Fast power supply recovery method and device based on source-load-storage coordinated power grid partition optimization

Abstract

The application belongs to the technical field of power supply recovery, and particularly relates to a fast power supply recovery method and device based on source-load-storage collaborative power grid partition optimization, which comprises the following steps: S1, constructing an electrical coupling degree index, a source-load-storage power balance index, a grid structure index and a node scale index respectively, and comprehensively forming a power grid partition comprehensive index; S2, taking the power grid partition comprehensive index as a fitness function, and designing an energy entropy driven hybrid selection mechanism to improve a simulated annealing genetic algorithm, and using the improved simulated annealing genetic algorithm to solve the power grid partition; S3, on the basis of the power grid partition, taking each partition as a unit, and establishing an optimization scheduling model with the goal of maximizing the recovery of important loads; and S4, using a particle swarm algorithm to solve the optimization scheduling model, and obtaining a fast power supply recovery scheme. The application can realize the priority recovery of key loads, and enhance the fast recovery capability and reliability of the power grid.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power supply recovery, and particularly relates to a fast power supply recovery method and device based on source-load-storage collaborative power grid partition optimization. BACKGROUND

[0002] When the power grid is in normal operation, it may encounter unconventional events, which refer to sudden disasters (such as extreme weather, earthquakes, cyber attacks) or major equipment failures (such as main transformer explosion, transmission line chain breakage) that exceed the disaster tolerance capacity of the conventional design of the power grid. Such events often lead to the destruction of the power grid topology structure, local or even global power outage, and therefore, it is necessary to quickly restore power supply. Traditional power supply recovery technology mainly relies on three types of methods, namely, relay protection device, centralized control and distributed control. Relay protection passively isolates faults by presetting thresholds, and cannot dynamically adapt to complex and variable unconventional event scenarios; centralized control relies on a central decision system, and is prone to response delay and single-point failure when communication is interrupted or large-scale failure occurs; distributed control improves local autonomy, but lacks a global coordination mechanism, and conflicts between regional strategies occur frequently. The existing methods generally have defects such as slow recovery speed, poor strategy adaptability, and low important load guarantee rate in extreme events.

[0003] The collaborative recovery method based on power grid partition improves the recovery efficiency by dividing autonomous regions, but the existing research has significant limitations: the partition index focuses on a single dimension (such as electrical distance, source-load static balance), and ignores key factors such as dynamic adjustment capability of energy storage and node voltage stability; the partition algorithm (such as traditional genetic algorithm, tabu search) is prone to local optimum, and it is difficult to balance the topology tightness and functional autonomy; the cross-region mutual aid mechanism is missing after partition, leading to insufficient distributed power consumption and limited flexibility of recovery strategy.

[0004] Research on distributed power island operation mainly focuses on local power balance, and uses real-time monitoring and predictive scheduling to achieve island division. However, existing models oversimplify the objective function, and do not include power supply reliability, voltage deviation, and energy storage state of charge decay into the multi-objective optimization framework. In addition, the collaborative mechanism of island division and power grid partition has not been established, and a three-dimensional recovery architecture of "partition autonomy-island complement-global mutual aid" cannot be formed, which restricts the spatiotemporal coordination potential of source-load-storage resources.

[0005] In summary, the current technical system has three major contradictions: structural imbalance between single partition index and multi-dimensional operation demand; game dilemma between algorithm convergence and global optimization ability; and lack of coordination between local recovery strategy and global resource allocation. These defects make it difficult for existing methods to achieve fast and accurate power supply recovery in unconventional events, and there is an urgent need to build a new solution that integrates multi-dimensional evaluation index, intelligent optimization algorithm and cross-region collaborative mechanism. SUMMARY

[0006] In view of the above deficiencies in the prior art, the purpose of the present application is to provide a method and device for fast power restoration based on source-load-storage collaborative power grid partition optimization, which can achieve priority restoration of critical loads and enhance the fast recovery capability and reliability of the power grid.

[0007] To achieve the above purpose, the present application provides a method for fast power restoration based on source-load-storage collaborative power grid partition optimization, comprising the following steps:

[0008] S1, respectively constructing an electrical coupling degree index C, a source-load-storage power balance index , a grid structure index , and a node size index , and comprehensively forming a comprehensive index of power grid partition , expressed as:

[0009] (1);

[0010] In the formula, , , , are the weight coefficients of C, , , respectively;

[0011] S2, taking as the fitness function, and designing an energy entropy driven hybrid selection mechanism to improve the simulated annealing genetic algorithm, and using the improved simulated annealing genetic algorithm to solve the power grid partition;

[0012] S3, on the basis of the power grid partition, taking each partition as a unit, an optimization scheduling model is established to maximize the restoration of important loads as the target;

[0013] S4, using a particle swarm algorithm to solve the optimization scheduling model to obtain a fast power restoration scheme.

[0014] As a preferred scheme of the present application, in S1, the construction process of the electrical coupling degree index C is:

[0015] The electrical coupling degree between node i and node j is expressed as:

[0016] (2);

[0017] In the formula, is the power interaction between node i and node j; , are the node voltages of node i and node j respectively;

[0018] In the power grid partition, when node i is in partition A and node j is in partition B, the inter-area coupling degree of partition A and partition B is expressed as:

[0019] (3);

[0020] In the formula, is the number of connection node pairs between partition A and partition B;

[0021] When node i and node j are both in partition A, the intra-area coupling degree of partition A is expressed as:

[0022] (4);

[0023] In the formula, is the number of node pairs in partition A;

[0024] The electrical coupling degree index C is formed comprehensively:

[0025] (5);

[0026] In the formula, is the weight coefficient of ; is the weight coefficient of .

