A Two-Stage Adaptive Power Flow Convergence Method for Power Systems with Time-Sequenced Faults

By employing a two-stage adaptive power flow convergence method, an initial solution is obtained under time-series faults using linearized power flow calculation and optimization models. This solves the divergence problem of power flow calculation under time-series faults and achieves efficient power flow convergence.

CN123073809APending Publication Date: 2026-11-10NORTH CHINA ELECTRIC POWER UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611239600.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-17
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies are prone to divergence in power flow calculations when faced with time-series faults, making it difficult to quickly converge to a feasible solution under complex disturbances, which leads to the interruption of time-series simulation.

Method used

A two-stage adaptive power flow convergence method is adopted. First, the initial solution of the system state variables is obtained through linearized power flow calculation. If no feasible solution is obtained, some constraints are relaxed and optimization is performed again to obtain the benchmark solution. The benchmark solution is used as a guide for optimization to ensure the convergence of nonlinear power flow calculation.

Benefits of technology

It provides high-quality initial running points that are easy to converge, avoids solution divergence under timing faults, ensures the convergence of nonlinear power flow calculation under extreme conditions, and achieves power flow calculation with high convergence rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123073809A_ABST
    Figure CN123073809A_ABST
Patent Text Reader

Abstract

This invention provides a two-stage adaptive power flow convergence method for power systems facing time-series faults, comprising: obtaining initial solutions for system state variables through linearized power flow calculation based on operating data; constructing an optimization model to provide an initial point for nonlinear power flow calculation based on the initial solution, and using the optimization model to perform feasible solution optimization under set power system physical operating constraints; using the optimized feasible solution as the initial point for nonlinear power flow calculation; if no feasible solution is obtained through optimization, relaxing some constraints in the system physical operating constraints, and re-optimizing under the relaxed system physical operating constraints to obtain a baseline solution; canceling the relaxation of some constraints, and using the baseline solution as a guide for optimization, and using the optimization result as the initial point for nonlinear power flow calculation to achieve convergence of nonlinear power flow calculation. This method overcomes the solution divergence problem caused by the shrinking feasible region under time-series faults, ensuring the convergence of nonlinear power flow calculation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power system steady-state analysis technology, and in particular to a two-stage adaptive power flow convergence method for power systems with time-series faults. Background Technology

[0002] Power flow calculation is a core foundational program for power system planning, design, operation mode arrangement, and security and stability analysis. With the continuous expansion of modern power grids, the integration of high proportions of renewable energy, and the frequent occurrence of extreme operating conditions, the operating environment faced by power systems is becoming increasingly complex. Especially when the power grid suffers severe time-series faults (such as N-1-1-1 faults), the grid topology undergoes drastic changes, and the steady-state physical boundaries of the system are often severely disrupted. This makes conventional nonlinear power flow calculation programs, such as the Newton-Raphson method, prone to encountering singular or unsolvable Jacobian matrices, leading to interruptions in time-series simulations. Therefore, it is urgent to study adaptive convergence adjustment strategies for power flow under time-series faults to ensure that power flow calculations can quickly converge to feasible solutions under complex disturbances, laying the foundation for time-series simulations of new power systems under extreme events.

[0003] Currently, there has been extensive research in academia and engineering on adjustment techniques for non-convergence in power flow calculations, which can be mainly divided into the following three technical approaches: The first approach is a mathematical optimization method based on underlying algorithms. Existing research focuses on improving the underlying numerical stability of nonlinear equation systems and the Jacobian matrix. This is achieved by introducing adaptive Levenberg-Marquardt damping factors, Anderson acceleration algorithms, or analyzing the impact of negative reactance on the conjugate gradient method to suppress iterative oscillations and expand the convergence region; or by introducing power droop nodes to modify the underlying Jacobian matrix of the system; or by improving the fully embedded solver through constant terms and prior nodes; or by using robust second-order cone optimal power flow and specific mathematical programming models to handle uncertainties and convergence correction problems in power grid operation. Furthermore, to address the problem of initial value sensitivity and easy divergence in nonlinear power flow solutions, existing research has also proposed initial value improvement methods based on intermediate residual diagnosis, zero-injection startup methods for distribution networks, and zero-injection initialization numerical methods for ill-conditioned power flows.

