A hierarchical and phased fault recovery method and system for active power distribution network

Abstract

The application discloses a kind of layered stage fault recovery method and system for active distribution network, method includes: S1, synchronous acquisition multi-source data, based on current mutation and voltage drop double criterion positioning fault;S2, establish mixed integer programming by second order cone conversion reconfiguration topology, according to three-level weight priority restoration key load;S3, establish the full-dimension power flow constraint containing DG, SNOP, energy storage and flexible load;S4, generate four boundary scenarios with interval number to describe uncertainty, solve recovery scheme by robustness and multi-objective weighting;S5, check by scene and adjust according to DG energy storage, load reduction, topology fine-tuning three progressive correction to all-scenario feasible;S6, according to time sequence execution and closed loop check, abnormal fast correction.The application considers speed, economy and robustness.

Description

Technical Field

[0001] This invention mainly relates to the field of power distribution network fault recovery technology, specifically to a hierarchical and phased fault recovery method and system for active power distribution networks. Background Technology

[0002] With the rapid development of renewable energy, a large number of distributed generation (DG) sources have been connected to the distribution network, leading to the transformation of traditional distribution networks into active distribution networks. However, the randomness of DG output and the volatility of loads cause power flow fluctuations and voltage deviations in the distribution network, increasing the difficulty of power restoration after a fault. Power restoration in traditional distribution networks mainly relies on network reconfiguration or islanding, restoring power through the operation of mechanical tie switches. However, mechanical switches have slow response speeds, limited number of operations, and do not fully consider the uncertainties of DG and loads, resulting in insufficient reliability of the restoration scheme.

[0003] Soft normally open points (SNOPs), as a novel type of equipment based on fully controlled power electronic devices, offer advantages such as fast response speed and flexible control. They can achieve load balancing between feeders and uninterrupted power supply during faults, significantly improving the control performance of distribution networks compared to traditional tie switches. Existing research on power restoration using SNOPs often focuses on single-objective optimization or fails to adequately consider uncertainties. When using probabilistic models to handle uncertainties, a large amount of random information is required, limiting their practicality. While robust optimization methods do not require probability distributions, traditional methods are computationally complex and struggle to obtain optimal solutions.

[0004] Furthermore, existing recovery methods do not fully integrate the active / reactive power control capabilities of distributed generation (DG), making it difficult to meet the high-precision, real-time operational optimization requirements in scenarios with frequent load and DG fluctuations. Therefore, there is an urgent need for a power supply recovery method that balances load loss minimization, operational constraint satisfaction, and uncertainty handling to improve the fault recovery capabilities of active distribution networks. Summary of the Invention

[0005] To address the technical problems existing in the prior art, this invention provides a hierarchical and phased fault recovery method and system for active distribution networks that improves the speed, safety, and reliability of fault recovery.

[0006] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows: A hierarchical and phased fault recovery method for active distribution networks includes the following steps: Step S1: When a fault occurs in the distribution network, multi-source operation data such as branch current, node voltage, real-time output of distributed generation (DG), state of charge (SOC) of energy storage system (ESS), and load power are collected synchronously before and after the fault. Based on the dual criteria of current mutation rate and voltage drop amplitude, the faulty branch is located collaboratively, and a switch control command is generated to drive the sectionalizing switch to complete the disconnection action to isolate the faulty section and output the non-faulty area range. Step S2: Using the non-faulty area range output in Step S1 as the input boundary, a mixed integer programming model is established with the goal of minimizing weighted load loss. The branch switch state variables and the power transmission range of each terminal of the soft normally open point SNOP are defined as decision variables. The nonlinear AC power flow constraints are transformed into convex second-order cone constraints through second-order cone transformation. A two-stage solution strategy combining heuristic algorithm and linearization optimization is adopted to output a network topology scheme that satisfies connectivity constraints, acyclic constraints, and capacity constraints, as well as the power operation range of each terminal of the SNOP. Step S3: Based on the network topology scheme output in step S2, establish a power flow calculation model for the distribution network, construct a full-dimensional operation constraint system covering node voltage limits, branch power flow limits, DG inverter capacity, SNOP power balance, ESS charging and discharging, and the adjustable range of flexible loads, and output the complete constraint feasible domain boundary. Step S4: Within the feasible region of full-dimensional constraints output in step S3, the prediction uncertainty of DG output and load power is described using interval numbers. Based on the prediction error rate, multiple extreme boundary scenarios are generated. An interval robust optimization model is established with the goal of minimizing the average voltage deviation and network loss. A multi-objective weighted comprehensive optimization model is constructed by combining the load recovery rate. By solving the model, the active / reactive output setpoints of each DG, the charging and discharging power of each ESS, the flexible load transfer / reduction amount, and the terminal power setpoints of each SNOP are obtained. The output is the optimal recovery scheme. Step S5: Based on the optimal recovery scheme output in step S4, check whether each scenario in each extreme boundary scenario meets the constraints in the full-dimensional operation constraint system constructed in step S3. If any scenario violates any constraint, it is determined that there is a robustness defect and it is corrected in a three-level priority manner. The correction process is iteratively executed until all scenarios meet all constraints or the preset iteration limit is reached, and the final device control parameters are output. Step S6: Based on the final equipment control parameters output in step S5, design a conflict-free timing execution process according to the equipment response characteristics, and sequentially issue switch action commands, DG output commands, ESS charging and discharging commands, and flexible load transfer commands; simultaneously collect real-time operating data of the distribution network after restoration, and verify whether the restoration effect meets the preset constraint compliance rate, load restoration rate, and power quality indicators; if all verifications pass, the entire fault restoration process is completed; if any indicator fails to meet the standard, return to step S5 for correction, forming a closed-loop control.