[0027] As a preferred scheme of the present application, in S1, the source-load-storage power balance index is expressed as:

[0028] (6);

[0029] In the formula, , , is a weight coefficient; represents the difference between the minimum net power of partition k; is the total load of partition k; is the energy margin of the energy storage of partition k; is the total capacity of the energy storage of partition k; is the supply-demand matching degree of partition k; is the maximum allowed matching deviation;

[0030] wherein, is expressed as:

[0031] (7);

[0032] In the formula, is the set of power generation nodes in partition k, and x represents one of the power generation nodes; is a set of energy storage nodes within partition k, and m represents one of the energy storage nodes; is a set of load nodes within partition k, and y represents one of the load nodes; is the output of generation node x; is the discharge power of energy storage node m; is the demand of load node y;

[0033] is represented as:

[0034] (8);

[0035] In the formula, is the current state of charge of energy storage node m; is the minimum allowable state of charge of energy storage node m; is the capacity of energy storage node m;

[0036] is represented as:

[0037] (9);

[0038] In the formula, T is a time window; represents the generation power of partition k at time t; represents the energy storage discharge power of partition k at time t; represents the load demand of partition k at time t.

[0039] As a preferred scheme of the present application, in S1, the network structure index is is represented as:

[0040] (10);

[0041] In the formula, Q is a modularity index; when node i and node j are located in partition k, is the electrical distance between node i and node j, is the power transmission capability between node i and node j; is a weight coefficient for balancing the importance of topology tightness and power transmission capability.

[0042] As a preferred scheme of the present application, in S1, the node scale index is is represented as:

[0043] (11);

[0044] In the formula, represents the average number of nodes: K is the total number of partitions; is the number of nodes of partition k;

[0045] (12) ;

[0046] N is the total number of nodes in the whole network.

[0047] As a preferred scheme of the present application, in S2, the fitness function is used to optimize the partition result, and the cooling function and Metropolis criterion in the simulated annealing genetic algorithm are expressed as:

[0048] (13) ;

[0049] (14) ;

[0050] In the formula, Tz and Tz+1 respectively represent the temperature of the zth generation and the (z+1) th generation; a is an annealing temperature coefficient, 0 < a < 1. , In the formula, Fz and Fz+1 respectively represent the fitness value of the zth generation and the (z+1) th generation; P represents the acceptance probability in the simulated annealing algorithm from the zth generation to the (z+1) th generation. In the formula, Fz and Fz+1 respectively represent the fitness value of the zth generation and the (z+1) th generation; P represents the acceptance probability in the simulated annealing algorithm from the zth generation to the (z+1) th generation.

[0051] A hybrid selection mechanism driven by energy entropy is designed, and the energy entropy calculation formula is as follows:

[0052] (15) ;

[0053] In the formula, E represents the energy entropy of the current population. is the Boltzmann probability of the individual u, that is, the probability of the individual u being selected at the current temperature Tz; M is the population size, that is, the total number of individuals in the population. , ,respectively represent the individual energy of the individual u and the individual v, that is, the reciprocal of the fitness value. The threshold value of the energy entropy is set as E0, and according to the selection-acceptance plan:

[0054] If E0 < E, roulette selection is adopted; if E0 > E, tournament selection is adopted.

[0055] The new solution is accepted in combination with the Metropolis criterion, but the acceptance probability P is negatively related to the energy entropy E, and is expressed as:

[0056] The new solution is accepted in combination with the Metropolis criterion, but the acceptance probability P is negatively related to the energy entropy E, and is expressed as:

[0057] (16) ;

[0058] In the formula,​​​​​​ is the energy difference between the new solution and the current solution; is the maximum possible value of the energy entropy.

[0059] As a preferred scheme of the present application, in the S3, the objective function F of the optimization scheduling model is:

[0060] (17);

[0061] (18);

[0062] wherein, is the weight of the load h, the load h is located in the partition k, and K is the total number of partitions; is the load recovery amount of the load h; is the time penalty coefficient; is the total time from the occurrence of the fault to the restoration of power supply; is the start-up time of the black-start power supply; is the operation time of the key switch, the key switch including a sectional switch, a tie switch, a black-start power supply access switch, a bus tie switch, and an automatic reclosing switch; is the power balance adjustment time within the partition;

[0063] The calculation method of is as follows:

[0064] (19);

[0065] wherein, is the start-up time of the black-start power supply in the partition k; is a 01 variable indicating whether the black-start power supply in the partition k is enabled, indicates that the black-start power supply in the partition k is enabled, indicates that the black-start power supply in the partition k is not enabled;

[0066] The calculation method of is as follows:

[0067] (20);

[0068] wherein, is the operation time of the switch s; is a 01 variable indicating whether the switch s is operated, indicates that the switch s is operated, indicates that the switch s is not operated; S is the total number of switches;

[0069] The calculation method of is as follows:

[0070] (21);

[0071] wherein, is the power supply capacity of zone k; is the energy storage output in zone k; is the load demand in zone k; is the regulation rate of zone k.