[0004] The second approach is based on heuristic algorithms. Existing research proposes that when there is no solution for the power flow, some load nodes are forcibly converted into generator nodes, and the reactive power deficit index of the system is quantified. Then, the shortest path algorithm, alternating bisection method, or dynamic programming is used to find the optimal reactive power compensation point in the local area for switching; or reactive power and voltage are injected iteratively; or the active and reactive power imbalance characteristics in the intermediate iteration process of power flow calculation are used to identify and adjust the key network variables that cause non-convergence in a data-driven manner.

[0005] The third approach is based on artificial intelligence and deep reinforcement learning. Existing research encapsulates the power flow calculation process into a reinforcement learning environment using a hybrid action space and introduces physically prior adjustment strategies; or improves deep deterministic policy gradient algorithms and lightweight neural networks to accelerate agent interaction; or, based on deep Q-networks, introduces a decoupled adjustment approach of "adjusting active power first, then reactive power," using electrical indicators such as impedance ratio to pre-identify voltage weak areas to narrow the search space; or combines node type conversion with deep reinforcement learning, designing the agent's action boundaries by quantifying reactive power non-convergence indicators to achieve reactive power adjustment. These methods currently mainly target system static ground state recovery or single N-1 fault adjustment.

[0006] While the aforementioned technologies have achieved some success in single-fault or conventional power flow divergence scenarios, they are difficult to apply directly when dealing with time-series faults (such as N-1-1-1). Existing technologies suffer from the following significant drawbacks: Conventional power flow calculation methods are highly dependent on initial points. When a system suffers from timing faults, traditional flat-start initial values ​​often lead to direct divergence in the solution. Existing technologies struggle to provide high-quality initial operating points for nonlinear solvers that are easy to converge under extreme conditions where the physical feasible region is compressed, resulting in low power flow convergence rates in timing fault scenarios. Most existing heuristic algorithms rely on greedy search, and single adjustment methods cannot handle the optimal coordination of active and reactive resources across the entire network when facing scenarios with overlapping timing faults. Furthermore, no effective method has yet emerged that can fully support convergence recovery from timing faults, limiting the ability of large power grids to conduct safety simulations for timing faults in practical engineering. Summary of the Invention

[0007] This invention provides a two-stage adaptive power flow convergence method for power systems with time-series faults, which solves the problem of solution divergence caused by the shrinking feasible region under time-series faults in the prior art.

[0008] On the one hand, this invention provides a two-stage adaptive power flow convergence method for power systems oriented towards time-series faults, comprising: Acquire operational data under power system timing faults; Based on the aforementioned operating data, the initial solution for the system state variables is obtained through linearized power flow calculation; Based on the initial solution, an optimization model is constructed to provide an initial point for nonlinear power flow calculation. The optimization model is then used to find feasible solutions under the set physical operating constraints of the power system. The feasible solution obtained through optimization is used as the initial point for the nonlinear power flow calculation. If no feasible solution is found through optimization, some of the constraints in the physical operation constraints of the system are relaxed, and the optimization model is used to re-optimize under the relaxed physical operation constraints of the system to obtain the baseline solution. The relaxation of some constraints is cancelled, and the optimization is guided by the benchmark solution. The optimization result is used as the initial point for the nonlinear power flow calculation so that the nonlinear power flow calculation converges.

[0009] Optionally, the operating data includes grid topology parameters and node injected power, and the initial solution of the system state variables is obtained through linearized power flow calculation based on the operating data, including: Based on the power grid topology parameters, construct the node admittance matrix; Based on the node admittance matrix and the node injected power, establish the node injected power equation; By using the square term of the node voltage as a variable, the node injected power equation is linearized to obtain a power flow model expanded with the square term of the voltage. Based on the power flow model, the initial solution for the system state variables is obtained.