[0007] Preferably, in step S1, the fault location branch based on the dual criteria of current mutation rate and voltage drop amplitude specifically includes: Calculate the transient current change ΔI and voltage change ΔU during the fault. When the current change rate ΔI / IN>1.5 and the voltage drop amplitude ΔU / UN>0.7 are simultaneously satisfied, the corresponding branch is determined to be a faulty branch, where IN is the rated current of the branch and UN is the nominal rated voltage of the 10kV distribution network.

[0008] Preferably, in step S2, the mixed-integer programming model minimizes the weighted load loss. The objective function is:

[0009] in, N L This represents the total number of load nodes in the non-faulty area. w i For nodes i Load weighting factor P L_i For nodes i Rated active load, x L_i This is a load recovery state variable.

[0010] Preferably, in step S3, the power flow calculation model of the distribution network is constructed using the Fast Decoupled Power Flow (FDLF) method, and its modified equation is:

[0011] Where, Δ P Δ Q These represent the active power and reactive power deviations at each node. , To simplify the admittance matrix of the distribution network, Δ d The node voltage phase angle deviation, Δ U For node voltage amplitude deviation, U This is the rated voltage for a 10kV distribution network.

[0012] Preferably, in step S3, the inverter capacity constraint of DG is:

[0013] Among them, S DG_k_N For the first k The rated capacity of the DG unit, cos The rated power factor of DG, P DG_k They actually contributed to DG. Q DG_k For DG, this is actually a waste of effort; , For the first k The minimum and maximum capacity of the DG (Distributed Generation) system.

[0014] Preferably, in step S4, the use of interval numbers to describe the prediction uncertainty of DG output and load power specifically includes:

[0015] in, P DG_k It has actually contributed to DG; P DG_k_pred For the first k The predicted active power output of Taiwan DG P L_i For nodes i Active load; P L_i_pred For nodes i The predicted active load value, d DG For DG output prediction error rate, d L This represents the load power prediction error rate.

[0016] Preferably, in step S4, the multiple extreme boundary scenarios are four boundary scenarios, each corresponding to the maximum and minimum combination of DG output and load power in the uncertain parameter space: Scene 1: P DG_k = P DG_k_pred (1+ d DG ), P L_i = P L_i_pred (1+ d L ); Scene 2: P DG_k = P DG_k_pred (1+ d DG ), P L_i = P L_i_pred (1- d L ); Scene 3: P DG_k = P DG_k_pred (1- d DG ), P L_i = P L_i_pred (1+ d L ); Scene 4: P DG_k = P DG_k_pred (1- d DG ), P L_i = P L_i_pred (1- d L ); The interval robust optimization model takes the worst case among the four boundary scenarios as the robust optimization result to ensure that the solution can still meet the operational constraints under the most unfavorable conditions.

[0017] Preferably, in step S4, the objective function of the multi-objective weighted comprehensive optimization model is... for:

[0018] in, a、b、c As weight, P L_i For nodes i Active load, For the maximum network loss, Δ U avg This represents the average voltage deviation. P loss The total active power loss of the system; x L_i This is a load recovery state variable.

[0019] Preferably, in step S5, the three-level priority progressive correction is performed sequentially in order of increasing cost: The first level of correction adjusts the reactive power output of DG and the charging and discharging power of each ESS, without involving load reduction. After the adjustment is completed, the power flow calculation and constraint verification are performed again by substituting into each boundary scenario. The second-level correction only reduces the third-level load by a ratio of n, prioritizing the uninterrupted power supply to the first and second-level loads, and re-verifies the constraints based on the results of the first-level correction. The third-level correction achieves topology fine-tuning by adjusting the switching status of tie switches or changing the location of DG access bus. After the adjustment, connectivity constraints and acyclic constraints need to be re-verified before the full-scenario constraint verification is completed.

[0020] The present invention also discloses a hierarchical and phased fault recovery system for active distribution networks, including an interconnected memory and a processor. The memory stores a computer program, which, when run by the processor, executes the steps of the method described above.

[0021] Compared with the prior art, the advantages of the present invention are as follows: The proposed method for hierarchical and phased fault recovery in active distribution networks with soft normally open points (SNOPs) and distributed generation (DGs) is based on a three-layer architecture design of "rapid structural recovery - coordinated optimization recovery - robustness enhancement and verification". It simplifies the mixed integer nonlinear programming problem by using second-order cone transformation and combines the active / reactive power control capabilities of DGs. While ensuring rapid power restoration after a fault and priority restoration of important loads, it achieves a balance between load restoration ratio, operational economy and uncertainty robustness, ultimately improving the speed, safety and reliability of fault recovery in active distribution networks.

[0022] After a fault occurs, this invention first collects key data before and after the fault through the feeder terminal unit (FTU) and synchronous phasor measurement unit (PMU). Based on the dual criteria of current mutation rate and voltage drop amplitude, the fault area is located and isolated within 3 seconds. Then, the network topology is quickly reconstructed through an algorithm combining heuristic rules and linear optimization. According to the three-level load priority optimization, the first and second level critical loads are restored first to ensure the core power demand. Next, a refined model is established based on the fast decoupling power flow method. With load recovery rate, network loss and voltage deviation as multiple objectives, the output of distributed generation, charging and discharging of energy storage system and flexible load transfer are coordinated and optimized. Then, considering the prediction uncertainty of DG output and load, the robustness of the scheme is checked through four boundary scenarios. The three-level iterative correction is carried out according to the priority of adjusting equipment output, reducing non-critical loads and fine-tuning network topology. Finally, control commands are issued in sequence according to the equipment response characteristics to complete the recovery execution. The indicators are checked through closed-loop monitoring. If the indicators are not met, the correction is quickly returned. Attached Figure Description

[0023] Figure 1 This is a flowchart of a hierarchical and phased fault recovery method for active distribution networks according to an embodiment of the present invention.

