Device for dividing three regions after power failure of power grid based on high-order Isin solver

Abstract

The invention discloses a device for dividing three regions after power failure of a power grid based on a high-order Isin solver. According to the specific implementation mode, a power grid topology is modeled into an undirected graph, and a partition target is converted into a minimum cut problem of minimizing the sum of line weights between different regions under the condition that engineering constraints are met. According to the method, constraint and targets are divided by constructing an Isin Hamiltonian quantity containing a high-order coupling term and a plurality of regions of native coding. The Isin Hamiltonian can be directly deployed on a high-order Isin solver realized by adopting a complementary metal oxide semiconductor architecture, and an optimal partition scheme is quickly searched by utilizing the characteristics of large-scale parallelism and low power consumption of the Isin solver. According to the method, the problem of rapid and accurate partition of parallel recovery after blackout is effectively solved, and the power grid toughness is remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of power grid optimization control and circuit design, specifically involving a three-region division device for power grid after power outage based on a high-order Ising solver. It is suitable for the rapid and efficient three-region division of the power grid during the power system fault recovery phase. Background Technology

[0002] In modern large-scale power systems, the rapid and orderly restoration of power supply after a severe fault or cascading blackout is crucial for ensuring the smooth operation of society. To improve the parallelism and flexibility of system restoration, the de-energized power grid is typically divided into several structurally sound sub-regions with independent start-up capabilities; this process is called grid partitioning. Appropriate partitioning not only facilitates parallel black start and load restoration in each sub-region but also reduces inter-regional interference, improving the stability and efficiency of the overall restoration process.

[0003] From an electrical coupling perspective, an ideal zoning scheme should cut as few interconnecting lines as possible, or lines with the smallest possible electrical weights, to reduce mutual interference between regions. Therefore, the power grid zoning problem can be mathematically modeled as a graph-based minimum cut problem: minimizing the sum of the equivalent weights of the cut lines while satisfying the requirement for the number of zoning areas. It is worth noting that actual power grid restoration often requires dividing the grid into three or more sub-regions to match the number of available start-up power sources, load distribution, and topology. This problem belongs to the NP-hard combinatorial optimization problem, and its solution complexity increases sharply with the number of zoning areas and the scale of the power grid.

[0004] Traditional digital processors based on the von Neumann architecture, due to their serial execution mechanism, memory access bottlenecks, and energy consumption limitations, are no longer able to meet the real-time solution requirements of high-dimensional optimization problems. To overcome the limitations of the computing architecture, Ising solvers based on the physical energy function minimization mechanism have gradually become a research focus in the industry. However, existing Ising machines are mostly limited to processing binary quadratic unconstrained optimization models. Solving higher-order problems requires order reduction, which not only significantly increases the problem size but may also distort the energy landscape of the original problem, affecting the quality of the solution.

[0005] To address this, the proposed method directly utilizes a high-order Ising solver based on complementary metal-oxide-semiconductor (CMOS). This solver natively supports high-order coupling terms and can directly map the complex constraints and objective functions of the three-part grid problem without order reduction. By implementing high-order spin product operations at the circuit level, this hardware architecture can complete dynamic evolution at microsecond speeds, accurately and efficiently searching for the optimal or near-optimal grid partitioning scheme, providing powerful hardware acceleration capabilities for rapid coordinated recovery after large-scale grid faults. Summary of the Invention

[0006] To address the issues of low efficiency in regional recovery after a major power outage and the need for order reduction in traditional solvers to handle high-order constraints, this invention provides a three-region partitioning device for power grids after an outage based on a high-order Ising solver. This method first abstracts the outage power grid topology into a weighted undirected graph and establishes a minimum cut problem model. Then, it constructs a high-order Ising Hamiltonian containing a unique partitioning constraint term, a minimum cut objective term, and an augmented term embedded with engineering constraints. The method derives the node dynamics equations, designs an extended circuit based on quantized node cross-resistance coupling Ising machine nodes, and uses logic gates to implement multi-spin product operations. Finally, it introduces an intelligent perturbation mechanism to shield coupling paths that already satisfy constraints, guiding the system out of local optima. This invention leverages the parallel characteristics of the complementary metal-oxide-semiconductor (CMOS) architecture Ising solver to quickly output the optimal partitioning scheme, improving the resilience of power grid recovery.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A three-region division device for power grid after power outage based on a high-order Ising solver includes the following steps:

[0009] S10. Abstract the power grid topology after the power outage into a weighted undirected graph, and establish a minimum cut problem model with multiple regions as the objective of minimizing the total weight of the cut lines.