[0010] Optionally, the linearization of the node injection power equation using the squared term of the node voltage as a variable includes: The trigonometric function terms in the node injection power equation are approximated to obtain the approximate node injection power equation. By replacing the voltage in the approximate nodal injection power equation with the voltage squared term, a linearized nodal injection power equation expressed in terms of voltage squared terms is obtained.

[0011] Optionally, the nodal injection power equation before approximation is expressed as: ; ; The trigonometric function terms are approximated and expressed as follows: , ; The approximate nodal injection power equation is expressed as: ; ; By substituting the voltage squared term into the approximate nodal injection power equation, we obtain a linearized nodal injection power equation expressed in terms of the voltage squared term, which is: ; ; ; in, For nodes The active power injected; For nodes The reactive power injected; For nodes in the node admittance matrix With nodes The electrical conductivity components between; For nodes in the node admittance matrix With nodes j The susceptance component between; For nodes The voltage amplitude; For nodes j The voltage amplitude; For nodes With nodes j The voltage phase angle difference between them; For the node Flow to Node j The branch road has contributed to the trend; For the node Flow to Node j The branch road is ineffective in tidal current; This represents the total number of nodes in the power system.

[0012] Optionally, the step of using the benchmark solution as a guide for optimization, and using the optimization result as the initial point for the nonlinear power flow calculation, includes: A reactive power approximation term is introduced into the objective function of the optimization model; wherein, the reactive power approximation term is used to constrain the reactive power distribution of the system to approximate the benchmark solution; The solution result of the objective function after introducing the reactive power approximation term is used as the initial point for the nonlinear power flow calculation.

[0013] Optionally, the reactive power approximation term is represented as: ; in, This represents the actual reactive load. This is the baseline solution; The weighting coefficients are for the reactive approximation term.

[0014] Optionally, the step of constructing an optimization model based on the initial solution to provide an initial point for nonlinear power flow calculation, and using the optimization model to find feasible solutions under the set physical operating constraints of the power system, includes: Based on the initial solution, an optimization model is set with the objective function of minimizing the overall cost; Under the preset physical operation constraints of the power system, the objective function is solved to obtain a feasible solution.

[0015] Optionally, the objective function with the reactive power approximation term can be expressed as: ; in, As a comprehensive cost; This represents the difference between the node voltage and its per-unit value of 1.0. The node voltage phase angle; Estimated value for reactive power injection; The unit provides active power for actual optimization; Contributed to the original scheduling plan; The actual force output borne by the balancing node; For balanced node indexes.

[0016] Optionally, the operational data includes line transmission capacity constraints, and relaxing some of the constraints in the system physical operational constraints includes: The line transmission capacity constraint boundary in the physical operation constraints is enlarged by a preset factor.

[0017] Optionally, the constraint on the line transmission capacity is linearized using a polygonal linear approximation method; The linearization process of the polygon linear approximation method includes: Based on the active and reactive power flow of the branch, the apparent power constraints of the branch are projected to multiple directions in space using trigonometric function projection coefficients, so as to transform the non-convex apparent power square form constraint into a set of linear polyhedral boundary constraints.

[0018] This invention provides a two-stage adaptive power flow convergence method for power systems facing time-series faults. It obtains initial solutions for system state variables through linearized power flow calculations, providing a high-quality initial operating point for the nonlinear solver that is easy to converge, thus avoiding direct divergence of initial values ​​under time-series faults. If no feasible solution is found through optimization, some constraints are relaxed and optimization is performed again to obtain a benchmark solution, thereby obtaining the approximate distribution trend of the system state within the expanded feasible region. The relaxation is then canceled, and optimization is performed again guided by the benchmark solution. The optimization result is used as the initial point for nonlinear power flow calculations. Using a defined benchmark to guide the system state gradually towards a feasible solution, it overcomes the divergence problem caused by the shrinking feasible region under time-series faults, ensuring the convergence of nonlinear power flow calculations. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1This is a flowchart illustrating the two-stage adaptive power flow convergence method for power systems oriented towards time-series faults provided in this embodiment of the invention. Figure 2 This is a logical architecture diagram of the two-stage power flow adaptive method provided in the embodiments of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] Figure 1 This is a flowchart illustrating the two-stage adaptive power flow convergence method for power systems oriented towards time-series faults provided in this embodiment of the invention.