[0024] Figure 2This is a schematic diagram of energy flow under different conditions provided in this invention.

[0025] Figure 3 This is a diagram of three IEEE 33 bus power distribution systems connected via SNOP, as provided in this invention.

[0026] Figure 4 This is a power deviation diagram caused by relaxation, provided in this invention. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0028] like Figure 1 As shown in the figure, the hierarchical and phased fault recovery method for active distribution networks provided by this invention includes the following steps: Step S1: Collect distribution network operation and distributed resource data through FTU / DTU and PMU, and transmit the data in accordance with the IEC 61850 GOOSE protocol; locate the fault branch based on dual criteria and complete the isolation of the fault section within 3 seconds.

[0029] Step S1 provides accurate and real-time data support for all subsequent fault recovery decisions and is the foundation of the entire process.

[0030] When a fault occurs in the distribution network, the data acquisition system is activated to simultaneously acquire data on branch current, node voltage, real-time output of DG, energy storage SOC, and load power before and after the fault. The acquisition time covers 50ms before the fault to 1s after the fault to ensure the integrity of the fault transient data.

[0031] Branch three-phase current and node three-phase voltage data are collected at 10ms sampling intervals using distribution automation terminals (FTU / DTU), with a sampling accuracy of 16 bits. Simultaneously, real-time output of distributed generation (DG), state of charge (SOC) data of energy storage systems (ESS), and power data of critical loads are collected at 50ms sampling intervals using phasor measurement units (PMU), with a time synchronization accuracy ≤1μs. All collected data are transmitted according to the IEC 61850 GOOSE protocol, with a transmission delay ≤5ms, ensuring rapid uploading of fault-related information to the distribution network automation master station. The multi-source data collected in this step is the core basis for fault location in step S2; the completeness and timeliness of the data directly determine the accuracy of fault location and the response speed of fault isolation.

[0032] After completing the above-mentioned synchronous acquisition of multi-source data, the next step is to locate the faulty branch.

[0033] The distribution network automation system collects instantaneous data of branch three-phase current and node three-phase voltage at a sampling frequency of 10kHz through feeder terminal units (FTUs). It uses the fault component method to locate the fault area. The specific judgment conditions are that the current change rate ΔI / IN of the faulty branch is greater than 1.5 and the voltage of the faulty node drops below 0.3UN. The control system ensures that the sectionalizing switches on both sides of the faulty section complete the disconnection action within 3 seconds after the fault occurs, isolating the faulty section and preserving the non-faulty power supply area.

[0034] Based on the transient characteristics of the fault, the current change ΔI and voltage change ΔU are calculated. When the two criteria are met (ΔI / IN>1.5 and ΔU / UN>0.7, where IN is the rated current of the branch and UN is the rated voltage), the branch is determined to be a faulty branch, and the fault area is defined as the set of nodes between the nearest sectionalizing switches on both sides of the branch.

[0035] The data processing center generates a switch control command within 1 second and sends it to the sectional switches on both sides of the faulty section via the FTU. The control switches complete the disconnection action within 3 seconds, isolating the faulty area and preventing the fault from spreading.

[0036] During the fault isolation phase, the FTU sampling accuracy is 16 bits, the data transmission delay is ≤5ms, and the data is transmitted using the IEC 61850 GOOSE protocol; the sectionalizing switch uses an electric operating mechanism with an action time of ≤0.1s to ensure that isolation is completed within 3s after a fault occurs.

[0037] The above step S1, through multi-source data synchronous acquisition and dual-criteria collaborative positioning, clarifies the scope of the non-fault area, defining a feasible operational boundary for network topology reconstruction in step S2. The timeliness and accuracy of fault isolation directly affect the implementation efficiency of subsequent recovery plans.

[0038] Step S1 accurately locates and quickly isolates the fault, creating a safe boundary for the recovery of the non-faulty area. Dual-criteria collaborative localization is based on the transient characteristics of the fault: criterion one is the current surge rate ΔI / IN > 1.5, and criterion two is the voltage drop amplitude ΔU / UN > 0.7. When both criteria are met, the branch is identified as faulty. The master station generates a control command within 1 second, driving the sectionalizing switches on both sides of the faulty section to disconnect within 3 seconds, isolating the faulty area. This step clearly defines the scope of the non-faulty area, defining a feasible boundary for network topology reconstruction in step S2. The timeliness and accuracy of fault isolation directly affect the efficiency of subsequent recovery scheme implementation.

[0039] Step S2: Define switch state variables, construct a connectivity, acyclic, and capacity constraint system, and use heuristic algorithms and linearization optimization to complete topology reconstruction; according to the three-level load priority, achieve priority recovery of critical loads through objective function optimization.

[0040] This embodiment uses three IEEE 33-node power distribution systems connected via SNOP as the test object for illustration. Figure 3 As shown, three IEEE 33-node standard distribution systems are interconnected via a single three-terminal back-to-back voltage source converter (SNOP). Each distribution system contains 33 nodes and 32 branches, including multiple distributed generation (DG) access nodes and tie-off switch nodes. This test system can comprehensively verify the effectiveness of the method of this invention in topology reconfiguration and power supply restoration in scenarios with multiple feeder interconnections and multiple DG access. It also provides clear physical system boundaries for the generation and verification of uncertainties in subsequent steps S4 and S5.

[0041] After completing fault isolation and defining the boundaries of non-faulty areas in step S1, this step establishes a mixed-integer programming model with the objective of minimizing weighted load loss, as follows: Establish a mixed-integer programming model with the objective of minimizing weighted load loss, where decision variables include branch switch state variables. x ij The transmission range of active power PSNOP_k and reactive power QSNOP_k at each terminal of the soft normally open point (SNOP) is determined. The nonlinear power flow constraints are transformed into convex second-order cone constraints through second-order cone transformation. The model is built using the YALMIP toolbox and solved using the CPLEX12.10 solver to obtain the initial network topology and SNOP power operating range.

