Load recovery method and system for fault area of power distribution network

Abstract

The invention relates to the technical field of micro-grids, and provides a load recovery method and system for a fault area of a power distribution network, and the method comprises the steps: carrying out the network reconstruction and load reduction of the power distribution network after obtaining the disaster information of the power distribution network; modeling the power distribution network by adopting a graph representation learning method, and abstracting the power distribution network into a graph structure; modeling a fault management problem of the power distribution network into a Markov decision process model based on the graph structure; based on a Markov decision process model, rapid load recovery of a micro-grid cluster in a power distribution network fault is realized through cooperative control. According to the method, graph representation learning and a Markov decision process are deeply fused with a micro-grid cluster, a closed-loop scheme of topology dynamic reconstruction-source grid load storage cooperation-important load zero power failure is formed, hardware does not need to be newly added, an existing power distribution automation terminal can be embedded, and the method is simple and convenient to operate. And the toughness, the new energy consumption rate and the key load power supply reliability of the power distribution network under extreme disasters are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of micro-grid, and in particular to a load recovery method and system for a fault area of a power distribution network. BACKGROUND

[0002] With the frequent occurrence of global extreme weather events, strong disturbances such as typhoons, snowstorms, and thunderstorms pose an increasingly serious challenge to the safe operation of power distribution networks. At present, power distribution networks usually adopt a radial structure with "closed-loop design and open-loop operation", relying on a small number of tie switches and sectionalizing switches to achieve fault isolation and load transfer. However, in extreme disaster scenarios, multiple lines fail simultaneously, switch stations lose power, and communication is interrupted, among other phenomena.

[0003] In related technologies, the recovery of a power distribution network treats network reconfiguration, load shedding, and micro-grid scheduling as three independent sub-problems. Network reconfiguration is mostly based on offline optimization or static power flow models, which takes a long time to solve and is difficult to provide a satisfactory reconfiguration scheme within seconds. Moreover, it does not adequately consider the randomness of distributed power output, which can easily lead to reverse power flow overload. Load shedding relies on overcurrent / low voltage protection to act locally or on the experience of dispatchers, lacking quantitative assessment of load importance, voltage sensitivity, and overall network power supply capacity, resulting in "unnecessary load shedding and excessive load shedding" and a high risk of power outage for critical loads such as hospitals and communication base stations. After a disaster, distributed power sources are often forced to go offline due to the lack of reasonable voltage / frequency support paths, resulting in a large amount of idle clean power generation capacity. Meanwhile, there is a lack of unified coordination framework between the power distribution network and the micro-grid (MG), with micro-grids operating independently and failing to form a cluster effect. There is also a lack of unified modeling approach for uncertain factors such as multiple line failures and switch failures, resulting in poor strategy robustness and missing opportunities to utilize black start, island operation, and flexible power support to accelerate system recovery. SUMMARY

[0004] To solve the above technical problems, the first aspect of the present application provides a load recovery method for a fault area of a power distribution network.

[0005] The second aspect of the present application provides a load recovery system for a fault area of a power distribution network.

[0006] The technical solutions adopted by the present application are as follows: The first aspect embodiment of the present application provides a load recovery method for a fault area of a power distribution network, comprising the following steps: receiving and analyzing meteorological information released from a meteorological center; after obtaining disaster information of the power distribution network according to the meteorological information, implementing two control measures of network reconfiguration and load reduction on the power distribution network; modeling the power distribution network by using a graph representation learning method, and abstracting the power distribution network into a graph structure; modeling a fault management problem of the power distribution network into a Markov decision process model based on the graph structure; and realizing rapid load recovery of a microgrid cluster in the fault of the power distribution network through collaborative control based on the Markov decision process model.

[0007] The load recovery method for the fault area of the power distribution network described above further has the following additional technical features: According to one embodiment of the present application, the control measure of implementing network reconfiguration on the power distribution network specifically comprises: establishing an optimization function with the target of minimizing network loss and maximizing distributed power consumption; establishing the relationship between line power adjustment and node injection power based on a sensitivity matrix to evaluate power flow changes; and implementing network reconfiguration on the power distribution network based on key operation constraints according to the optimization function and the power flow changes, wherein the key operation constraints include: maintaining a radial topology structure, node voltage safety constraints, line thermal stability constraints and extreme scenario robustness guarantee.

[0008] According to one embodiment of the present application, the graph representation learning method is used to model the power distribution network and abstract the power distribution network into a graph structure, specifically comprising: abstracting the power distribution network into a graph structure G=(N, E), wherein: the node set N includes all bus nodes, load nodes, distributed energy access nodes and zero-power injection nodes in the power distribution network; the edge set E includes: physical connections of power distribution lines and internal transformers; the edge set E is divided into two functional subsets: controllable lines with switching capability, including sectionalizing switch lines and tie switch lines, whose on-off states are adjustable; and fixed connection lines, whose connection states are immutable.

[0009] According to one embodiment of the present application, the fault management problem of the power distribution network is modeled into a Markov decision process model based on the graph structure, specifically comprising: generating a state K according to the node set N and the edge set E of the graph structure; designing a combination of all possible switch and load operations in the power distribution network to generate an action B; learning the transition probability P of the power distribution network state in a data-driven manner tr ; setting a reward function O with the target of maximizing power supply and minimizing voltage constraint out-of-limit; and generating a Markov decision process model M according to .