[0023] See Figure 1 The two-stage adaptive power flow convergence method for power systems facing time-sequenced faults may include the following steps 101 to 106.

[0024] Step 101: Obtain operating data under power system timing faults.

[0025] In this step, the operational data may include at least one of the following: power grid topology parameters, such as the basic data required to construct the node admittance matrix, including node and branch connection relationships and line resistance, reactance, susceptance, etc.; node injected power, including the injected active and reactive power of each node, generator output data, and load power data; generator operating parameters, such as the upper and lower limits of active power output, reactive power output range, dispatch plan output value, and active power output undertaken by balancing nodes; load parameters, including the active and reactive load of each node and the ideal reactive load target value; line transmission capacity, i.e., the active and reactive power transmission limits of each branch; node voltage and phase angle parameters, including node voltage amplitude, phase angle, and its safe operating range; and time-series fault sequence information, including fault type, occurrence time sequence, and the resulting topology changes.

[0026] Step 102: Based on the running data, obtain the initial solution of the system state variables through linearized power flow calculation.

[0027] The operational data may include grid topology parameters and node injected power. Step 102 above may include: Step 1: Construct the node admittance matrix based on the power grid topology parameters; Step 2: Based on the nodal admittance matrix and nodal injected power, establish the nodal injected power equation; Step 3: Linearize the node injected power equation using the nodal voltage square term as a variable to obtain the power flow model expanded with the voltage square term; Step 4: Based on the power flow model, obtain the initial solution of the system state variables.

[0028] For step three above, linearizing the nodal injected power equation using the squared term of the nodal voltage as a variable can include: The trigonometric function terms in the nodal injection power equation are approximated to obtain the approximate nodal injection power equation. By replacing the voltage in the approximate nodal injection power equation with the voltage squared term, we obtain the linearized nodal injection power equation expressed in terms of the voltage squared term.

[0029] Through the above steps, the node injected power equation is linearized using the square term of the node voltage as a variable, resulting in a power flow model expanded with the square term of the voltage. This transforms the nonlinear power flow equation into a linearized form that is easy to solve, yielding a high-precision initial solution. This avoids the divergence problem caused by improper initial points when directly using nonlinear power flow calculations in time-series fault scenarios.

[0030] Based on the fundamental Newton-Laurel method, the nodal injection power equations before approximation include: The nodal active power injection equation before approximation is expressed as formula (1): (1); The nodal reactive power injection equation before approximation is expressed as formula (2): (2); The branch active power flow equations before approximation are expressed as formula (3): (3); The branch reactive power flow equation before approximation is expressed as formula (4): (4); in, For nodes The active power injected; For nodes The reactive power injected; For nodes in the node admittance matrix With nodes The electrical conductivity components between; For nodes in the node admittance matrix With nodes j The susceptance component between; For nodes The voltage amplitude; For nodes j The voltage amplitude; For nodes With nodes j The voltage phase angle difference between them; For the node Flow to Node j The branch road has contributed to the trend; For the node Flow to Node j The branch road is ineffective in tidal current; This represents the total number of nodes in the power system.