[0072] As a preferred scheme of the present application, the constraint conditions of the optimization scheduling model include:

[0073] Power balance constraint:

[0074] (22);

[0075] wherein, is the load of zone k at time t; is the wind power output of zone k at time t; is the photovoltaic output of zone k at time t; is the charge / discharge power of the energy storage in zone k at time t; is the transmission power of zone k with the superior power grid at time t; is the power transmission between each zone of zone k at time t;

[0076] Energy storage power and capacity constraint:

[0077] (23);

[0078] wherein, , respectively is the energy storage capacity at time t and time t-1; is the charge / discharge variable of the energy storage at time t, when the energy storage is in the charging state, and when the energy storage is in the discharging state; is the energy storage charging efficiency; is the time interval; is the energy storage discharging efficiency;

[0079] Voltage and frequency constraint:

[0080] (24);

[0081] wherein, U represents voltage, and f represents frequency;

[0082] Energy storage state of charge:

[0083] (25);

[0084] wherein, , SoCmin and SoCmax are the upper and lower limits of state of charge; SoCini is the initial state of charge; SoCini is the initial state of charge;

[0085] Partition autonomy constraint:

[0086] (26);

[0087] In the formula, O is a partition autonomy constraint index; Ck is the available capacity in partition k; Pk is the maximum load in partition k.

[0088] As a preferred scheme of the present application, in S4, the solving process is:

[0089] S4.1, input the partition result, that is, input the pre-divided power grid partition information, including regional node list, source-load-storage configuration parameters, and partition interconnection line capacity limit;

[0090] S4.2, initialize the PSO algorithm parameters, that is, set the particle swarm optimization algorithm parameters, including particle swarm size, particle dimension, maximum iteration number, inertia weight initial value, learning factor, and variable range;

[0091] S4.3, randomly initialize particle position and velocity, and randomly generate an initial particle swarm according to the variable range, wherein the position of each particle represents a recovery strategy;

[0092] S4.4, calculate the particle fitness value, and evaluate the performance of each particle based on the objective function and constraint condition of the optimization scheduling model;

[0093] S4.5, update the individual historical optimal solution, compare the current fitness value of each particle with its historical optimal value, and if it is better, update the individual historical optimal position and individual historical optimal fitness value;

[0094] S4.6, update the group historical optimal solution, and select the global optimal solution from all particles, the group historical optimal position, and the group historical optimal fitness value;

[0095] S4.7, judge whether the condition is met, that is, whether the maximum iteration number is reached or the fitness value difference between two iterations is minimum, if the condition is met, output the optimal solution and terminate iteration, and if the condition is not met, update the position vector and velocity vector of each particle, and return to step S4.4 for iteration.

[0096] The rapid power supply recovery device based on source-load-storage collaborative power grid partition optimization includes a memory, a processor, and a computer program stored on the memory and capable of running on the processor, and the computer program is executed by the processor to realize the above-mentioned method.

[0097] The present application has the beneficial effects of:

[0098] The present application constructs a set of power grid partition indexes comprehensively considering electrical coupling degree, source-load-storage power balance, grid structure and node scale, compared with the traditional single-index partition method, can more comprehensively and accurately reflect the actual operation state and power supply capacity of the power grid, effectively avoids the problems of power imbalance and voltage out-of-limit caused by single index, significantly improves the scientificity and rationality of power grid partition, and lays a solid foundation for subsequent optimal scheduling.

[0099] The present application introduces the idea of simulated annealing to improve the genetic algorithm, forms the SAGA algorithm, effectively overcomes the defects of easy premature convergence and local optimum of the traditional genetic algorithm, significantly improves the optimization ability and optimization performance of the algorithm, makes the power grid partition result more high-quality, can better meet the partition demand of the power grid under different operation conditions, and enhances the flexibility and adaptability of the power grid operation.

[0100] The present application establishes a multi-objective partition rapid recovery optimization model by taking each partition as a unit, gives higher weight to the key load, and preferentially restores its power supply, so as to effectively prevent and reduce the harm caused by power failure, relieve the pressure on the power grid, minimize economic loss, significantly improve the rapid recovery ability and power supply reliability of the power grid when encountering irregular events, and better meet the urgent demand of modern power grid for high-reliability and rapid recovery power supply.

[0101] The present application also fully considers the dynamic characteristics and uncertainty factors of power grid operation, leaves optimization space for further developing dynamic partition method, introducing robust optimization or stochastic optimization method, etc., so that the whole scheme is more forward-looking, practical and sustainable, and can better adapt to the new trends and challenges of future power grid development. BRIEF DESCRIPTION OF DRAWINGS

[0102] Figure 1 is the flow principle diagram of the present application;

[0103] Figure 2 is the SAGA algorithm flow chart in the present application;

[0104] Figure 3 is the particle swarm optimization solution flow chart combined with power grid partition in the present application;

[0105] Figure 4 is the structure diagram of each partition of the power grid containing distributed energy storage in the verification process of the present application;

[0106] Figure 5 is the structure diagram corresponding to the power grid partition of scheme 1 in the verification process of the present application;

[0107] Figure 6This is a structural diagram corresponding to the power grid partition in Scheme 2 during the verification process of this invention;

[0108] Figure 7 This is a schematic diagram of the comprehensive index iteration curve during the verification process of this invention;

[0109] Figure 8 This is a schematic diagram comparing the load recovery time of the two schemes during the verification process of this invention;