[0042] It should be noted that defining switch state variables x ij ∈[0,1], x ij =1 indicates a branch. ij When the switch is closed, x ij =0 indicates disconnection, establishing a constraint system:

[0043] in, E This is a set of branches in the non-faulty area. N This represents the number of nodes in the non-faulty region. E S For subset S The set of internal branches is solved using a combination of heuristic rules and linear optimization algorithms. This algorithm prioritizes the shortest path between connecting switches based on Dijkstra's algorithm (search time ≤ 0.5s), and corrects the switch states using the CPLEX 12.10 solver, with a solution time ≤ 7s. The resulting feasible network topology is output.

[0044] Based on the obtained feasible network topology, a weighted objective function is further constructed, and the recovery priority is optimized in stages according to the importance of the load.

[0045] Critical load recovery is classified into three levels according to GB 50052-2020 standard, with the first level having the highest weight. w 1=5, secondary weights w 2=3, Level 3 w 3=1, construct the objective function:

[0046] in, N L This represents the total number of load nodes in the non-faulty area. w i For nodes i Load weighting factor (Level 1 and Level 2 critical loads) w i =5, Level 3 Normal Load w i =1), P L_i For nodes i Rated active load, x L_i Ω is the load recovery state variable. i For nodes i The set of adjacent nodes.

[0047] The objective function of the upper-level optimization model is:

[0048] in, N L This represents the total number of load nodes in the non-faulty area. w i For nodes i Load weighting factor (Level 1 and Level 2 critical loads) w i =5, Level 3 Normal Load w i =1), P L_i For nodes i Rated active load, x L_i This is a load recovery state variable.

[0049] Solving the objective function requires accurate power flow constraints. This invention uses a second-order cone transformation method to handle nonlinear power flow equations.

[0050] The power flow constraints of the upper-level optimization model, after being transformed by a second-order cone, satisfy:

[0051] in, P ij ,Q ij Branch roads ij Transmitting active power and reactive power U i , U j They are nodes i , j The effective value of the line voltage, I ij branch road ij The effective value of the line current is given by , and Rij and Xij are the resistance and reactance of branch ij, respectively.

[0052] After the aforementioned second-order cone relaxation, the original nonlinear AC power flow equality constraint is transformed into a convex second-order cone inequality constraint. This relaxation process introduces quantifiable approximation errors. For example... Figure 4 As shown, under the typical operating conditions of the IEEE 33-bus test system, the branch power deviation caused by relaxation is controlled within a very small range, indicating that the second-order cone relaxation can ensure the solution efficiency while having a negligible impact on the accuracy of power flow calculation, thus verifying the rationality and feasibility of using the second-order cone transformation to handle nonlinear power flow constraints.

[0053] In the upper-level optimization stage, the solution parameters are set to integer precision. e= 10 -6 The maximum number of iterations is 1000, the linear solver used is CPLEX 12.10, the maximum solution time is 30 seconds, and the nodal voltage constraint is 0.95. U N ≤ U i ≤1.05 U N The voltage deviation range for critical load nodes, DG access nodes, and SNOP connection nodes has been narrowed to ±3%.

[0054] Step S3: The model is constructed using the fast decoupling power flow method, which clarifies the constraints of node voltage, branch power flow, DG, energy storage and flexible load in all dimensions, in accordance with relevant national standards.

[0055] Based on the topology output in step S2, a refined model is constructed using the fast decoupling power flow method, and the equations are corrected:

[0056] Where, Δ P Δ Q These represent the active power and reactive power deviations at each node. , To simplify the admittance matrix of the distribution network, Δ d The node voltage phase angle deviation, Δ UFor node voltage amplitude deviation, U= 11.4kV is the rated voltage of a 10kV distribution network. Full-dimensional constraint settings:

[0057] The above constraints include node voltage constraints, branch power flow constraints, distributed generation (DG) constraints, and energy storage constraints. Node voltage constraints are 0.95UN≤Ui≤1.05UN, with tighter limits of ±3% for critical load nodes, DG access nodes, and SNOP connection nodes. Branch power flow constraints ensure that the current in each branch does not exceed its rated current carrying capacity. ESS constraints include upper and lower limits for SOC within the 20% to 90% range, charging and discharging power constraints, and energy conservation constraints. Flexible load constraints include transferable load proportion constraints and interruptible load proportion constraints.

[0058] In addition to the basic constraints mentioned above, detailed operational constraints for distributed generation (DG) and soft normally open point (SNOP) need to be further established.

[0059] The reactive power regulation of the DG meets the inverter capacity constraint:

[0060] Among them, S DG_k_N For the first k The rated capacity of the DG unit, cos =0.9 is the rated power factor of DG. P DG_k It has actually contributed to DG. Q DG_k This is essentially a waste of DG's energy.

[0061] The SNOP adopts a three-terminal back-to-back voltage source converter structure, which satisfies the following capacity constraints and power balance constraints:

[0062] in, S SNOP_N =3MVA is the rated capacity of SNOP. P SNOP_k , Q SNOP_k SNOP No. k The active and reactive power transmission power of the terminal.