[0031] The voltage amplitude is approximated as 1 pu; according to the Taylor expansion, the trigonometric function terms are approximated and expressed as formulas (5) and (6): (5); (6); The approximate nodal injection power equations are expressed as equations (7) and (8): (7); (8); By substituting the voltage squared term into the approximate nodal injection power equation, we obtain a linearized nodal injection power equation expressed in terms of the voltage squared term, which is: The term is approximately equal to ,Will The term is approximately equal to The lower-order approximate expression for the voltage product term is given by formula (9): (9); The branch active power flow linearization equation expanded with voltage square terms is expressed as formula (10): (10); The branch reactive power flow linearization equation expanded with voltage square terms is expressed as formula (11). (11); The linearized power flow matrix model expanded with voltage square terms is expressed as equations (12) to (16): (12); in, Inject power vectors into the active power of nodes; Inject power vectors into the reactive power of nodes; The node voltage phase angle vector; This represents the node voltage magnitude vector; It is a unit vector; This is the sensitivity matrix of active power to the square term of node voltage; This is the sensitivity matrix of active power to the square term of node voltage; This is the sensitivity matrix of reactive power to the node voltage phase angle; This is the sensitivity matrix of reactive power to the square of the node voltage. This is the active power loss vector; This is the reactive power loss vector.

[0032] (13); (14); (15); (16); in, It is the set of all nodes in the power system; Step 103: Based on the initial solution, construct an optimization model to provide an initial point for nonlinear power flow calculation, and use the optimization model to find feasible solutions under the set physical operating constraints of the power system.

[0033] Step 103 above includes: Based on the initial solution, an optimization model is set with the objective function of minimizing the overall cost; Under the pre-defined physical operation constraints of the power system, the objective function is solved to obtain a feasible solution.

[0034] Step 104: Use the feasible solution obtained through optimization as the initial point for nonlinear power flow calculation.

[0035] Step 105: If no feasible solution is found through optimization, relax some of the constraints in the physical operation constraints of the system, and use the optimization model to re-optimize under the relaxed physical operation constraints of the system to obtain the baseline solution.

[0036] Operational data may include line transmission capacity constraints, and may relax some of the constraints in the system's physical operational constraints, including: The line transmission capacity constraint boundary in the physical operation constraints is enlarged by a preset multiple.

[0037] The preset multiplier is usually 5x.

[0038] Step 106: Remove the relaxation of some constraints and use the benchmark solution as a guide to perform optimization. Use the optimization result as the initial point for nonlinear power flow calculation so that the nonlinear power flow calculation converges.

[0039] In step 106 above, optimization is performed using the benchmark solution as a guide, and the optimization result is used as the initial point for nonlinear power flow calculation. This may include: Step 1: Introduce a reactive power approximation term into the objective function of the optimization model; the reactive power approximation term is used to constrain the reactive power distribution of the system to approximate the reference solution; The reactive approximation term is expressed as formula (17): (17); in, This represents the actual reactive load. This is the baseline solution. These are the weighting coefficients for the reactive power approximation term; Step 2: Use the solution of the objective function after introducing the reactive power approximation term as the initial point for nonlinear power flow calculation.

[0040] The objective function with the reactive power approximation term is expressed as formula (18): (18); in, As a comprehensive cost; This represents the difference between the node voltage and its per-unit value of 1.0. The node voltage phase angle; Estimated value for reactive power injection; The unit provides active power for actual optimization; It has contributed to the original scheduling plan.

[0041] In this embodiment, the initial solution of the system state variables is obtained through linearized power flow calculation, providing a high-quality initial running point for the nonlinear solver that is easy to converge, thus avoiding the direct divergence of the initial value under timing faults. If no feasible solution is obtained through optimization, some constraints are relaxed and optimization is performed again to obtain the benchmark solution, thereby obtaining the approximate distribution trend of the system state within the expanded feasible region. The relaxation is canceled and optimization is performed with the benchmark solution as the guide. The optimization result is used as the initial point for nonlinear power flow calculation. The determined benchmark guides the system state to gradually transition to a feasible solution, overcoming the problem of solution divergence caused by the shrinking feasible region under timing faults, and ensuring the convergence of nonlinear power flow calculation.