[0110] Figure 9 This is a schematic diagram comparing the voltage levels of the two schemes during the verification process of this invention. Detailed Implementation

[0111] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0112] Example 1: As Figure 1 As shown, the rapid power supply restoration method based on source-load-storage coordinated grid zoning optimization includes the following steps:

[0113] S1. Construct the electrical coupling index C and the source-load-storage power balance index respectively. Grid structure indicators and node size indicators And comprehensively form a power grid zoning index. , represented as:

[0114] (1);

[0115] In the formula, , , , C, respectively , , Weighting coefficients;

[0116] S2, will As the fitness function, an energy entropy-driven hybrid selection mechanism is designed to improve the simulated annealing genetic algorithm, and the improved simulated annealing genetic algorithm (SAGA algorithm) is used to solve the power grid partitioning problem;

[0117] S3. Based on the power grid zoning, establish an optimal scheduling model with each zone as a unit, aiming to maximize the recovery of important loads;

[0118] S4. The particle swarm optimization algorithm is used to solve the optimization scheduling model and obtain a fast power restoration scheme.

[0119] In S1, the process of constructing the electrical coupling index C is as follows:

[0120] One of the goals of partitioning is to minimize the electrical coupling between different regions. By reasonably dividing the regions, the electrical coupling between nodes within each region is high, while the electrical coupling between different regions is low, thereby achieving optimization of the system.

[0121] The electrical coupling between node i and node j is represented as:

[0122] (2);

[0123] wherein, is the power interaction between node i and node j; , Vi and Vj are the node voltages of node i and node j, respectively;

[0124] If the electrical coupling between two nodes is low, it means that the electrical connection between them is weak. When partitioning, these nodes can be preferentially divided into different regions, because their mutual influence is small, and partitioning will not have too much impact on the overall performance of the system. On the contrary, if the electrical coupling between two nodes is high, it means that the electrical connection between them is strong. When partitioning, these nodes should be avoided to be divided into different regions as much as possible, because their mutual influence is large, and partitioning may lead to a decrease in system performance or an increase in the risk of fault propagation. In power grid partitioning, minimizing inter-regional coupling and maximizing intra-regional coupling are two complementary goals. Minimizing inter-regional coupling is to reduce the mutual influence between partitions, while maximizing intra-regional coupling is to enhance the autonomy and stability within the partition.

[0125] In power grid partitioning, when node i is in partition A (here, the initial partition) and node j is in partition B, the inter-regional coupling between partition A and partition B

[0126] is represented as:

[0127] (3);

[0128] wherein, is the number of connected node pairs between partition A and partition B;

[0129] The value of can measure the average electrical coupling strength between two partitions. The smaller the value, the weaker the interaction between regions, and the stronger the autonomy of the partition.

[0130] When node i and node j are both in partition A, the intra-regional coupling of partition A is represented as:

[0131] ​ (4);

[0132] wherein, is the number of node pairs in partition A;

[0133] To optimize the minimization of inter-zone coupling degree and the maximization of intra-zone coupling degree simultaneously, the two indicators can be combined to form an electrical coupling degree indicator C:

[0134] (5);

[0135] wherein, is the weight coefficient of ; is the weight coefficient of , used to balance the importance of inter-zone and intra-zone coupling degrees.

[0136] The smaller the value of C is, the better. According to actual requirements, the weight is adjusted. If the power grid has a higher requirement for fault isolation, then is larger. If the power grid has a higher requirement for intra-partition stability, then is larger.

[0137] In S1, the source-load-storage power balance level is measured, not only considering static power matching, but also considering dynamic capacity margin, because it reflects the adjustable capacity of energy storage in the partition and the energy balance potential on the time scale. Similarly, supply-demand matching degree is very important to measure the power balance level, which can quantify the space-time matching characteristics of power generation, energy storage and load. Considering the inter-partition interaction capability can evaluate the power support capability of the partition in emergency, and the source-load-storage power balance indicator is expressed as:

[0138] (6);

[0139] wherein, , , is the weight coefficient (which can be set according to requirements, the same below); represents the minimum difference of net power of partition k; is the total load of partition k; is the available energy margin of energy storage of partition k; is the total capacity of energy storage of partition k; is the supply-demand matching degree of partition k; is the maximum allowed matching deviation;

[0140] wherein, is expressed as:

[0141] (7);

[0142] In the formula, is the set of generation nodes within the partition k, and x represents one of the generation nodes; is the set of energy storage nodes within the partition k, and m represents one of the energy storage nodes; is the set of load nodes within the partition k, and y represents one of the load nodes; is the output of the generation node x; is the discharge power of the energy storage node m (a positive value indicates discharging, and a negative value indicates charging); is the demand of the load node y;

[0143] is represented as:

[0144] (8);

[0145] In the formula, is the current state of charge of the energy storage node m (the proportion of the current storage capacity of the energy storage to its total capacity); is the minimum allowed state of charge of the energy storage node m; is the capacity of the energy storage node m;

[0146] is represented as:

[0147] (9);

[0148] In the formula, T is the time window (usually takes the value of 24, representing 1 day); represents the generation power of the partition k at time t; represents the energy storage discharge power of the partition k at time t; represents the load demand of the partition k at time t.