[0010] S20. Based on the target number of partitions, construct the unique partition constraint term and the minimum cut target term in the higher-order Ising Hamiltonian. The unique partition constraint term is used to encode the unique constraint that each node belongs to only one region, and the minimum cut target term ensures that the total weight of the edges connecting different subsets is minimized.

[0011] S30. The inequality constraints, including the minimum number of nodes in each sub-region, power balance requirements, and black-start power supply configuration requirements, are uniformly encoded into the augmented terms of the higher-order Ising Hamiltonian by using the penalty function method and introducing slack variables.

[0012] S40. Starting from the constructed high-order Ising Hamiltonian formula, the dynamic equation describing the evolution of node states over time is derived. The dynamic equation follows the law of gradually decreasing along the negative gradient direction of the Hamiltonian to approximate the lowest energy state, corresponding to the optimal power grid partitioning scheme.

[0013] S50. Based on the derived dynamic equations, an extended circuit structure based on the quantized node interactive resistance coupling Ising machine node is adopted. A logic function link for realizing multi-spin product is introduced into the circuit. Multi-spin variable multiplication operation considering the coupling coefficient sign is realized through XOR gates or XNOR gates, so that the circuit can directly reflect the role of higher-order coupling terms in dynamic evolution, thereby realizing the energy decrease process containing higher-order terms.

[0014] S60 introduces an intelligent disturbance mechanism, which generates corresponding control signals by real-time identification of whether the fourth-order objective sub-term is in an energy-optimal state. control signals Disturbance signal Obtained by performing logical operations;

[0015] S70, according to the signal By controlling the opening and closing of the transmission gate, the fourth-order objective sub-item in the energy optimal state is automatically shielded in the circuit through the transmission gate, the coupling path corresponding to the fourth-order objective sub-item in the energy optimal state is cut off, the influence of the fourth-order objective sub-item not in the energy optimal state on the system is retained, and the system is guided to effectively jump out of the local optimum.

[0016] S80. A high-order Ising solver is constructed by inputting the structural parameters of the weighted undirected graph. The evolution of the node voltage is achieved through the charging and discharging of the node capacitors in the circuit, and finally the node voltages stabilize at 0 or... ,in The circuit power supply voltage is the upper limit of the node voltage evolution, thus completing the solution to the problem and outputting the optimal partition result for each node.

[0017] Specifically, in step S10, the power grid topology after the power outage is abstracted as an undirected graph:

[0018]

[0019] In the above formula, V represents the set of nodes, and E represents the set of lines. For any line... ,line weight The equivalent influence weight of the line is represented, and a three-part minimum cut partition model is established with the optimization objective of minimizing the sum of the line weights across different regions.

[0020] Specifically, in step S20, for each node Introducing three spin variables These correspond to whether a node is assigned to region 1, 2, or 3, respectively. Whether a node belongs to a certain region is determined by the relative sign relationship between the three spins, and a unique partitioning constraint term for the node is constructed:

[0021]

[0022] In the above formula, A is the uniqueness constraint penalty coefficient, and A>0. It is a region index variable. , Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region v corresponding to node i, and V be the set of nodes. This is a unique partitioning constraint term for nodes, used to ensure that among the three spin variables corresponding to each node, there is exactly one spin that is opposite to the other two spins. The partition corresponding to the spin with the opposite value is the partition to which the node belongs.

[0023] Specifically, in step S20, for any line Construct the minimum cut objective term for the three partitions:

[0024]

[0025] In the above formula, B is the minimum cut objective weight coefficient, B<0, Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region u corresponding to node j. Let be the spin variable of the region v corresponding to node i. Let v be the spin variable of the region corresponding to node j. It is a region index variable. E represents the set of routes. To minimize the cut objective term, an energy penalty is imposed on connections between different regions through a four-spin product term, thereby minimizing the cross-regional line weight.

