV.

[0048] The line information is extracted from the database to obtain the line ID, rated voltage and topological structure. A topological data graph is constructed, where nodes A, B and C represent three connection points respectively, and edges AB and BC represent the lines connecting these nodes. The topological graph is verified and simplified using the measured data. The forward-backward algorithm is used to calculate the power flow, and the voltages of nodes A, B and C are 10 kV, 9.8 kV and 9.6 kV respectively, and the currents are 100 A, 98 A and 96 A respectively.

[0049] A three-phase short-circuit fault is set at node B, with a fault resistance of 0.1Ω and a fault duration of 0.1 seconds. Modify the topology data diagram and recalculate the power flow; observe and record the voltage and current changes of each node during the fault simulation process; it is found that the voltage of node B drops sharply and the current rises sharply; the voltage of node A drops slightly and the current rises; the voltage and current of node C are also affected to a certain extent; generate a change record report, including the voltage and current comparison before and after the fault, the fault propagation range and the protection action status, etc.

[0050] Example 2: Reference Figure 2 The difference between this embodiment and embodiment 1 is that before the first simulation step, a fault risk assessment step is also included: Fault risk assessment: including: building fault causal network steps, generating fault probability distribution steps and simulation optimization steps; Constructing a fault causal network: Define Bayesian network node variables, including: equipment state variables (insulation level, service life), environmental variables (temperature, humidity, lightning activity intensity), fault type variables (phase short circuit, ground fault, disconnection), and establish conditional probability relationships between nodes based on historical fault recording data and operation and maintenance records. The establishment of conditional probability relationships uses the EM algorithm and domain knowledge fusion method, which includes: For nodes with complete historical data, the CPT table is directly calculated by maximum likelihood estimation; For nodes with missing data, the Dirichlet prior distribution constrained by expert experience is introduced, and the probability parameters are completed through Bayesian reasoning; Specifically, equipment state variables: nodes are defined according to the equipment’s insulation grade, service life and other attributes, which reflect the equipment’s aging degree and potential failure risk.

[0051] Environmental variables: Nodes are defined based on meteorological conditions such as temperature, humidity, and intensity of lightning activity, which may affect the operating status and failure probability of the equipment.

[0052] Fault type variable: defines different types of faults, such as phase short circuit, ground fault, and line break. These fault types reflect specific problems that may occur in the equipment.

[0053] Generate fault probability distribution: input real-time meteorological data and equipment monitoring data into the fault causal network; calculate the fault probability of each line section ,in is the probability of occurrence of the kth type of failure, is the kth type of fault, for environmental / equipment evidence; Simulation optimization: Generate candidate fault sets by descending order of fault probability ,in represents the i-th line section in the distribution network, represents the possible fault type of the i-th line section, is the probability of the jth fault occurring in the ith line section, is the probability threshold (α≥0.2); Will The input parameters are passed to the first simulation step to simulate the fault in combination with the voltage and current distribution.

[0054] Specifically, the candidate fault set is generated according to the descending order of the fault probability. The candidate fault set contains those line sections and fault type combinations whose fault probability is higher than a certain threshold α; the threshold α should be selected based on risk management requirements and resource constraints, and is usually set to 0.2 or higher to ensure that the fault events of concern have sufficient severity and probability.

[0055] For example, it is found that the fault probability of phase-to-phase short circuit in a certain line section increases significantly due to the influence of high temperature and high humidity; the candidate fault set is generated according to the descending order of the fault probability. For example, the threshold α is set to 0.2, and the line sections and fault type combinations with a fault probability higher than 0.2 are screened out; the candidate fault set is passed as an input parameter to the first simulation step. For example, for the high-risk fault events screened out (such as phase-to-phase short circuit in a certain line section), detailed fault simulation and analysis are performed in combination with the distribution of voltage and current.

[0056] In the first simulation step, the impact of fault events on the distribution network is simulated based on the candidate fault set and voltage and current distribution; the voltage and current changes of each node during the fault simulation and the overall response of the distribution network are observed and recorded; a fault simulation report is generated, including comparisons before and after the fault, changes in key indicators, fault propagation range and protection action, etc., to provide decision support for subsequent fault prevention and response measures.

[0057] In other implementations, in the simulation optimization step, the probability adjustment threshold in the simulation optimization is dynamically adjusted, and the dynamic adjustment method is: ,in is the baseline threshold, is the meteorological influence coefficient, , It is the real-time meteorological disaster warning level. It is the maximum value of the meteorological disaster warning level.