[0149] In S1, based on the traditional modularity function used to measure the internal connection tightness and the isolation degree between partitions, electrical characteristics are introduced, and electrical distance and power transmission capacity are integrated into the calculation of the modularity function. According to the load distribution and power output, the weights of the topology and electrical characteristics are dynamically adjusted to improve the adaptability of the index. The grid structure index is represented as:

[0150] (10);

[0151] In the formula, Q is the modularity index used to measure the degree of partition modularity; when node i and node j are located in partition k, is the electrical distance between node i and node j, is the power transmission capacity between node i and node j; To balance the importance of the weight coefficient of the topology tightness and the power transmission ability. The modularity index is an existing index, which can be obtained according to the prior art.

[0152] In S1, the node size index is expressed as:

[0153] (11);

[0154] In the formula, , represents the average node number: K is the total number of partitions; is the number of nodes in the partition k;

[0155] (12);

[0156] In the formula, N is the total number of nodes in the whole network.

[0157] The genetic algorithm is a global optimization algorithm, and its global search capability can ensure that the optimal solution is gradually approached as the number of iterations increases. When the number of iterations increases, the evolution ability shows a recession phenomenon, resulting in a single population type, and the diversification is not obvious. The possibility of local convergence of the algorithm is large, thereby producing the premature phenomenon. The core idea of the simulated annealing algorithm (simulated annealing, SA) is to find the approximate global solution of the optimization problem by simulating the high-temperature annealing process, jump out of the local optimum, and continue to search until the global optimal solution is obtained.

[0158] In S2, the is used as the fitness function, and the partition result is used as the problem to be solved. The cooling function and Metropolis criterion in the simulated annealing genetic algorithm are expressed as:

[0159] (13);

[0160] (14);

[0161] In the formula, , respectively represent the temperature of the zth generation and the z+1th generation; a is the annealing temperature coefficient, 0 , respectively represent the fitness value of the zth generation and the z+1th generation; P represents the acceptance probability in the simulated annealing algorithm from the zth generation to the z+1th generation;

[0162] A hybrid selection mechanism driven by energy entropy is designed, and the energy entropy calculation formula is as follows:

[0163] (15);

[0164] In the formula, Energy entropy of the current population, which measures the diversity of the population. The greater the value, the higher the diversity. Boltzmann probability of individual u, that is, the probability of individual u being selected at the current temperature The probability of being selected next time; M is the population size, that is, the total number of individuals in the population. 、 The individual energy of individual u and individual v, respectively, that is, the inverse of the fitness value; ln represents the logarithm with base e.

[0165] Set the threshold value of energy entropy According to the selection-acceptance plan:

[0166] If , use roulette wheel selection, which focuses on diversity; if , use tournament selection, which focuses on convergence.

[0167] Roulette wheel selection is a selection method based on the proportion of fitness. The probability of each individual being selected is proportional to its fitness value. The higher the fitness value of an individual, the greater the probability of being selected. Tournament selection is a competitive selection method. Each time, a number of individuals are randomly selected from the population for comparison, and the individual with the highest fitness value is selected to enter the next generation population.

[0168] Accept the new solution in combination with the Metropolis criterion, but the acceptance probability is negatively related to the energy entropy, which is expressed as:

[0169] (16);

[0170] In the formula, is the energy difference between the new solution and the current solution; is the maximum possible value of energy entropy, which is used to normalize .

[0171] The threshold value , is an empirical coefficient, which is adjusted dynamically. The threshold value is set higher in the early stage to maintain diversity, and is reduced in the later stage to speed up convergence.

[0172] In combination with the existing genetic algorithm process, the complete SAGA algorithm flow chart is shown in Figure 2 . SA parameters refer to simulated annealing parameters, and GA parameters refer to genetic algorithm parameters. Each individual (solution) is composed of multiple genes, each of which represents a node partition label. Then, through the merging mutation and splitting mutation of genetic operations, the optimal partition result is finally output. The basis for accepting new solutions is , is the difference between the fitness values of the new solution and the current solution, is the termination temperature.

[0173] In S3, the objective function F of the optimization scheduling model is:

[0174] (17);

[0175] (18);

[0176] wherein, is the weight of the load h, which is located in the partition k; is the load recovery amount of the load h; is the time penalty coefficient, used to balance the recovery amount and the recovery speed; is the total time from the occurrence of the fault to the restoration of power supply; is the start-up time of the black-start power supply; is the critical switch operation time, including the sectionalizing switch, the tie switch, the black-start power supply access switch, the bus tie switch, and the automatic reclosing switch; is the power balance adjustment time within the partition;

[0177] The black-start power supply refers to a power supply that can start generating power without relying on external power supply after the entire power system is powered off due to a fault, and gradually drive other non-self-starting generator units to ultimately restore the entire power system.

[0178] The sectionalizing switch is a device used to divide a power line into several sections, usually installed on a distribution line.

[0179] The tie switch is a switching device that connects two or more distribution lines, allowing different lines to be connected when needed.

[0180] The black-start power supply access switch is a critical device for connecting the black-start power supply (such as hydroelectric power, diesel generators, etc.) to the power system.

[0181] The bus tie switch is a switching device that connects different buses within a substation, used to achieve electrical connection between buses.

[0182] The automatic reclosing switch is a switching device that can automatically reconnect a disconnected line after detecting a transient fault on the line.

[0183] The fast power restoration scheme obtained by solving includes switch operation decisions, which ensure the shortest restoration time under the minimum load cut-off condition and the set constraints; it also includes power restoration sequence, which considers the time required for power warming and climbing time to determine the black-start power supply wake-up sequence; and the transmission of tie lines during inter-zone assistance, which restores the area containing critical loads first to ensure continuous power supply for critical loads.