[0063] During fault recovery, the energy flow between the terminals of the SNOP exhibits different directions and magnitudes depending on the power supply and demand status of the connected feeders. For example... Figure 2As shown, energy flow exhibits the following typical operating conditions: When a feeder experiences a power deficit, the SNOP absorbs active power from the feeder with sufficient power and injects it into the feeder with the deficit; when the power supply and demand of each feeder are balanced, the power transmitted by each terminal of the SNOP approaches zero, maintaining only necessary reactive power support to stabilize the port voltage. In the multi-objective optimization in step S4 and the full-scenario verification in step S5, the power transmission range of the SNOP must satisfy the above-mentioned capacity constraints and power balance constraints to ensure the feasibility and safety of energy flow under various uncertain scenarios.

[0064] The complete constraint system established in step S3—covering power flow equations, node voltage and branch power flow limits, DG inverter capacity, SNOP power balance, energy storage SOC, and adjustable range of flexible loads—provides a complete feasible domain boundary for the multi-objective optimization in step S4.

[0065] Step S4: Establish a weighted objective function that includes load recovery rate, network loss, and voltage deviation, and use DG output and energy storage power as optimization variables to obtain the optimization scheme through professional tools.

[0066] Within the constraint framework of step S3, step S4 starts from uncertainty modeling and uses a combination of interval robust optimization and multi-objective weighted method to solve for the optimal recovery parameters.

[0067] Based on the fixed network topology of the upper layer output, the uncertainty of the distributed generation (DG) output PDG_k and load power PL_i is described by the interval number. An adaptive weighted optimization model is established with the goal of minimizing the average voltage deviation ΔUavg and the network loss Ploss. The optimal control parameters are determined by solving four boundary scenarios.

[0068] The interval description of the uncertainty parameter is as follows:

[0069] in, P DG_k_pred For the first k The predicted active power output of Taiwan DG P L_i_pred For nodes i The predicted active load value, d DG =30% is the DG output prediction error rate. d L =30% is the load power prediction error rate.

[0070] It should be noted that the output range and load range of the uncertainty parameters are described using interval numbers as follows:

[0071] in, d DG To predict the error rate, four boundary scenarios are generated for worst-case scenario construction and verification. Each scenario is checked to ensure that node voltage, branch power flow, and equipment capacity meet the constraints. Violations are identified as robustness deficiencies. This step builds upon the robustness verification results of step S4, aiming to satisfy all scenario constraints at the lowest cost. Corrections are implemented progressively according to priority, strictly adhering to the constraint system of step S3 and the scenario verification logic of step S4 to ensure the consistency and feasibility of the corrected solution.

[0072] The objective function of the lower-level optimization model is:

[0073] in, E This is the set of branches in the non-faulty area.

[0074] The specific parameter combinations for the four boundary scenarios are as follows: Scene 1: P DG_k = P DG_k_pred (1+ d DG ), P L_i = P L_i_pred (1+ d L ); Scene 2: P DG_k = P DG_k_pred (1+ d DG ), P L_i = P L_i_pred (1- d L ); Scene 3: P DG_k = P DG_k_pred (1- d DG ), P L_i = P L_i_pred (1+ d L ); Scene 4: P DG_k = P DG_k_pred (1- d DG ),P L_i = P L_i_pred (1- d L ); Based on the above-mentioned robust optimization of the interval, in order to further coordinate the trade-off between load recovery rate, network loss and voltage deviation, a multi-objective weighted comprehensive optimization model is constructed.

[0075] It should be noted that the objective function for multi-objective optimization will be constructed next.

[0076]

[0077] in, α= 0.5, β= 0.3, c= 0.2 is the weight. P loss For network loss, Δ U avg The average voltage deviation is given by the following formula:

[0078] After completing the mathematical formulation of the above optimization model, the solution strategy and calculation parameters are determined.

[0079] In the lower-level optimization stage, the Gurobi 9.5.0 solver was used, with a solution accuracy of e=10-6, a maximum solution time of 10s / scene, and the number of threads set to 8. The objective function was standardized, with the voltage deviation standardized in the range of [0%, 10%] and the network loss standardized in the range of [50kW, 100kW].

[0080] The optimal recovery scheme output in step S4 is based on the predicted values ​​of DG output and load power. Considering the inherent fluctuations in new energy output and the unavoidable errors in load forecasting, step S5 will perform uncertainty robustness verification on the above scheme and make progressive corrections in a hierarchical manner to ensure the feasibility of the project.

[0081] Step S5: Describe the uncertainty of DG output and load using interval numbers, generate four extreme scenarios to verify the compliance of constraints, and modify the scheme according to the hierarchical strategy to ensure robustness.

[0082] If the lower-level optimization results in node voltage or branch current exceeding limits, return to the upper-level optimization stage to adjust the switch state or DG access state, and re-execute the above optimization steps. The maximum number of iterations is 5, until a power supply restoration scheme that satisfies all constraints is obtained.

[0083] The specific modifications in this step are implemented in a three-tiered priority manner to eliminate constraint violations at the lowest cost.

[0084] The first level of correction adjusts the reactive power of distributed generation (DG) and the power of energy storage. It prioritizes constraint satisfaction through flexible resource regulation, without incurring load loss costs. The correction formula is as follows:

[0085] Substitute the four boundary scenarios from step S4 and re-execute the power flow calculation and constraint verification. If the constraints are satisfied in all scenarios, the correction is complete; otherwise, proceed to the second level of correction.

[0086] The second level of correction involves tertiary load reduction. When flexible resource allocation is insufficient, tertiary loads are appropriately reduced to prioritize power supply to primary and secondary loads. The reduction formula is as follows:

[0087] in, l n The load reduction ratio is used. Combining the first-level correction results with the load reduction amount, the four-scenario constraint verification is carried out again. If the constraints are met, the correction is complete; otherwise, the third-level correction is initiated.

[0088] The third level of correction involves fine-tuning the network topology. This involves adjusting the topology to change the power flow direction and resolve persistent overload or voltage exceedance issues. After fine-tuning, connectivity and acyclic constraints need to be re-verified, and then four boundary scenarios are substituted to complete constraint verification, ensuring the topology is feasible and that all scenarios satisfy the constraints.