[0058] Specifically, the real-time meteorological disaster warning level is obtained through official warning information released by the meteorological department, professional meteorological forecasting systems or third-party meteorological service providers. This information usually includes the warning type (such as heavy rain, lightning, strong wind, etc.), warning level (such as blue, yellow, orange, red, etc.) and the corresponding numerical representation; meteorological impact coefficient The determination can be made through historical data analysis, expert evaluation or research based on the correlation between meteorological disasters and power grid failures. Usually, this requires comprehensive consideration of factors such as the type, intensity, duration of meteorological disasters and the disaster resistance of power grid equipment; obtaining real-time meteorological data such as temperature, humidity, wind speed, wind direction, precipitation, etc. through meteorological monitoring equipment such as meteorological observation stations, radars, and satellites; using data services provided by meteorological service providers to obtain meteorological data for specific areas and specific times. These data usually include historical data and real-time data, which can be used for meteorological analysis and prediction.

[0059] For example, suppose a distribution network is undergoing fault risk assessment and simulation analysis. At a certain time t, the real-time meteorological disaster warning level is orange warning, and the corresponding value is 6 (assuming the maximum warning level is red, the value is 10). Benchmark threshold Set to 0.2, the meteorological influence coefficient It was determined to be 0.1 after expert evaluation.

[0060] Through the official warning information released by the meteorological department, the meteorological disaster warning level at time t is obtained as orange, and the value is 6; after expert evaluation, the meteorological impact coefficient β is determined to be 0.1.

[0061] According to the formula , calculate the dynamic threshold at time t is 0.26.

[0062] The dynamic threshold of 0.26 is used to replace the static threshold of 0.2, and the candidate fault set is regenerated. This includes those line sections and fault type combinations with a fault probability higher than 0.26. The candidate fault set is passed for simulation: the generated candidate fault set is passed as an input parameter to the first simulation step, and the fault is simulated and analyzed in combination with the distribution of voltage and current. The simulation results will be used to guide subsequent fault prevention and response measures.

[0063] Reference Figure 3 , in other embodiments, a model building step is further included after the first simulation step; Model construction: including: a first model construction step, a prediction step, a second model construction step, a training step and a second simulation step; First model construction: establish the first Bayesian network model, the first-level nodes of the first Bayesian network model are the historical distribution network operation data, and the directed edges of the Bayesian network model are the conversion probabilities between the historical operation data of each node; Specifically, the first-level node is set as follows: First, the first-level node is determined to be the historical distribution network operation data. These data may include historical records of key parameters such as voltage, current, power factor, load, temperature, etc.; then, based on the historical data, the correlation between each node (i.e., different operating parameters) is analyzed to determine the conversion probability between them. This usually involves statistical analysis and probabilistic modeling of historical data; using Bayesian network construction tools or software, the first Bayesian network model is constructed based on the first-level nodes and conversion probabilities. This model will be used to predict the operating status of the distribution network at the next moment.

[0064] Prediction: Input the real-time operation data of the distribution network into the first Bayesian network model to obtain the operation data at the next moment, which is recorded as simulation data; Second model construction: Establish a full-node dynamic model of the distribution network, integrating the real-time measurement data of the SCADA system and the synchronous phasor information of the PMU; Training: Input simulation data and historical operation data into the full-node dynamic model of the distribution network to train the model; Specifically, collect and integrate the real-time measurement data provided by the SCADA system and the synchronized phasor information provided by the PMU (phasor measurement unit). These data will be used to build a full-node dynamic model of the distribution network; build a full-node dynamic model based on the physical structure and dynamic behavior of the distribution network. This model should be able to reflect the dynamic characteristics of the distribution network under various operating conditions; input simulation data and historical operating data into the full-node dynamic model of the distribution network to train the model; input simulation data (from the prediction results of the first Bayesian network model) and historical operating data into the full-node dynamic model of the distribution network; use machine learning or optimization algorithms to train the model. The goal of the training is to enable the model to more accurately predict the operating status of the distribution network and adapt to different operating conditions and fault conditions.

[0065] Second simulation: The fault data output by the first simulation is input into the trained full-node dynamic model of the distribution network, and then the output step is performed.