[0184] The calculation method of is as follows:

[0185] (19);

[0186] In the formula, is the start-up time of the black start power source in the partition k; is a 01 variable indicating whether the black start power source in the partition k is enabled, indicates that the black start power source in the partition k is enabled, indicates that the black start power source in the partition k is not enabled;

[0187] The calculation method of is as follows:

[0188] (20);

[0189] In the formula, is the operation time of the switch s; is a 01 variable indicating whether the switch s is operated, indicates that the switch s is operated, indicates that the switch s is not operated; S is the total number of switches;

[0190] The calculation method of is as follows:

[0191] (21);

[0192] In the formula, is the power supply capacity of the partition k; is the energy storage output in the partition k; is the load demand in the partition k; is the regulation rate of the partition k.

[0193] The constraint conditions of the optimization scheduling model include:

[0194] Power balance constraint:

[0195] (22);

[0196] In the formula, is the load of the partition k at time t; is the wind power output of the partition k at time t; is the photovoltaic output of the partition k at time t; is the charge / discharge power of the energy storage in the partition k at time t; is the transmission power of the partition k at time t and the upper-level power grid; is the power transmission between each partition of the partition k at time t;

[0197] Energy storage power and capacity constraints:

[0198] (23);

[0199] wherein, , respectively is the energy storage capacity at time t, t-1; is the energy storage charge-discharge variable at time t, when the energy storage is in charging state, when the energy storage is in discharging state; is the energy storage charging efficiency; is the time interval; is the energy storage discharging efficiency;

[0200] Voltage and frequency constraints:

[0201] (24);

[0202] wherein, U represents voltage, f represents frequency;

[0203] p.u. (p.u) refers to the index unity, which is a unit for representing the relative value of electrical quantity.

[0204] Energy storage state of charge:

[0205] (25);

[0206] wherein, , is the upper and lower limit of the state of charge; is the initial state of charge; is the state of charge of the energy storage I;

[0207] Zoning autonomy constraints:

[0208] (26);

[0209] wherein, O is the zoning autonomy constraint index; is the available capacity within the zone k; is the maximum load of the zone k.

[0210] In S4, the solving process is:

[0211] S4.1, input the zoning result, that is, input the pre-divided power grid zoning information, including regional node list, source-load-storage configuration parameters and inter-zoning tie line capacity limit;

[0212] S4.2, initialize PSO algorithm parameters, that is, set particle swarm optimization algorithm parameters, including particle swarm size, particle dimension, maximum iteration number, inertia weight initial value, learning factor and variable range;

[0213] S4.3, randomly initialize particle position and velocity, randomly generate initial particle swarm according to variable range, wherein the position of each particle represents a recovery strategy;

[0214] S4.4, calculate particle fitness value, evaluate the performance of each particle based on the objective function and constraint condition of the optimization scheduling model;

[0215] S4.5, update individual historical optimal solution, compare the current fitness value of each particle with its historical optimal value, if better, update the individual historical optimal position and individual historical optimal fitness value;

[0216] S4.6, update group historical optimal solution, select the global optimal solution from all particles, group historical optimal position and group historical optimal fitness value;

[0217] S4.7, judge whether the condition is met, whether the maximum iteration number is reached or the fitness value difference between two iterations is minimum, if the condition is met, output the optimal solution and terminate iteration, if not, update the position vector and velocity vector of each particle, and return to step S4.4 to continue iteration.

[0218] The solution flowchart is shown in Figure 3 .

[0219] The verification process is: the structure diagram of each subarea of the power grid containing distributed energy storage is constructed, as shown in Figure 4 . Mainly including distributed wind power, photovoltaic and energy storage and conventional load, wherein according to the characteristics of the internal nodes of the main grid, the load, power plant and distributed energy storage are planned to different areas, different colored frame lines represent different subareas, according to the load to be handled by different subareas, the rapid recovery power supply strategy of each subarea is designed to ensure the rapid recovery of power supply for important loads in each subarea.

[0220] The power grid is partitioned by combining the given typical daily data with IEEE39 node system. Considering the electrical coupling degree, node scale, network structure and power balance index 4 elements. Different weight combinations will produce different power grid subarea results, in the case of no emphasis, to ensure performance, the weights can be set the same. Scheme 1: based on the source-load comprehensive index without considering energy storage for power grid partitioning, scheme 2: based on the comprehensive evaluation index considering energy storage for power grid partitioning. The results of different scheme partitioning are shown in Table 1, the structure diagrams corresponding to the power grid partitioning of scheme 1 and scheme 2 are shown in Figure 5 and Figure 6 .

[0221] Table 1. Results of different schemes of power grid partition

[0222]

[0223] As can be seen from Table 1, the power grid of both schemes is divided into 3 regions, and there is no isolated node, but the node distribution of scheme 2 is more balanced and reasonable than that of scheme 1. Under the same result of power grid partition, scheme 2 considers the power balance of source-load-storage, and the modularity index decreases by 0.02% compared with scheme 1, the regional node structure strength is slightly weaker, but the power balance index increases by 4.2%, the functional index is obviously improved, and the node size index also increases by 8%, and the node distribution becomes more reasonable. The comprehensive index of scheme 1 is 0.7089, and the comprehensive index of scheme 2 is 0.7306, which is 3.1% higher than that of the former. Therefore, the comprehensive index considering the modularity of source-load-storage, the electrical coupling degree index, the network structure index and the node size index can not only guarantee good structure of partition, but also fully utilize the power complementary characteristics among nodes, and can meet the comprehensive requirements of structure and functionality.