[0026] Specifically, in step S30, the engineering inequality constraints of each subdomain are uniformly encoded using a relaxation variable and penalty function approach, and the region indicator variable for node i belonging to region u is defined as:

[0027]

[0028] In the above formula, Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region v corresponding to node i. Let be a region indicator variable for which node i belongs. When node i belongs to region u... When node i does not belong to region u, and any form of The inequality constraints are transformed into equality constraints:

[0029]

[0030] In the above formula, 𝑟 represents the constraint type index, used to distinguish between region size constraints, power balance constraints, and black start configuration constraints. This represents the constraint weight parameter corresponding to node i. This indicates the minimum threshold requirement for region u under the given constraint type. Let V be a non-negative slack variable, representing the set of nodes, and construct a unified augmentation term:

[0031]

[0032] In the above formula, As a penalty factor corresponding to the constraint type, the inequality constraints are embedded in a higher-order Ising Hamiltonian in a unified form:

[0033]

[0034] In the above formula, The total Hamiltonian integrates three core elements: unique node partitioning constraints, minimum cut target of the power grid, and engineering operation constraints. This is a unique partitioning constraint for nodes, used to ensure that each node belongs to only one region; The minimum cut objective term is used to quantify the cross-regional line weight penalty to achieve the weight and minimize the objective. It is an augmentation term for engineering constraints, used to embed inequality constraints such as region size, power balance, and black-start power supply configuration.

[0035] Specifically, in step S40, the spin variable is... Mapped to node voltage According to the total Ising Hamiltonian For each spin variable and constrained slack variables Corresponding node voltage state Find the partial derivatives and construct the continuous-time dynamic equations:

[0036]

[0037]

[0038] In the above formula, This represents the voltage in region u corresponding to node i. rate of change over time Represents the slack variable voltage of region u corresponding to constraint type r. rate of change over time Represents the total Ising Hamiltonian For node voltage The first-order partial derivative reflects the rate and direction of change of the total energy with respect to the spin voltage of region u corresponding to node i. Represents the total Ising Hamiltonian For node voltage The first-order partial derivative reflects the rate and direction of change of the total energy with respect to the relaxation variable voltage of the region u corresponding to the constraint type r. The proportionality constant for the correlation quantization node cross-resistive coupling Ising machine circuit parameters. , The total Ising Hamiltonian is used to obtain the differential equations of the voltage changes of each node over time, and the circuit structure is built based on the differential equations of the voltage of each node.

[0039] Specifically, in step S50, the quantization node cross-resistive coupling Ising machine node circuit consists of a current transmitter and a quantizer. The current transmitter enables the node spin energy to sum the coupling current from the neighborhood and drive the node capacitor to update the voltage. spin voltage Before transmission, it is quantized by a cascaded inverter and compared with the threshold voltage. The comparison is mapped to discrete spin. This enables a continuous-time spin evolution process.

[0040] Specifically, in step S50, a resistive coupling extension circuit is constructed based on the differential equations of each node obtained in step S40 to simultaneously achieve both second-order coupling and higher-order coupling. The second-order coupling is achieved through the quantization output of adjacent nodes. This is achieved by inputting the voltage to the target node via a coupling resistor; the voltage at the node input is fixed as a voltage. , Threshold voltage, The circuit supply voltage represents the upper limit of the node voltage evolution. The output voltage, after being quantized by a comparator, yields a spin state of 0 or 0. When the adjacent spin voltage changes, the input current direction reverses accordingly to drive the target node voltage update; higher-order coupling terms include the fourth-order action in the minimum cut target term and the third and fourth-order actions in the augmented term, which realize multi-spin product operations by introducing logic function links, for the total Hamiltonian Multi-spin product terms in , Let m be the m-th spin variable in the multi-spin product term. The logic gate selection for the corresponding circuit is determined by the following formula:

[0041]

[0042] In the above formula, sgn is the extracted coupling coefficient. The sign function of the positive and negative attributes, where k is the product order. Let be the coupling coefficient of the k-th order term, when At that time, the symbol generation logic is implemented by an XOR gate. At that time, it is implemented by an XOR gate; because the spin variable in this device There exists a special simplification rule for the spin-squared term: for any spin variable The squared terms satisfy In the multi-spin product operation of second-order and higher-order coupling terms, if a square term of the same spin variable appears, the square term can be directly replaced with a constant 1, thereby simplifying the operation expression of the spin product term and the circuit implementation logic.

[0043] Specifically, in step S60, the intelligent perturbation mechanism targets the fourth-order minimum cut objective term, and the fourth-order minimum cut objective term obtained in step S20... The corresponding circuit structure is as follows:

[0044]

[0045] In the above formula, B is the minimum cut target weight coefficient, and B < 0; Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region u corresponding to node j. Let be the spin variable of the region v corresponding to node i. Let v be the spin variable of the region corresponding to node j. It is a region index variable. ; , , , Spin variables , , , The corresponding node voltage quantization value is used to introduce an analog switch between the logic gate output and the coupling resistor, controlled by the energy value state of the fourth-order sub-term in the minimum cut target term. A transmission gate is used as the control element; the switching state of the transmission gate is determined by the control signal. The decision, expressed logically, is as follows:

[0046]

[0047] In the above formula, , , , Spin variables , , , The corresponding node voltage quantization value, The control signal for the switching state of the transmission gate is used. A four-input XNOR gate is used to determine the energy value state of the fourth-order sub-term in the minimum cut objective term: when the fourth-order objective sub-term is -1, it indicates that the fourth-order objective sub-term has reached the energy optimal state, and the XNOR gate output is 1; when the fourth-order product term is +1, it indicates that the fourth-order objective sub-term has not yet reached the energy optimal state, and the XNOR gate output is 0. This is a global disturbance control signal used to control whether the intelligent disturbance mode is enabled. When the system enters the intelligent perturbation state, if the fourth-order objective sub-term is already in the energy-optimal state, then the XNOR output is 1. When the transmission gate is turned off, the higher-order coupling paths corresponding to the fourth-order target sub-term are automatically shielded, preventing these paths from contributing to the node current injection. If the fourth-order target sub-term has not yet reached its energy optimum, the XNOR output is 0. The transmission gate remains open, and the fourth-order objective sub-term continues to participate in the energy descent process; when At that time, the intelligent perturbation function is turned off, and the coupling paths corresponding to all fourth-order target sub-items return to normal operation.

[0048] Specifically, in step S60, the disturbance control signal A square wave modulation strategy with decreasing pulse width is adopted to effectively balance the conflicting needs of global exploration and local convergence.

[0049] Specifically, in step S80, circuit-level simulation verification is performed in Cadence Virtuoso using a 45nm process, and the partitioning result is determined by observing the evolution trajectory of the three voltage curves of the observed nodes.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] (1) The present invention constructs a high-order Ising Hamiltonian, which can natively encode the constraints and objectives of multi-region partitioning without the need for order reduction processing, thus avoiding the variable inflation problem caused by order reduction in traditional methods and significantly reducing the complexity of problem solving. At the same time, relying on the massive parallelism and low power consumption characteristics of the complementary metal-oxide-semiconductor architecture Ising machine, it can quickly search for the optimal partitioning scheme, effectively improving the efficiency and accuracy of parallel grid recovery after a major power outage.

[0052] (2) The present invention designs an extended circuit structure based on the quantized node interactive resistive coupling Ising machine node, and implements multi-spin product operation through logic gates to accurately reflect the role of higher-order coupling terms in dynamic evolution. The introduced intelligent disturbance mechanism can identify the constraint satisfaction state in real time, automatically shield the coupling path that has satisfied the constraint, guide the system to jump out of the local optimum, ensure the solution quality, and significantly enhance the resilience and stability of power grid fault recovery. Attached Figure Description

[0053] Figure 1 This is an undirected graph of a 6-node power grid topology based on unit edge weights and its partitioning results.

[0054] Figure 2 A schematic diagram of a high-order Ising solver circuit based on quantized node cross-resistive coupling Ising machine nodes (with...) (For example, a node).

[0055] Figure 3 This is a schematic diagram of the circuit implementation of the intelligent disturbance mechanism.

[0056] Figure 4 for Figure 1 The problem considers the optimal partitioning result under the black-start power inequality constraint. Detailed Implementation

[0057] This invention proposes a three-region division device based on a high-order Ising solver after a power outage, which is described in detail below with reference to the accompanying drawings:

[0058] Figure 1 This presents an undirected graph and partitioning results for a 6-node power grid topology based on unit edge weights. Specifically, Figure 1 (a) is a weighted undirected graph input model of a 6-node power grid topology. The power grid topology after a power outage is abstracted as a weighted undirected graph:

[0059]

[0060] The node set is as follows:

[0061]

[0062] The routes are respectively:

[0063]

[0064] Each path in the graph uses a unit edge weight, and the nodes... The node is a black-starting power source. This invention requires dividing the 6-node undirected graph into 3 sub-regions, and minimizing the sum of the weights of the cross-region lines while satisfying the inequality constraint that each sub-region has at least one black-starting power source. Figure 1 (b) is the optimal partitioning result obtained by the high-order Ising solver. The result satisfies the black-start power supply configuration constraint and the minimum cut objective, and intuitively presents the final region assignment of each node.