[0089] The adjustment phase of the scheme adopts a layered and progressive adjustment strategy: the first level adjusts the reactive power output of DG, with an adjustment range of ±20kVar; the second level adjusts the SNOP power setting, with active power ±5% and reactive power ±10%; the third level adjusts the switch status, disconnecting the tie switch or DG access switch associated with the over-limit node.

[0090] Correcting the number of iterations k Each iteration lasts ≤ 2 seconds, and the total correction time is ≤ 6 seconds. If the constraints for the entire scenario cannot be satisfied after 3 iterations, a suboptimal solution is output with the objective of "maximizing the number of scenarios that satisfy the constraints," and scenarios that violate the constraints and require manual intervention are marked. After correction, the final device control parameters are output as the basis for executing step S6.

[0091] The final solution, after robust verification and hierarchical correction in step S5, has been confirmed to be feasible across all scenarios within the prediction error range. Step S6 will issue control commands sequentially according to the device response characteristics and perform closed-loop verification of the execution effect.

[0092] Step S6: Issue equipment control commands according to the time sequence, collect operating data synchronously to verify the recovery effect, and complete the recovery if the constraints and performance indicators are met; otherwise, return for correction.

[0093] It should be noted that this step follows the final revised solution of step S5, issuing control commands according to the principle of matching equipment response timing, and simultaneously conducting closed-loop monitoring to ensure that the execution effect is consistent with the optimization goal. The execution timing is designed based on the equipment response characteristics to avoid action conflicts, and the execution commands correspond one-to-one with the final parameters of step S5.

[0094] Before issuing the command, the fault mode switching and control strategy configuration of the soft normally open point (SNOP) must be completed first.

[0095] In fault mode, SNOP employs a droop control strategy, with the control equation being:

[0096] in, P SNOP_k_ref , Q SNOP_k_ref SNOP, respectively k Reference values ​​for active and reactive power at the terminal. d k No. k The voltage phase angle of the terminal, d ref =0 rad is the phase angle reference value. U k For the first k The voltage amplitude of the terminal, U ref = U N =11.4kV is the voltage reference value. K p =10kW / rad, K q =5kVar / kV.

[0097] After SNOP completes the fault mode switch, the control commands for each device are executed sequentially according to the following timing sequence.

[0098] In step S6, control commands are issued in a sequential order to avoid equipment action conflicts. SNOP devices, i.e., three-terminal back-to-back voltage source converters, switch to fault mode. Specific timing: 25-26s: switch operation command issued; 26-27s: DG operation command issued; 27-28s: energy storage charging / discharging command issued; 28-29s: flexible load transfer command issued; 29-30s: real-time data collected via FTU / PMU. Three key indicators are verified: 1) All constraints meet the standards; 2) Primary load recovery rate ≥99%, Secondary load ≥90%, Total ≥95%; 3) Average voltage deviation ≤3%, Network loss ≤ 80% of historical average. If verification passes, fault recovery is complete. If an anomaly occurs, return to step S5 for correction, with a correction time ≤5s, forming a closed-loop control.

[0099] By issuing control commands through the distribution network automation system, the switching of SNOP operation mode, the action of branch switches, and the reduction and regulation of active power output of DG are controlled to achieve safe and rapid power restoration of loads in non-faulty areas.

[0100] The timing sequence for the resumption execution phase is as follows: t=1.0s: SNOP mode switching, switching time ≤20ms; t=1.05s: Issue the redundant switch disconnect command, and the disconnection is completed in t=1.15s; t=1.2s: The command to close the communication switch is issued, and the closing is completed in t=1.28s; t=1.3s: Issue SNOP power reference value command; SNOP power stabilizes at t=1.35s. t=1.36s: DG reactive power adjustment command issued; t=1.38s: DG reactive power output stabilized. t=1.5s: Collect data to verify the recovery effect and confirm that the recovery is complete.

[0101] The proposed method for hierarchical and phased fault recovery in active distribution networks with soft normally open points (SNOPs) and distributed generation (DGs) is based on a three-layer architecture design of "rapid structural recovery - coordinated optimization recovery - robustness enhancement and verification". It simplifies the mixed integer nonlinear programming problem by using second-order cone transformation and combines the active / reactive power control capabilities of DGs. While ensuring rapid power restoration after a fault and priority restoration of important loads, it achieves a balance between load restoration ratio, operational economy and uncertainty robustness, ultimately improving the speed, safety and reliability of fault recovery in active distribution networks.

[0102] After a fault occurs, this invention first collects key data before and after the fault through the feeder terminal unit (FTU) and synchronous phasor measurement unit (PMU). Based on the dual criteria of current mutation rate and voltage drop amplitude, the fault area is located and isolated within 3 seconds. Then, the network topology is quickly reconstructed through an algorithm combining heuristic rules and linear optimization. According to the three-level load priority optimization, the first and second level critical loads are restored first to ensure the core power demand. Next, a refined model is established based on the fast decoupling power flow method. With load recovery rate, network loss and voltage deviation as multiple objectives, the output of distributed generation, charging and discharging of energy storage system and flexible load transfer are coordinated and optimized. Then, considering the prediction uncertainty of DG output and load, the robustness of the scheme is checked through four boundary scenarios. The three-level iterative correction is carried out according to the priority of adjusting equipment output, reducing non-critical loads and fine-tuning network topology. Finally, control commands are issued in sequence according to the equipment response characteristics to complete the recovery execution. The indicators are checked through closed-loop monitoring. If the indicators are not met, the correction is quickly returned.