[0224] Taking scheme 2 as an example, the comprehensive index iteration curve of GA algorithm and improved SAGA algorithm is shown in Figure 7 , the GA algorithm reaches the maximum value 0.6952 at about the 20th time. The SAGA algorithm reaches the maximum value 0.7192 at about the 10th time, and obviously the improved algorithm is better than GA in optimization speed and result.

[0225] The multi-objective partition fast recovery optimization model (optimization scheduling model) proposed in this embodiment adopts particle swarm optimization algorithm for solution. Taking the power grid partition condition corresponding to scheme 2 as an example, in order to more intuitively compare the effectiveness and superiority of power grid partition, two different schemes are compared. Scheme 1: without considering power grid partition, the power supply is restored, and scheme 2: considering each partition of the power grid, the power supply is restored based on each partition as a unit.

[0226] As shown in Figure 8 , the process of restoring power supply without partitioning takes 7 minutes, and based on each partition as a unit for power supply restoration, the load recovery time is about 5.5 minutes, which is 23% higher than scheme 1, greatly saving the time of restoring power supply.

[0227] The voltage level comparison of the two schemes is shown in Figure 9 . Without considering power grid partition, the average level of system node voltage is low, the minimum value is 0.9134 p.u, and the average voltage fluctuation rate of the system is 5.06%. After large power grid partition, the minimum value of the system node is 0.9291 p.u, and the average voltage fluctuation rate of the system is 4.33%. Therefore, power grid partition has a certain lifting effect on node voltage, and can also inhibit voltage fluctuation to a certain extent.

[0228] Embodiment 2: Based on embodiment 1, the following improvements are made:

[0229] On the basis of formula (8), the energy storage available energy margin correction value of partition k is further calculated , and in the calculation of formula (6), replace with , wherein:

[0230] (27);

[0231] In the formula, is the cycle attenuation coefficient of the energy storage node m, which is related to the battery chemical properties; represents the absolute value of the state of charge change in the current charge and discharge cycle; is the maximum allowable discharge depth threshold; is the total number of energy storage devices in partition k. The energy storage life attenuation is considered synchronously in the partition optimization stage, realizing the dual optimization of power supply recovery economy and device service life.

[0232] Embodiment 3: A rapid power supply recovery device based on source-load-storage collaborative power grid partition optimization, comprising a memory, a processor, and a computer program stored on the memory and capable of running on the processor, and the computer program is executed by the processor to realize the method in embodiment 1 or embodiment 2.

Claims

1. A method for fast power supply recovery based on source-load-storage coordination and power grid partition optimization, characterized in that The method comprises the following steps: S1, respectively constructing the electrical coupling degree index C, the source-load-storage power balance index , the grid structure index , and the node scale index , and comprehensively forming the power grid partition comprehensive index , expressed as: （1）； wherein, , , , are weight coefficients of C, , , , respectively. S2, to As fitness function, and design energy entropy driven hybrid selection mechanism to improve simulated annealing genetic algorithm, using improved simulated annealing genetic algorithm to solve power grid partitioning; S3, on the basis of the power grid partition, taking each partition as a unit, an optimization scheduling model is established to maximize the recovery of important loads; S4, a particle swarm optimization algorithm is used to solve the optimization scheduling model to obtain a rapid power supply recovery scheme.

2. The method of claim 1, wherein, In the S1, the construction process of the electrical coupling degree index C is: The degree of electrical coupling between node i and node j is represented as: （2）； wherein Pijis the power interaction between node i and node j; , Vi and Vjare the node voltages of node i and node j, respectively. In the power grid partition, when node i is located in partition A and node j is located in partition B, the inter-area coupling degree of partition A and partition B is is represented as: （3）； In the formula, is the number of connection nodes between the partition A and the partition B; When both node i and node j are located in partition A, the intra-partition A coupling degree is represented as: （4）； wherein is the number of node pairs within partition A; The electrical coupling degree index C is formed by comprehensive: （5）； wherein is a weight coefficient of is a weight coefficient of 3.The method of claim 1, wherein, In the S1, the source-load-storage power balance index is represented as: （6）； wherein, , , is a weight coefficient; represents the difference of minimum net power of partition k; is the total load of partition k; is the energy storage available energy margin of partition k; is the total capacity of energy storage of partition k; is the supply-demand matching degree of partition k; is the maximum allowed matching deviation; wherein is represented by: （7）； wherein, is the set of generation nodes within partition k, x represents one of the generation nodes; is the set of energy storage nodes within partition k, m represents one of the energy storage nodes; is the set of load nodes within partition k, y represents one of the load nodes; is the output of generation node x; is the discharge power of energy storage node m; is the demand of load node y; is represented as: （8）； wherein is the current state of charge of the energy storage node m; is the minimum allowed state of charge of the energy storage node m; is the capacity of the energy storage node m; is represented as: （9）； In the formula, T is a time window; represents the power generation of the partition k at the time t; represents the power discharge of the energy storage of the partition k at the time t; represents the load demand of the partition k at the time t.