[0066] Specifically, in the trained full-node dynamic model of the distribution network, preset fault data are input. These fault data may include different types of short circuits, open circuits, overloads and other fault conditions. The model is run to simulate the impact of the fault on the distribution network. The output includes key parameters such as voltage, current, power distribution after the fault, as well as possible fault propagation paths and impact ranges.

[0067] Reference Figure 4 In other embodiments, before the second model building step, a modeling step is also included; Modeling: including data screening step and third model building step; Data screening: Screen historical operation data to obtain historical fault data; Specifically, first, determine the screening criteria, such as fault type, fault occurrence time, fault duration, etc., and then use database query or data processing software to screen the historical operation data according to the screening criteria to obtain a historical fault data set.

[0068] Distributed power source judgment: judge whether there is a distributed power source in the distribution network, if yes, perform the third model optimization step, otherwise, perform the third model construction step; Specifically, by consulting the design documents and operation records of the distribution network or consulting the distribution network management personnel, determine whether there is a distributed power source in the distribution network. If so, the third model optimization step needs to be performed; if not, the third model construction step is directly performed; The third model construction: establish the second Bayesian network model. The first-level nodes of the second Bayesian network model are historical fault data, and the second-level nodes are the constraints of the distribution network in each period. The constraints include load level, voltage range, current limit, etc., and establish directed edges from the first-level nodes to the second-level nodes. Optimization of the third model: intervene in the construction steps of the third model, add three-level nodes, the three-level nodes are the distributed power supply fault ride-through control strategy, and establish directed edges between the first-level nodes and the third-level nodes, and directed edges between the second-level nodes and the third-level nodes respectively; among them, the fault ride-through control strategy includes: LVRT voltage support curve of photovoltaic inverter; virtual synchronous machine inertia response of energy storage converter; doubly fed wind turbine crowbar circuit action logic timing.

[0069] Specifically, directed edges between the first-level nodes and the third-level nodes are established, indicating that the fault data will affect the fault-crossing control strategy of the distributed power source; at the same time, directed edges between the second-level nodes and the third-level nodes are established, indicating that the constraints will also affect the fault-crossing control strategy of the distributed power source.

[0070] Data processing: input the fault data into the second Bayesian network model to obtain the processed fault data, and then perform the second simulation step.

[0071] Reference Figure 5 , in other embodiments, a fault construction step is further included before the first simulation step; Fault construction: Construct an extended admittance matrix with fault impedance parameters, solve the network power flow distribution under fault boundary conditions, and intervene in the first simulation step.

[0072] The expanded admittance matrix in the fault construction step satisfies: ; in, is the basic admittance matrix, is the time-varying impedance of the nth fault point, is the fault correlation matrix.

[0073] Specifically, firstly, according to the topological structure and line parameters of the distribution network, the basic admittance matrix is ​​constructed. . This matrix reflects the admittance relationship of the network under normal conditions. For example, there is a simple three-node distribution network, in which there is a line connecting node 1 and node 2, there is also a line connecting node 2 and node 3, and there is no direct line connecting node 1 and node 3. Assume that the admittance of all lines is the same, and ignore the reactance part of the line, only consider resistance and conductance; we can regard each node as an electrical node, and each line as an admittance element. According to Kirchhoff's current law and voltage law, we can write the current balance equation for each node. Then, by eliminating the voltage variable, we can get a set of equations containing only current and admittance. The coefficient matrix of this set of equations is the basic admittance matrix: .

[0074] For each potential fault point n, determine the time-varying impedance according to its fault type and severity ,This impedance may include components such as resistance, inductance, and capacitance, and is used to simulate the electrical impact of a fault on the network.

[0075] Constructing the fault correlation matrix , which is used to describe the relationship between the fault point n and other nodes in the network. The elements of this matrix are usually 0 or 1, indicating whether the node is affected by the fault point n.

[0076] According to the formula , calculate the extended admittance matrix, where represents the change in admittance due to the fault, Represents the Kronecker product operation of matrices.

[0077] The extended admittance matrix is ​​used to solve the network power flow distribution under fault boundary conditions in order to evaluate the impact of the fault on the distribution network.

[0078] Implementation method: Boundary condition setting: According to the fault type and location, set the corresponding boundary conditions, such as the voltage, current or power at the fault point; Power flow calculation: Use power flow calculation algorithms (such as Newton-Raphson method, DC power flow method, etc.), combined with the extended admittance matrix and boundary conditions, to solve the network power flow distribution. This includes the voltage, current, power and other parameters of each node.