[0103] The core technological advantages of this method are reflected in the following three aspects: First, it excels in real-time performance. The entire fault recovery decision-making process is completed within 30 seconds: fault isolation takes 3 seconds, topology reconfiguration responds rapidly through a combination of Dijkstra's heuristic search (≤0.5s) and CPLEX linearization correction (≤7s), refined power flow calculation and multi-objective optimization are completed within 30 seconds through Gurobi parallel solution, robust correction takes no more than 6 seconds (3 iterations × 2s / iteration), and execution verification takes approximately 5 seconds. The tight connection and coordinated optimization of each link meet the stringent time sensitivity requirements of distribution network fault recovery.

[0104] Second, it exhibits strong robustness. Interval-based modeling comprehensively covers uncertainties in DG output and load power (error rate ±30%), generating four extreme boundary scenarios for worst-case verification. After three-level hierarchical progressive correction, it maintains low voltage over-limit rates and branch overload rates within DG output fluctuations of ±30% and load error ranges of ±30%, significantly improving robustness compared to traditional deterministic recovery methods. Furthermore, SNOP employs a droop control strategy in fault modes, enabling autonomous response to port voltage and frequency deviations without master station communication, further enhancing the autonomy and robustness of the recovery scheme.

[0105] Third, it is highly practical. It is compatible with various types of active distribution networks with voltage levels of 10kV, node scales of 30 to 100, and DG penetration rates of 10% to 50%. It can be directly integrated into existing distribution automation systems without large-scale hardware modifications. The standardized data acquisition mechanism is based on FTU / DTU and PMU, which can be quickly adapted to existing equipment. The two-layer architecture of mixed integer programming and interval robust optimization takes into account both load recovery integrity and operational economy, and achieves synergistic optimization of high recovery rate of critical loads (Level 1 ≥ 99%, Level 2 ≥ 90%, Total ≥ 95%) with low network loss and small voltage deviation.

[0106] The layered and phased fault recovery method proposed in this invention, in addition to being suitable for the above-mentioned target scenarios, still possesses significant advantages and practical value when extended to the following power system-related scenarios: Firstly, in county-level power distribution networks and industrial park microgrids where the penetration rate of distributed power sources is constantly increasing, its standardized data acquisition mechanism can be quickly adapted to existing equipment. It can achieve rapid fault isolation and priority restoration of critical loads without additional modifications, effectively solving the recovery problems caused by large fluctuations in renewable energy output and dense loads in small and medium-sized power grids.

[0107] Secondly, for integrated energy systems that include energy storage power stations and virtual power plants, their multi-objective optimization model and flexible resource coordination strategy can achieve optimal operation scheduling in scenarios where electricity, heat, and cooling are complementary. At the same time, relying on a robust hierarchical correction mechanism, it can effectively cope with the strong uncertainty of multiple energy outputs and ensure the stability of system operation.

[0108] Third, in cross-regional interconnected distribution networks or isolated power grids in remote areas, the closed-loop monitoring and anomaly backtracking functions of this method can make up for the response delay defects of remote operation and maintenance; the ability to make full-process decisions within 30 seconds and make rapid corrections within no more than 5 seconds can significantly reduce the risk of fault expansion and significantly improve the reliability of power grid supply.

[0109] Fourth, for upgrading and transforming old power distribution networks, this method can significantly reduce transformation costs by eliminating the need for large-scale hardware replacement and allowing direct integration into existing automation systems. At the same time, through refined power flow modeling and national standard constraint systems, it helps old power grids meet the safe operation requirements of new power systems.

[0110] Fifth, in emergency power supply and temporary power supply guarantee scenarios, the method's time-series execution mechanism and high load recovery rate characteristics can quickly restore power supply to core areas; combined with robust design to cope with output fluctuations of temporary power supply equipment, it improves the accuracy and reliability of emergency power supply.

[0111] This invention also discloses a hierarchical and phased fault recovery system for active distribution networks, comprising an interconnected memory and a processor. The memory stores a computer program, which, when run by the processor, executes the steps of the method described above. The system of this invention corresponds to the method described above and also possesses the advantages described therein.

[0112] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0113] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A hierarchical and phased fault recovery method for active distribution networks, characterized in that, Including the following steps: Step S1: When a fault occurs in the distribution network, multi-source operation data such as branch current, node voltage, real-time output of distributed generation (DG), state of charge (SOC) of energy storage system (ESS), and load power are collected synchronously before and after the fault. Based on the dual criteria of current mutation rate and voltage drop amplitude, the faulty branch is located collaboratively, and a switch control command is generated to drive the sectionalizing switch to complete the disconnection action to isolate the faulty section and output the non-faulty area range. Step S2: Using the non-faulty area range output in Step S1 as the input boundary, a mixed integer programming model is established with the goal of minimizing weighted load loss. The branch switch state variables and the power transmission range of each terminal of the soft normally open point SNOP are defined as decision variables. The nonlinear AC power flow constraints are transformed into convex second-order cone constraints through second-order cone transformation. A two-stage solution strategy combining heuristic algorithm and linearization optimization is adopted to output a network topology scheme that satisfies connectivity constraints, acyclic constraints, and capacity constraints, as well as the power operation range of each terminal of the SNOP. Step S3: Based on the network topology scheme output in step S2, establish a power flow calculation model for the distribution network, construct a full-dimensional operation constraint system covering node voltage limits, branch power flow limits, DG inverter capacity, SNOP power balance, ESS charging and discharging, and the adjustable range of flexible loads, and output the complete constraint feasible domain boundary. Step S4: Within the feasible region of full-dimensional constraints output in step S3, the prediction uncertainty of DG output and load power is described using interval numbers. Based on the prediction error rate, multiple extreme boundary scenarios are generated. An interval robust optimization model is established with the goal of minimizing the average voltage deviation and network loss. A multi-objective weighted comprehensive optimization model is constructed by combining the load recovery rate. By solving the model, the active / reactive output setpoints of each DG, the charging and discharging power of each ESS, the flexible load transfer / reduction amount, and the terminal power setpoints of each SNOP are obtained. The output is the optimal recovery scheme. Step S5: Based on the optimal recovery scheme output in step S4, check whether each scenario in each extreme boundary scenario meets the constraints in the full-dimensional operation constraint system constructed in step S3. If any scenario violates any constraint, it is determined that there is a robustness defect and it is corrected in a three-level priority manner. The correction process is iteratively executed until all scenarios meet all constraints or the preset iteration limit is reached, and the final device control parameters are output. Step S6: Based on the final equipment control parameters output in step S5, design a conflict-free timing execution process according to the equipment response characteristics, and sequentially issue switch action commands, DG output commands, ESS charging and discharging commands, and flexible load transfer commands; simultaneously collect real-time operating data of the distribution network after restoration, and verify whether the restoration effect meets the preset constraint compliance rate, load restoration rate, and power quality indicators; if all verifications pass, the entire fault restoration process is completed; if any indicator fails to meet the standard, return to step S5 for correction, forming a closed-loop control.