[0065] Figure 2 This is a schematic diagram of a high-order Ising solver circuit based on a quantized node cross-resistive coupled Ising machine node (with...). (Taking a node as an example). In three-part coding, whether a node belongs to a certain region is determined by the relative sign relationship between spins. The indicator variable for node i belonging to region u=1 is defined as:

[0066]

[0067] In the above formula, Let be the region indicator variable for node i belonging to region 1. , , Let be the spin variables of node i in regions 1, 2, and 3, respectively. When node i belongs to region 1... relatively and Take the opposite sign, and thus When node i does not belong to region 1, we have To ensure that each sub-region contains at least one black-start power node, black-start inequality constraints need to be constructed. Taking region 1 as an example:

[0068]

[0069] In the above formula, Let node i be a region indicator variable belonging to region 1. This inequality indicates that region 1 should contain at least one element from set 1. The black start node. Introduce a slack variable. Transform the above inequality constraints into equation form:

[0070]

[0071] In the above formula , , Let the spin variables of nodes i in regions 1, 2, and 3 be respectively, and then we can further construct augmented terms:

[0072]

[0073] In the above formula The black-start constraint penalty factor is an augmented term that enables the system to automatically satisfy the black-start constraint during the energy minimization process.

[0074] The node uniqueness constraint is:

[0075]

[0076] The minimum cut objective term is:

[0077]

[0078] The final formula for the higher-order Ising Hamiltonian is:

[0079]

[0080] In the above formula, A is the uniqueness constraint penalty coefficient, A>0, and B is the minimum cut objective weight coefficient, B<0. Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region u corresponding to node j. Let be the spin variable of the region v corresponding to node i. Let v be the spin variable of the region corresponding to node j. For the total Hamiltonian, This is a unique partition constraint for the node. For the minimum cut objective term, For the augmented term of the engineering constraints, based on the total Ising Hamiltonian For each spin variable and constrained slack variables Corresponding node voltage state Find the partial derivatives and construct the continuous-time dynamic equations:

[0081]

[0082]

[0083] In the above formula, This represents the voltage in region u corresponding to node i. rate of change over time Represents the slack voltage rate of change over time Represents the total Ising Hamiltonian For node voltage The first-order partial derivative reflects the rate and direction of change of the total energy with respect to the spin voltage of region u corresponding to node i. Represents the total Ising Hamiltonian For node voltage The first-order partial derivative reflects the rate and direction of change of the total energy with respect to the relaxation voltage. The proportionality constant for the correlation quantization node cross-resistive coupling Ising machine circuit parameters. ,by Figure 1 Spin variable of vertex 1 For example, the differential equation for its voltage evolution is:

[0084]

[0085] In the above formula, A is the uniqueness constraint penalty coefficient, A>0, and B is the minimum cut objective weight coefficient, B<0. The proportionality constant for the correlation quantization node cross-resistive coupling Ising machine circuit parameters. Similarly, the voltage differential equations for vertices 2 to 6 can be derived in this way.

[0086] The node circuit of this invention employs a quantized node cross-resistive coupled Ising machine node structure, which generates voltage through the charging and discharging of node capacitors by current. When the polarity of the input current changes, the capacitor voltage flips accordingly, achieving spin-current switching. Switching between nodes. The quantized node, an inter-resistive coupled Ising machine node, mainly consists of current transmitters and quantizers. The current transmitter is composed of field-effect transistors M1–M8, where M1–M4 provide local feedback to achieve input potential self-balancing, and M5–M8 are two pairs of current mirrors, enabling the node to sum the coupled currents from its neighborhood and use them to drive the capacitors to update the node voltage. Node voltage. Before transmission, the signal is quantized by a cascaded inverter consisting of M9–M12 and compared with the threshold voltage. The comparison is mapped to discrete spin. ,in The circuit power supply voltage enables a continuous-time spin evolution process.

[0087] The system energy consists of second-order and higher-order terms, expressed as a spin variable. For example, when implementing second-order terms, nodes and nodes The quantization output is achieved through a programmable coupling resistor. Connected, and satisfying the following relationship:

[0088]

[0089] In the above formula, R is the reference resistance value. The coupling weights of nodes i and j are set so that the coupling current directly drives the node capacitor voltage update, thereby realizing the two-body interaction of the traditional Ising machine. By adjusting... The resistance value can be precisely matched to the coupling weight of the second-order coefficients.

[0090] Higher-order coupling terms are implemented through the introduction of logical functional links to achieve multi-spin product operations. Spin variables... The three-spin product in the differential equation is generated by a three-input XOR structure and injected into the node input through a coupling resistor. Together with other input currents, it determines the voltage gradient of the node, enabling the node voltage to evolve based on continuous dynamics and achieving a precise reduction of the higher-order Hamiltonian.