[0079] Result analysis: Analyze the calculation results and evaluate the impact of the fault on the network, such as voltage fluctuation, current overload, power imbalance, etc.

[0080] The purpose of the intervention in the first simulation step is to apply the results of the fault construction step to the first simulation step in order to more accurately simulate the distribution network behavior under fault conditions.

[0081] Implementation method: Data transfer: The extended admittance matrix and network power flow distribution results obtained in the fault construction step are transferred to the first simulation step.

[0082] Model adjustment: In the first simulation step, the distribution network model is adjusted according to the delivered data so that it reflects the network status under fault conditions.

[0083] Simulation operation: Use the adjusted model to perform simulation operation to obtain the simulation results of distribution network behavior under fault conditions.

[0084] For example, suppose there is a potential short-circuit fault point in a distribution network, located on a feeder. In order to evaluate the impact of the fault on the distribution network, the following steps are performed: According to the topological structure and line parameters of the distribution network, the basic admittance matrix is ​​constructed , determine the time-varying impedance of the fault point , is assumed to be a pure resistance value and is used to simulate a short-circuit fault.

[0085] Constructing the fault correlation matrix , which represents the relationship between the fault point and other nodes in the network; according to the formula , calculate the extended admittance matrix , set the fault boundary conditions, such as the voltage drop at the fault point is 0 (short circuit fault).

[0086] Use the power flow calculation algorithm combined with the extended admittance matrix and boundary conditions to solve the network power flow distribution; analyze the calculation results and find that the fault causes the voltage drop, current increase and power imbalance of nearby nodes; pass the extended admittance matrix and network power flow distribution results to the first simulation step; in the first simulation step, adjust the distribution network model so that it can reflect the network status under fault conditions; use the adjusted model to perform simulation operations to obtain the simulation results of the distribution network behavior under fault conditions, which may include voltage distribution, current distribution, power flow, etc. after the fault, which are used to further analyze and evaluate the impact of the fault on the distribution network.

[0087] Traditional fault simulation is mostly based on deterministic models or static probabilities. This solution constructs a fault causal network (Bayesian network) and combines historical data, real-time meteorological and equipment monitoring data to dynamically calculate the fault probability distribution. It also introduces the Dirichlet prior distribution and EM algorithm constrained by expert experience to solve the problem of missing data and improve the robustness of probabilistic reasoning.

[0088] In addition, existing technologies mostly use fixed thresholds, while this solution proposes to dynamically adjust the probability threshold, realize the linkage between the threshold and the external environment through the meteorological impact coefficient, and adaptively optimize the candidate fault set in combination with the real-time meteorological disaster level to improve the real-time and accuracy of fault simulation.

[0089] Through fault-crossing control strategies (such as crowbar circuit action logic timing), the dynamic behavior of distributed power sources during faults is simulated, which is closer to the actual scenarios of modern high-penetration renewable energy distribution networks; by describing the relationship between the fault point and the network topology, it supports accurate simulation of complex fault scenarios (such as multi-point time-varying impedance faults).

[0090] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A distribution network fault simulation method based on power flow calculation, characterized in that: The following steps are involved: First acquisition: acquiring line information of the target distribution network, wherein the line information includes a line ID, a rated voltage, and a topological structure of each line; First construction: constructing a topological data map of the distribution network according to the line information, and correcting and simplifying the topological data map through the measurement data; First calculation: Based on the corrected and simplified topological data graph, the forward-backward algorithm is used to perform power flow calculation to obtain the voltage and current distribution of each node; First simulation: simulating the fault situation in the distribution network according to the voltage and current distribution, including setting the fault point, fault type and fault parameters to form fault data; Output: Observe and record the voltage and current changes of each node during the fault simulation, as well as the overall response of the distribution network, and output the change record report.