4. The method of claim 1, wherein, In the S1, the grid structure index is represented as: （10）； where Q is a modularity index; when nodes i and j are in partition k, is the electrical distance between nodes i and j, is the power transfer capability between nodes i and j; is a weight coefficient balancing the importance of the topological closeness and the power transfer capability.

5. The method of claim 1, wherein, In the S1, the node scale index is represented as: （11）； wherein represents the average number of nodes: K is the total number of partitions; is the number of nodes of partition k; （12）； In the formula, N is the total number of nodes in the whole network.

6. The method of claim 1, wherein, In S2, with The fitness function is used to find the optimal solution by taking the partitioning result as the problem to be solved. The cooling function and Metropolis criterion in the simulated annealing genetic algorithm are expressed as follows: （13）； （14）； In the formulae, , respectively represent the temperature of the zth generation, the z+1th generation. a is an annealing temperature coefficient, 0 < a < 1; , respectively represent the fitness value of the zth generation, the z+1th generation; P represents the acceptance probability in the simulated annealing algorithm from the zth generation to the z+1th generation; A hybrid selection mechanism driven by energy entropy is designed, and the energy entropy calculation formula is as follows: （15）； wherein, represents the energy entropy of the current population; is the Boltzmann probability of the individual u, i.e. the probability that the individual u is selected at the current temperature M is the population size, i.e. the total number of individuals in the population; , are the individual energies of the individual u, respectively of the individual v, i.e. the inverse of the fitness value. Setting a threshold for energy entropy , according to the selection-accept plan: If roulette wheel selection is employed; if tournament selection is employed. New solutions are accepted with probability combining the Metropolis criterion The negative correlation with energy entropy is expressed as: （16）； wherein is the energy difference between the new solution and the current solution; is the maximum possible value of the energy entropy.

7. The method of claim 1, wherein, In the S3, the objective function F of the optimization scheduling model is: （17）； （18）； wherein, is the weight of the load h, which is located in the partition k, K is the total number of partitions; is the load recovery amount of the load h; is the time penalty coefficient; is the total time from the occurrence of the fault to the restoration of power supply; is the start-up time of the black-start power supply; is the operation time of the key switch, which includes the sectionalizing switch, the tie switch, the black-start power supply access switch, the bus tie switch, and the automatic reclosing switch; is the power balance adjustment time within the partition; The calculation method is as follows: （19）； wherein is the start-up time of the black start power source within the partition k; is a variable for whether the black start power source is enabled in the partition k, and is 01, indicates that the black start power source is enabled in the partition k, indicates that the black start power source is not enabled in the partition k; The calculation method is as follows: （20）； wherein is the operating time of switch s; denotes whether switch s is operated, is a 01 variable, denotes that switch s is operated, denotes that switch s is not operated; S is the total number of switches; The calculation method is as follows: （21）； wherein is the power capability for zone k; is the energy storage output within zone k; is the load demand within zone k; is the regulation rate for zone k.

8. The method of claim 7, wherein, The constraint conditions of the optimization scheduling model include: Power balance constraint: （22）； wherein, is the load of zone k at time t; is the wind power output of zone k at time t; is the photovoltaic power output of zone k at time t; is the charge / discharge power of energy storage within zone k at time t; is the transmission power of zone k to the upper-level power grid at time t; is the power transmission between each zone of zone k at time t; Energy storage power and capacity constraint: （23）； In the formula, , respectively is the energy storage capacity at time t and time t-1; is the energy storage charge-discharge variable at time t, when the energy storage is in a charging state, when the energy storage is in a discharging state; is the energy storage charging efficiency; is the time interval; is the energy storage discharging efficiency; Voltage and frequency constraint: （24）； In the formula, U represents voltage, and f represents frequency; Energy storage state of charge: （25）； wherein , are upper and lower limits of the state of charge; is the initial state of charge; is the state of charge of the energy storage I; Partition autonomy constraint: （26）； In the formula, O is a partition autonomy constraint index; is the available capacity within the partition k; is the maximum load of the partition k.

9. The method of claim 8, wherein, In the S4, the solving process is: S4.1, input the partition result, that is, input the pre-divided power grid partition information, including regional node list, source-load-storage configuration parameters, and partition interconnection line capacity limit; S4.2, initialize the PSO algorithm parameters, that is, set the particle swarm optimization algorithm parameters, including particle swarm size, particle dimension, maximum iteration number, inertia weight initial value, learning factor and variable range; S4.3, randomly initialize particle position and velocity, and randomly generate an initial particle swarm according to the variable range, wherein the position of each particle represents a recovery strategy; S4.4, calculate the particle fitness value, and evaluate the performance of each particle based on the objective function and constraint conditions of the optimization scheduling model; S4.5, update the individual historical optimal solution, compare the current fitness value of each particle with its historical optimal value, and if it is better, update the individual historical optimal position and individual historical optimal fitness value; S4.6, update the group historical optimal solution, and select the global optimal solution from all particles, the group historical optimal position and the group historical optimal fitness value; S4.7, judge whether the condition is met, whether the maximum iteration number is reached or the fitness value difference between two iterations is minimum, if the condition is met, output the optimal solution and terminate iteration, and if the condition is not met, update the position vector and velocity vector of each particle, and return to step S4.4 for iteration.

10. A device for fast power supply recovery based on source-load-storage coordination and power grid partition optimization, characterized in that: The computer program is stored in the memory and can be run on the processor, and the computer program is executed by the processor to realize the method in any one of claims 1-9.

Images