[0091] Figure 3 This is a schematic diagram of the circuit implementation of the intelligent perturbation mechanism. During the perturbation phase, the system determines the energy state of each fourth-order objective sub-item in the minimum cut objective term of the total Ising Hamiltonian. Each sub-item corresponds to a path formed by an XOR gate and a coupling resistor connected in series. Only fourth-order objective sub-items that have not yet reached their optimal energy state maintain their corresponding coupling path. Specifically, a transmission gate controlled by the energy state of the fourth-order objective sub-item is connected in series between the XOR gate output and the coupling resistor. The switching state of the transmission gate is controlled by a control signal. Decide:

[0092]

[0093] In the above formula, , , , Spin variables , , , The corresponding node voltage quantization value, The control signal for the switching state of the transmission gate is used. The four-input XOR gate is used to determine whether the fourth-order objective sub-item has reached the energy optimal state: when the product value is -1, it means that the fourth-order objective sub-item has reached the energy optimal state, and the XOR gate output is 1; when the product value is 1, it means that the fourth-order objective sub-item has not yet reached the optimal state, and the XOR gate output is 0. This is a global disturbance control signal used to uniformly control the start and stop of the intelligent disturbance mode. When When the system enters the intelligent perturbation state, if the fourth-order objective sub-term is already in the energy-optimal state, the XOR gate output is 1. The corresponding transmission gate is turned off, thereby automatically shielding the higher-order coupling paths corresponding to the fourth-order target term, so that the fourth-order target term no longer contributes to the node current; if the fourth-order target term has not yet reached the energy optimal state, the XOR gate output is 0. The transmission gate remains open, and the fourth-order objective sub-term continues to participate in the system's energy reduction process. At that time, the intelligent perturbation function is turned off, and the coupling paths corresponding to all fourth-order target sub-items return to normal operation.

[0094] Figure 4 for Figure 1 The problem considers the optimal partitioning results under the black-start power inequality constraint. The figure shows a set of partitioning results for... Figure 1 The node voltage evolution results for the problem shown are presented. Six different line types correspond to six nodes, and three markers—circles, squares, and triangles—represent the quantized voltage trajectory of each node mapped to three candidate subsets. When the marker corresponding to one of the three curves of the same line type differs from the markers of the other two curves, the partition represented by that marker can be determined as the final partition to which the node belongs. The corresponding final partitioning results are shown in [the table / document / etc.]. Figure 1 (b), where node 6 belongs to partition 1, nodes 3, 4 and 5 belong to partition 2, and nodes 1 and 2 belong to partition 3.

[0095] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any changes made based on the design principles of the present invention, or any non-creative modifications made thereon, shall fall within the scope of protection of the present invention.

Claims

1. A three-region division device for power grid after power outage based on a high-order Ising solver, comprising the following steps: S10. Abstract the power grid topology after the power outage into a weighted undirected graph, and establish a minimum cut problem model with three regions as the objective of minimizing the total weight of the cut lines. S20. Based on the target number of partitions, construct the unique partition constraint term and the minimum cut target term in the higher-order Ising Hamiltonian. The unique partition constraint term is used to encode the unique constraint that each node belongs to only one region, and the minimum cut target term ensures that the total weight of the edges connecting different subsets is minimized. S30. The inequality constraints, including the minimum number of nodes in each sub-region, power balance requirements, and black-start power supply configuration requirements, are uniformly encoded into the augmented terms of the higher-order Ising Hamiltonian by using the penalty function method and introducing slack variables. S40. Starting from the constructed high-order Ising Hamiltonian formula, the dynamic equation describing the evolution of node states over time is derived. The dynamic equation follows the law of gradually decreasing along the negative gradient direction of the Hamiltonian to approximate the lowest energy state, corresponding to the optimal power grid partitioning scheme. S50. Based on the derived dynamic equations, an extended circuit structure based on the quantized node interactive resistance coupling Ising machine node is adopted. A logic function link for realizing multi-spin product is introduced into the circuit. Multi-spin variable multiplication operation considering the coupling coefficient sign is realized through XOR gates or XNOR gates, so that the circuit can directly reflect the role of higher-order coupling terms in dynamic evolution, thereby realizing the energy decrease process containing higher-order terms. S60 introduces an intelligent disturbance mechanism, which generates corresponding control signals by real-time identification of whether the fourth-order objective sub-term is in an energy-optimal state. control signals Disturbance signal Obtained by performing logical operations; S70, according to the signal By controlling the opening and closing of the transmission gate, the fourth-order objective sub-item in the energy optimal state is automatically shielded in the circuit through the transmission gate, the coupling path corresponding to the fourth-order objective sub-item in the energy optimal state is cut off, the influence of the fourth-order objective sub-item not in the energy optimal state on the system is retained, and the system is guided to effectively jump out of the local optimum. S80. A high-order Ising solver is constructed by inputting the structural parameters of the weighted undirected graph. The evolution of the node voltage is achieved through the charging and discharging of the node capacitors in the circuit, and finally the node voltages stabilize at 0 or... ,in The circuit power supply voltage is the upper limit of the node voltage evolution, thus completing the solution to the problem and outputting the optimal partition result for each node.