2. The distribution network fault simulation method based on power flow calculation according to claim 1 is characterized in that: Before the first simulation step, a failure risk assessment step is also included: Fault risk assessment: including: building fault causal network steps, generating fault probability distribution steps and simulation optimization steps; Constructing a fault causal network: defining Bayesian network node variables, including equipment state variables, environmental variables, and fault type variables, and establishing conditional probability relationships between nodes based on historical fault recording data and operation and maintenance records; Generate fault probability distribution: input real-time meteorological data and equipment monitoring data into the fault causal network; calculate the fault probability of each line section ,in is the probability of occurrence of the kth type of failure, is the kth type of fault, for environmental / equipment evidence; Simulation optimization: Generate candidate fault sets by descending order of fault probability ,in represents the i-th line section in the distribution network, represents the possible fault type of the i-th line section, is the probability of the jth fault occurring in the ith line section, is the probability threshold, and α≥0.2; Will The input parameters are passed to the first simulation step to simulate the fault in combination with the voltage and current distribution.

3. The distribution network fault simulation method based on power flow calculation according to claim 2 is characterized in that: In the step of building the fault causal network, the conditional probability relationship is established by using the EM algorithm and the domain knowledge fusion method, which specifically includes: For nodes with complete historical data, the CPT table is directly calculated by maximum likelihood estimation; For nodes with missing data, the Dirichlet prior distribution constrained by expert experience is introduced, and the probability parameters are completed through Bayesian reasoning.

4. The distribution network fault simulation method based on power flow calculation according to claim 2 is characterized in that: In the simulation optimization step, the probability adjustment threshold in the simulation optimization is dynamically adjusted. The dynamic adjustment method is: ,in is the baseline threshold, is the meteorological influence coefficient, , It is the real-time meteorological disaster warning level. It is the maximum value of the meteorological disaster warning level.

5. The distribution network fault simulation method based on power flow calculation according to any one of claims 1 to 4, characterized in that: After the first simulation step, a model building step is also included; Model construction: including: a first model construction step, a prediction step, a second model construction step, a training step and a second simulation step; First model construction: establish the first Bayesian network model, the first-level nodes of the first Bayesian network model are the historical distribution network operation data, and the directed edges of the Bayesian network model are the conversion probabilities between the historical operation data of each node; Prediction: Input the real-time operation data of the distribution network into the first Bayesian network model to obtain the operation data at the next moment, which is recorded as simulation data; Second model construction: Establish a full-node dynamic model of the distribution network, integrating the real-time measurement data of the SCADA system and the synchronous phasor information of the PMU; Training: Input simulation data and historical operation data into the full-node dynamic model of the distribution network to train the model; Second simulation: The fault data is input into the trained full-node dynamic model of the distribution network, and then the output step is performed.

6. The distribution network fault simulation method based on power flow calculation according to claim 5 is characterized in that: Before the second model building step, a modeling step is also included; Modeling: It includes data screening step and third model building step; Data screening: Screen historical operation data to obtain historical fault data; The third model construction: establish the second Bayesian network model, the first-level nodes of the second Bayesian network model are historical fault data, the second-level nodes are the constraints of the distribution network in each period, and establish directed edges from the first-level nodes to the second-level nodes; Data processing: input the fault data into the second Bayesian network model to obtain the processed fault data, and then perform the second simulation step.

7. The distribution network fault simulation method based on power flow calculation according to claim 6 is characterized in that: After the data screening step, a distributed power source determination step is also included; Distributed power source judgment: judge whether there is a distributed power source in the distribution network, if yes, perform the third model optimization step, otherwise, perform the third model construction step; Optimization of the third model: intervene in the construction steps of the third model, add three-level nodes, the three-level nodes are the distributed power supply fault ride-through control strategy, and establish directed edges between the first-level nodes and the third-level nodes, and directed edges between the second-level nodes and the third-level nodes respectively.

8. The distribution network fault simulation method based on power flow calculation according to claim 6 is characterized in that: A fault construction step is also included before the first simulation step; Fault construction: Construct an extended admittance matrix with fault impedance parameters, solve the network power flow distribution under fault boundary conditions, and intervene in the first simulation step.

9. The distribution network fault simulation method based on power flow calculation according to claim 8, characterized in that: The expanded admittance matrix in the fault construction step satisfies: ； in, is the basic admittance matrix, is the time-varying impedance of the nth fault point, is the fault correlation matrix.

10. The distribution network fault simulation method based on power flow calculation according to claim 7, characterized in that: The fault ride-through control strategy in the third model optimization step includes: LVRT voltage support curve of photovoltaic inverter; Virtual synchronous machine inertia response of energy storage converter; Action logic sequence of the crowbar circuit of a doubly-fed wind turbine.

Images