[0042] In one embodiment of this specification, the constraint on line transmission capacity is linearized using a polygonal linear approximation method; The linearization process for polygon linear approximation includes: Based on the active and reactive power flow of the branch, the apparent power constraints of the branch are projected to multiple directions in space using trigonometric function projection coefficients, so as to transform the non-convex apparent power square form constraint into a set of linear polyhedral boundary constraints.

[0043] In this embodiment, "multiple" generally refers to 12. When the apparent power constraint of the line power flow is non-convex, a 12-sided linearization method is introduced to transform the nonlinear apparent power constraint into a linear constraint set, ensuring the convergence and computational efficiency of convex optimization.

[0044] In one embodiment of this specification, the preset physical operating constraints of the power system include at least one of the following: Kirchhoff power balance constraints, node voltage safety range constraints, node phase angle safety range constraints, generator active power output limit constraints, generator reactive power output range constraints, and line transmission capacity constraints.

[0045] The following section describes the multi-objective power flow dynamic optimization adjustment of this invention.

[0046] The optimization model can employ different control strategies based on the current operating state of the system to find the global minimum cost. Global minimum cost This can include the cost of the basic strategy. and the cost of guidance strategies ; .

[0047] Basic strategy cost It can include the following: When the feasible region of the system is sufficient, the model will constrain the system state to be as close as possible to the ideal operating point before the failure. It includes the following three sub-items: (1) The state offset is expressed as formula (19): (19); Variable definition: This represents the difference between the node voltage and its per-unit value of 1.0. The node voltage phase angle; The estimated value for reactive power injection.

[0048] Physical significance: Prioritize ensuring the grid voltage support level, prevent voltage collapse, and suppress drastic fluctuations during power flow redistribution.

[0049] (2) The generator output is expressed by formula (20): (20); To contribute to the actual optimization of the generating units; Contributed to the original scheduling plan; The actual force output borne by the balancing node; For balanced node indexes; Physical meaning: PV nodes should not change the original active power output plan as much as possible; the unbalance of the system should be supported by the balancing nodes first.

[0050] (3) The load power factor is expressed by formula (21): (twenty one); This represents the actual reactive load. This represents the ideal reactive load.

[0051] Physical meaning: To ensure that the equivalent impedance of the load does not undergo severe distortion.

[0052] Guiding Strategy Cost include: When the physical feasible region of the system shrinks significantly or disappears, the model expands the original physical boundary, obtaining a general distribution trend of the system state within the expanded feasible region. This guides the system state to be smoothly projected along a homologous trajectory to the true topological boundary. It includes the following sub-items: The reactive power approximation is expressed by formula (22): (twenty two); Variable definition: The optimal distribution, or benchmark solution, is obtained when the system relaxes the line transmission limit by a preset multiple (such as 5 times).

[0053] Physical meaning: When the boundary shrinks back to the true topology, the forced reactive power variable iteration direction is towards the optimal baseline solution. By moving closer together, the problem of numerical solution divergence caused by tightening constraints is avoided to some extent.

[0054] The constraints include: Kirchhoff power balance and network coupling constraints: (1) Node injection equation: The slack node is represented by equation (23): (twenty three); The generator node (PV) is represented by formula (24): (twenty four); in, The active power injection amount for the balancing node; The actual active power output undertaken by the balancing node; To balance the active power load of the original scheduling plan at the node; This refers to the active power injection at the generator node (PV node). The unit provides active power for actual optimization; This represents the active load of the original scheduling plan.

[0055] (2) Network Equivalent Coupling: Active Power Injection at Nodes With reactive power injection It must be exactly equal to the calculated value of the power flow model (i.e.) wait). This is the active load vector of the node.

[0056] The node voltage and phase angle safety range constraints include: (1) Balance node boundary: The phase angle is strictly limited to the target value, and the voltage amplitude is equal to the target set value, serving as the absolute phase and voltage reference for the entire network.