2. The power grid three-region partitioning device based on a high-order Ising solver according to claim 1, wherein in step S10, the power grid topology after the power outage is abstracted as an undirected graph: In the above formula, V represents the set of nodes, and E represents the set of lines. For any line... ,line weight The equivalent influence weight of the line is represented, and a three-part minimum cut partition model is established with the optimization objective of minimizing the sum of the line weights across different regions.

3. The power grid outage three-region division device based on a high-order Ising solver according to claim 1, in step S20, for each node... Introducing three spin variables , respectively corresponding to nodes Whether assigned to region 1, 2, or 3, node Whether a node belongs to a certain region is determined by the relative sign relationship between the three spins, and a unique partitioning constraint term for the node is constructed: In the above formula, A is the uniqueness constraint penalty coefficient, and A>0. It is a region index variable. , Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region v corresponding to node i, and V be the set of nodes. This is a unique partitioning constraint term for nodes, used to ensure that among the three spin variables corresponding to each node, there is exactly one spin that is opposite to the other two spins. The partition corresponding to the spin with the opposite value is the partition to which the node belongs.

4. The power grid outage three-region division device based on a high-order Ising solver according to claim 1, wherein in step S20, for any line Construct the minimum cut objective term for the three partitions: In the above formula, B is the minimum cut objective weight coefficient, B<0, Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region u corresponding to node j. Let be the spin variable of the region v corresponding to node i. Let v be the spin variable of the region corresponding to node j. It is a region index variable. E represents the set of routes. To minimize the cut objective term, an energy penalty is imposed on connections between different regions through a four-spin product term, thereby minimizing the cross-regional line weight.

5. The three-region division device for power grid after power outage based on a high-order Ising solver according to claim 1, in step S30, the engineering inequality constraints of each subdomain are uniformly encoded using slack variables and penalty functions, and the region indicator variable for node i belonging to region u is defined as: In the above formula, Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region v corresponding to node i. Let be a region indicator variable for which node i belongs. When node i belongs to region u... When node i does not belong to region u, and any form of Inequality constraints are transformed into equality constraints: In the above formula, 𝑟 represents the constraint type index, used to distinguish between region size constraints, power balance constraints, and black start configuration constraints. This represents the constraint weight parameter corresponding to node i. This indicates the minimum threshold requirement for region u under the given constraint type. Let V be a non-negative slack variable, representing the set of nodes, and construct a unified augmentation term: In the above formula, As a penalty factor corresponding to the constraint type, the inequality constraints are embedded in a higher-order Ising Hamiltonian in a unified form: In the above formula, The total Hamiltonian integrates three core elements: unique node partitioning constraints, minimum cut target of the power grid, and engineering operation constraints. This is a unique partitioning constraint for nodes, used to ensure that each node belongs to only one region; The minimum cut objective term is used to quantify the cross-regional line weight penalty to achieve the weight and minimize the objective. It is an augmentation term for engineering constraints, used to embed inequality constraints such as region size, power balance, and black-start power supply configuration.

6. The three-region division device for power grid outage based on a high-order Ising solver according to claim 1, wherein in step S40, the spin variable is... Mapped to node voltage According to the total Ising Hamiltonian For each spin variable and constrained slack variables Corresponding node voltage state Find the partial derivatives and construct the continuous-time dynamic equations: In the above formula, This represents the voltage in region u corresponding to node i. rate of change over time Represents the slack variable voltage of region u corresponding to constraint type r. rate of change over time Represents the total Ising Hamiltonian For node voltage The first-order partial derivative reflects the rate and direction of change of the total energy with respect to the spin voltage of region u corresponding to node i. Represents the total Ising Hamiltonian For node voltage The first-order partial derivative reflects the rate and direction of change of the total energy with respect to the relaxation variable voltage of the region u corresponding to the constraint type r. The proportionality constant for the correlation quantization node cross-resistive coupling Ising machine circuit parameters. , The total Ising Hamiltonian is used to obtain the differential equations of the voltage changes of each node over time, and the circuit structure is built based on the differential equations of the voltage of each node.