2. The hierarchical and phased fault recovery method for active distribution networks according to claim 1, characterized in that, In step S1, the fault location branch based on the dual criteria of current mutation rate and voltage drop amplitude specifically includes: Calculate the transient current change ΔI and voltage change ΔU during the fault. When the current change rate ΔI / IN>1.5 and the voltage drop amplitude ΔU / UN>0.7 are simultaneously satisfied, the corresponding branch is determined to be a faulty branch, where IN is the rated current of the branch and UN is the nominal rated voltage of the 10kV distribution network.

3. The hierarchical and phased fault recovery method for active distribution networks according to claim 1, characterized in that, In step S2, the mixed-integer programming model minimizes the weighted load loss. The objective function is: in, N L This represents the total number of load nodes in the non-faulty area. w i For nodes i Load weighting factor P L_i For nodes i Rated active load, x L_i This is a load recovery state variable.

4. The hierarchical and phased fault recovery method for active distribution networks according to claim 1, characterized in that, In step S3, the power flow calculation model of the distribution network is constructed using the Fast Decoupled Power Flow (FDLF) method, and its modified equation is: Where, Δ P Δ Q These represent the active power and reactive power deviations at each node. , To simplify the admittance matrix of the distribution network, Δ δ The node voltage phase angle deviation, Δ U For node voltage amplitude deviation, U This is the rated voltage for a 10kV distribution network.

5. The hierarchical and phased fault recovery method for active distribution networks according to claim 4, characterized in that, In step S3, the inverter capacity constraint of DG is: Among them, S DG_k_N For the first k The rated capacity of the DG unit, cos The rated power factor of DG, P DG_k They actually contributed to DG. Q DG_k For DG, this is actually a waste of effort; , For the first k The minimum and maximum capacity of the DG (Distributed Generation) system.

6. The hierarchical and phased fault recovery method for active distribution networks according to any one of claims 1-5, characterized in that, In step S4, the use of interval numbers to describe the prediction uncertainty of DG output and load power specifically includes: in, P DG_k It has actually contributed to DG; P DG_k_pred For the first k The predicted active power output of Taiwan DG P L_i For nodes i Active load; P L_i_pred For nodes i The predicted active load value, δ DG For DG output prediction error rate, δ L This represents the load power prediction error rate.

7. The hierarchical and phased fault recovery method for active distribution networks according to claim 6, characterized in that, In step S4, the multiple extreme boundary scenarios are four boundary scenarios, each corresponding to the maximum and minimum combination of DG output and load power in the uncertain parameter space: Scene 1: P DG_k = P DG_k_pred (1+ δ DG ), P L_i = P L_i_pred (1+ δ L ); Scene 2: P DG_k = P DG_k_pred (1+ δ DG ), P L_i = P L_i_pred (1- δ L ); Scene 3: P DG_k = P DG_k_pred (1- δ DG ), P L_i = P L_i_pred (1+ δ L ); Scene 4: P DG_k = P DG_k_pred (1- δ DG ), P L_i = P L_i_pred (1- δ L ); The interval robust optimization model takes the worst case among the four boundary scenarios as the robust optimization result to ensure that the solution can still meet the operational constraints under the most unfavorable conditions.

8. The hierarchical and phased fault recovery method for active distribution networks according to any one of claims 1-5, characterized in that, In step S4, the objective function of the multi-objective weighted comprehensive optimization model is... for: in, α, β, γ As weight, P L_i For nodes i Active load, For the maximum network loss, Δ U avg This represents the average voltage deviation. P loss The total active power loss of the system; x L_i This is a load recovery state variable.

9. The hierarchical and phased fault recovery method for active distribution networks according to any one of claims 1-5, characterized in that, In step S5, the three-level priority progressive correction is performed sequentially in order of increasing cost: The first level of correction adjusts the reactive power output of DG and the charging and discharging power of each ESS, without involving load reduction. After the adjustment is completed, the power flow calculation and constraint verification are performed again by substituting into each boundary scenario. The second-level correction only reduces the third-level load by a ratio of n, prioritizing the uninterrupted power supply to the first and second-level loads, and re-verifies the constraints based on the results of the first-level correction. The third-level correction achieves topology fine-tuning by adjusting the switching status of the tie switch or changing the location of the DG access bus. After the adjustment, connectivity constraints and acyclic constraints need to be re-verified before the full-scenario constraint verification is completed.

10. A hierarchical and phased fault recovery system for active distribution networks, comprising an interconnected memory and a processor, wherein the memory stores a computer program, characterized in that... The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-9.

Images