[0057] (2) Safe intervals between PV and PQ nodes: The voltage difference limit of PV nodes is expressed by formula (25): (25); in, It is a voltage lower limit relaxation variable; This is a relaxation variable for the upper voltage limit; The constraint on the phase angle limit is expressed by formula (26): (26); The phase angle is a relaxation variable; The voltage difference limit at the PQ node is expressed by formula (27): (27); The phase angle limit is expressed as formula (28): (28); Generator output limits and ramping constraints include: The limit of active power output is expressed by formula (29): (29).

[0058] The range of reactive power output is expressed by formula (30): (30); in, The unit provides active power for actual optimization; This is the per-unit value representing the upper limit of the unit's active power output; and These are the lower limit and upper limit of generator reactive power output, respectively; Line transmission constraints include: Dodecagonal linear approximation model: To avoid the non-convex square form of apparent power, this invention extracts branch active power in the optimization space. and no merit Using trigonometric projection coefficients (such as...) Project it onto 12 directions in space. These 12 directions can include orthogonal directions, 30° angle directions, and 60° angle directions, as shown in formula (31): (31); in, The trend of contributing to the development of branch roads; The branch road was not contributing to the current. This is the per-unit value of the maximum power that the line can transmit; These are slack variables for line constraints; In summary, this invention offers the following technical advantages: It achieves high convergence rates under timing faults, effectively overcoming the divergence problem caused by the shrinking feasible region. This invention first utilizes a one-stage linear simplification followed by a two-stage optimization adjustment to provide a high-quality initial operating point. When the physical feasible region is severely limited by the fault, a homotopy projection mechanism is introduced: first, the line capacity constraint is magnified several times to solve for the optimal power distribution; then, when restoring the true capacity constraint, the reactive power distribution is forced to approximate this benchmark by controlling the model. This mechanism replaces blind iteration with a defined benchmark, ensuring that the system still achieves good convergence even under timing faults.

[0059] High efficiency and robustness of global optimization. This invention introduces a 12-sided linearization method, transforming the highly non-convex apparent power constraint into a linear polyhedral boundary constraint set. This improvement reduces the extremely difficult nonlinear programming problem to a convex optimization model with extremely fast solution speed, achieving optimal coordinated allocation of active and reactive power across the entire network globally, and providing reliable algorithmic support for rapid testing of subsequent time-series faults.

[0060] Figure 2 This is a logical architecture diagram of the two-stage power flow adaptive method provided in an embodiment of the present invention. Figure 2 As shown, step 1: data and parameter preparation, by inputting basic parameters, constructing the admittance matrix and calculating the sensitivity matrix, and outputting preset parameters, lays the foundation for subsequent calculations.

[0061] Step 2: Convex Optimization and Feasibility Check. In this stage, a convex optimization model is constructed and a solution is sought. The system will determine if a feasible solution exists; if so, the initial point is obtained directly and the process ends. The convex optimization model is the optimization model used to provide the initial point for nonlinear power flow calculations.

[0062] If no feasible solution exists (corresponding to a situation where a severe fault shrinks the physical feasible region), then step 3: homotopy projection and restoration is executed. This step first temporarily expands the constraints to solve for the baseline solution in the relaxed region, then restores the expanded constraints to the true physical boundary, and uses the homotopy projection mechanism to pull the state back. Finally, the projection obtains the initial point that satisfies the true physical constraints, thereby ensuring that the power flow calculation can still obtain effective convergent initial values ​​under extreme conditions.

[0063] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units.

[0064] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A two-stage adaptive power flow convergence method for power systems with time-series faults, characterized in that, include: Acquire operational data under power system timing faults; Based on the aforementioned operating data, the initial solution for the system state variables is obtained through linearized power flow calculation; Based on the initial solution, an optimization model is constructed to provide an initial point for nonlinear power flow calculation. The optimization model is then used to find feasible solutions under the set physical operating constraints of the power system. The feasible solution obtained through optimization is used as the initial point for the nonlinear power flow calculation. If no feasible solution is found through optimization, some of the constraints in the physical operation constraints of the system are relaxed, and the optimization model is used to re-optimize under the relaxed physical operation constraints of the system to obtain the baseline solution. The relaxation of some constraints is cancelled, and the optimization is guided by the benchmark solution. The optimization result is used as the initial point for the nonlinear power flow calculation so that the nonlinear power flow calculation converges.