7. The three-region division device for power grid after power outage based on a high-order Ising solver according to claim 1, in step S50, the quantization node interactive resistive coupled Ising machine node circuit consists of a current transmitter and a quantizer. The current transmitter enables the node spin energy to sum the coupling current from the neighboring region and drive the node capacitor to update the voltage. spin voltage Before transmission, it is quantized by a cascaded inverter and compared with the threshold voltage. The comparison is mapped to discrete spin. This enables a continuous-time spin evolution process.

8. The three-region division device for power grid after power outage based on a high-order Ising solver according to claim 1, in step S50, a resistive coupling extension circuit is built according to the differential equations of each node obtained in step S40 to simultaneously realize the two types of effects of second-order coupling terms and high-order coupling. The second-order coupling is achieved through the quantization output of adjacent nodes. This is achieved by inputting the voltage to the target node via a coupling resistor, with the voltage value at the node input terminal fixed as a voltage. , Threshold voltage, The circuit supply voltage represents the upper limit of the node voltage evolution. The output voltage, after being quantized by a comparator, yields a spin state of 0 or 0. When the adjacent spin voltage changes, the input current direction reverses accordingly to drive the target node voltage update; higher-order coupling terms include the fourth-order action in the minimum cut target term and the third and fourth-order actions in the augmented term, which realize multi-spin product operations by introducing logic function links, for the total Hamiltonian Multi-spin product terms in , Let m be the m-th spin variable in the multi-spin product term. The logic gate selection for the corresponding circuit is determined by the following formula: Where sgn is the extraction coupling coefficient The sign function of the positive and negative attributes, where k is the product order. Let be the coupling coefficient of the k-th order term, when At that time, the symbol generation logic is implemented by an XOR gate. At that time, it is implemented by an XNOR gate; because the spin variable in this device There exists a special simplification rule for the spin-squared term: for any spin variable The squared terms satisfy In the multi-spin product operation of second-order and higher-order coupling terms, if a square term of the same spin variable appears, the square term can be directly replaced with a constant 1, thereby simplifying the operation expression of the spin product term and the circuit implementation logic.

9. The power grid outage three-region division device based on a high-order Ising solver according to claim 1, wherein in step S60, the intelligent disturbance mechanism targets the fourth-order minimum cut objective term, and the fourth-order minimum cut objective term obtained in step S20... The corresponding circuit structure is as follows: In the above formula, B is the minimum cut target weight coefficient, and B < 0; Let be the spin variable of the region u corresponding to node i. Let be the spin variable of the region u corresponding to node j. Let be the spin variable of the region v corresponding to node i. Let v be the spin variable of the region corresponding to node j. It is a region index variable. ; , , , Spin variables , , , The corresponding node voltage quantization value is used to introduce an analog switch between the logic gate output and the coupling resistor, controlled by the energy value state of the fourth-order sub-term in the minimum cut target term. A transmission gate is used as the control element; the switching state of the transmission gate is determined by the control signal. The decision, the logical expression is: In the above formula, , , , Spin variables , , , The corresponding node voltage quantization value, The control signal for the switching state of the transmission gate is used. A four-input XNOR gate is used to determine the energy value state of the fourth-order sub-term in the minimum cut objective term: when the fourth-order objective sub-term is -1, it indicates that the fourth-order objective sub-term has reached the energy optimal state, and the XNOR gate output is 1; when the fourth-order objective sub-term is +1, it indicates that the fourth-order objective sub-term has not yet reached the energy optimal state, and the XNOR gate output is 0. This is a global disturbance control signal used to control whether the intelligent disturbance mode is enabled. When the system enters the intelligent perturbation state, if the fourth-order objective sub-term is already in the energy-optimal state, then the XNOR output is 1. When the transmission gate is turned off, the higher-order coupling paths corresponding to the fourth-order target sub-term are automatically shielded, preventing these paths from contributing to the node current injection. If the fourth-order target sub-term has not yet reached its energy optimum, the XNOR output is 0. The transmission gate remains open, and the fourth-order objective sub-term continues to participate in the energy descent process; when At that time, the intelligent perturbation function is turned off, and the coupling paths corresponding to all fourth-order target sub-items return to normal operation.

10. The three-region division device for power grid after power outage based on a high-order Ising solver according to claim 1, wherein in step S60, the disturbance control signal... A square wave modulation strategy with decreasing pulse width is adopted, which effectively balances the conflicting needs of global exploration and local convergence.

Images