2. The method according to claim 1, characterized in that, The operational data includes grid topology parameters and node injected power. The initial solution for the system state variables is obtained through linearized power flow calculation based on the operational data, including: Based on the power grid topology parameters, construct the node admittance matrix; Based on the node admittance matrix and the node injected power, establish the node injected power equation; By using the square term of the node voltage as a variable, the node injected power equation is linearized to obtain a power flow model expanded with the square term of the voltage. Based on the power flow model, the initial solution for the system state variables is obtained.

3. The method according to claim 2, characterized in that, The linearization of the node injection power equation using the squared term of the node voltage as a variable includes: The trigonometric function terms in the node injection power equation are approximated to obtain the approximate node injection power equation. By replacing the voltage in the approximate nodal injection power equation with the voltage squared term, a linearized nodal injection power equation expressed in terms of voltage squared terms is obtained.

4. The method according to claim 3, characterized in that, The nodal injection power equation before approximation is expressed as: ; ; The trigonometric function terms are approximated and expressed as follows: , ; The approximate nodal injection power equation is expressed as: ; ; By substituting the voltage squared term into the approximate nodal injection power equation, we obtain a linearized nodal injection power equation expressed in terms of the voltage squared term, which is: ; ; ; in, For nodes The active power injected; For nodes reactive power injection; For nodes in the node admittance matrix With nodes The electrical conductivity components between; For nodes in the node admittance matrix With nodes j The susceptance component between; For nodes The voltage amplitude; For nodes j The voltage amplitude; For nodes With nodes j The voltage phase angle difference between them; For the node Flow to Node j The branch road has contributed to the trend; For the node Flow to Node j The branch road is ineffective in tidal current; This represents the total number of nodes in the power system.

5. The method according to claim 1, characterized in that, The optimization process guided by the benchmark solution, using the optimization result as the initial point for the nonlinear power flow calculation, includes: A reactive power approximation term is introduced into the objective function of the optimization model; wherein, the reactive power approximation term is used to constrain the reactive power distribution of the system to approximate the benchmark solution; The solution result of the objective function after introducing the reactive power approximation term is used as the initial point for the nonlinear power flow calculation.

6. The method according to claim 5, characterized in that, The reactive power approximation term is represented as follows: ; in, This represents the actual reactive load. This is the baseline solution; The weighting coefficients are for the reactive approximation term.

7. The method according to claim 5, characterized in that, The step of constructing an optimization model based on the initial solution to provide an initial point for nonlinear power flow calculation, and using the optimization model to find feasible solutions under the set physical operating constraints of the power system, includes: Based on the initial solution, an optimization model is set with the objective function of minimizing the overall cost; Under the preset physical operation constraints of the power system, the objective function is solved to obtain a feasible solution.

8. The method according to claim 7, characterized in that, The objective function with the reactive power approximation term is expressed as: ; in, As a comprehensive cost; This represents the difference between the node voltage and its per-unit value of 1.

0. The node voltage phase angle; Estimated value for reactive power injection; The unit provides active power for actual optimization; Contributed to the original scheduling plan; The actual force output borne by the balancing node; For balanced node indexes.

9. The method according to claim 1, characterized in that, The operational data includes line transmission capacity constraints, and the relaxation of some constraints in the system physical operational constraints includes: The line transmission capacity constraint boundary in the physical operation constraints is enlarged by a preset factor.

10. The method according to claim 9, characterized in that, The constraint on the line transmission capacity is linearized using a polygonal linear approximation method. The linearization process of the polygon linear approximation method includes: Based on the active and reactive power flow of the branch, the apparent power constraints of the branch are projected to multiple directions in space using trigonometric function projection coefficients, so as to transform the non-convex apparent power square form constraint into a set of linear polyhedral boundary constraints.