[0010] According to one embodiment of the present application, based on the Markov decision process model, the micro-grid cluster realizes the load rapid recovery in the power distribution network fault through cooperative control, specifically including: acquiring the fault information in real time to determine the fault partition, and controlling the system to operate the tie switch to disconnect the electrical connection on both sides of the fault area, forming physical isolation; generating the recovery capability evaluation S of the power distribution network based on the state K of the Markov decision process model; acquiring the load recovery optimization model, solving the load recovery optimization model, and generating the load recovery strategy.

[0011] The second aspect embodiment of the present application provides a load recovery system of a power distribution network fault area, including: an analysis module, the analysis module is used for receiving and analyzing meteorological information from a meteorological center; a control module, the control module is used for implementing two control measures of network reconstruction and load reduction on the power distribution network after obtaining disaster information of the power distribution network according to the meteorological information; a learning module, the learning module is used for modeling the power distribution network by using a graph representation learning method, and abstracting the power distribution network into a graph structure; a modeling module, the modeling module is used for modeling the fault management problem of the power distribution network into a Markov decision process model based on the graph structure; a cooperative module, the cooperative module is used for realizing the load rapid recovery of the micro-grid cluster in the power distribution network fault through cooperative control based on the Markov decision process model.

[0012] The load recovery system of the power distribution network fault area has the following additional technical features: According to one embodiment of the present application, the control module is specifically used for: establishing an optimization function with the target of minimizing network loss and maximizing distributed power consumption; based on the sensitivity matrix, establishing the relationship between the line power adjustment amount and the node injection power to evaluate the power flow change; according to the optimization function and the power flow change, implementing network reconstruction on the power distribution network based on key operation constraints, wherein the key operation constraints include: radial topology structure maintenance, node voltage safety constraint, line thermal stability constraint and extreme scenario robustness guarantee.

[0013] According to one embodiment of the present application, the learning module is specifically used for: abstracting the power distribution network into a graph structure G=(N,E), wherein: the node set N includes all bus nodes, load nodes, distributed energy access nodes and zero power injection nodes in the power distribution network; the edge set E includes: the physical connection of the distribution line and the internal transformer; the edge set E is divided into two functional subsets: controllable lines with switching capability, including sectionalizing switch lines and tie switch lines, whose on-off state is adjustable; fixed connection lines, whose connection state is invariable.

[0014] According to one embodiment of the present application, the modeling module is specifically used for: generating a state K according to a node set N and an edge set E of the graph structure; designing a combination of all possible switch and load operations in the power distribution network to generate an action B; learning a transition probability P of the power distribution network state in a data-driven manner tr ; setting a reward function O with the target of maximizing power supply and minimizing voltage constraint out-of-limit; and generating a Markov decision process model M according to .

[0015] According to one embodiment of the present application, the coordination module is specifically used for: acquiring fault information in real time to determine a fault partition, and controlling a system to operate a tie switch to disconnect electrical connections on both sides of the fault area to form physical isolation; generating a recovery capability evaluation S of the power distribution network based on the state K of the Markov decision process model; and acquiring a load recovery optimization model, solving the load recovery optimization model, and generating a load recovery strategy.

[0016] The present application has the following beneficial effects: The present application deeply integrates "graph representation learning + Markov decision process" with micro-grid clusters to form a closed-loop scheme of "topology dynamic reconstruction-source network load storage coordination-important load zero power outage", which does not need to add new hardware, can be embedded in existing power distribution automation terminals, and significantly improves the resilience, new energy consumption rate and key load power supply reliability of the power distribution network under extreme disasters. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a flowchart of a load recovery method for a fault area of a power distribution network according to one embodiment of the present application; Figure 2 is a block schematic diagram of a load recovery system for a fault area of a power distribution network according to one embodiment of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0019] Figure 1 is a flowchart of a load recovery method for a fault area of a power distribution network according to one embodiment of the present application, and the method comprises the following steps: Step S1, receiving and analyzing meteorological information released from a meteorological center.

[0020] Specifically, when an extreme disaster strikes the power distribution network, first, the multi-source data fusion mechanism is started under the "cloud-edge-end" collaborative perception framework: the cloud high-performance server cluster receives and analyzes the kilometer-level grid forecast data from the meteorological center in real time through the power dedicated communication network. The data dimension not only covers traditional meteorological elements such as typhoon path, wind speed, and icing thickness, but also extends to derived parameters such as precipitation probability, temperature gradient, and sunshine intensity, providing macro and prior environmental situational awareness for predicting the physical impact of disasters on power distribution lines, towers and substation facilities.

[0021] Step S2, after obtaining the disaster information of the power distribution network according to the meteorological information, the power distribution network is implemented with network reconfiguration and load reduction control measures.

[0022] Specifically, as an emergency control means, voltage collapse and large-scale power outage are avoided. When the power distribution network encounters an extreme disaster (such as typhoon, snowstorm, etc.), network reconfiguration (DNR) and load shedding are two key control measures.

[0023] Further, in the power distribution network failure scenario caused by extreme disasters, network reconfiguration as a key control means, by adjusting the state of sectionalizing switches and tie switches in the power distribution network, dynamically changes the network topology, and then realizes the optimal distribution of power flow. Its core goals include three aspects: first, by transferring the load from the overloaded line to the lightly loaded line, effectively reducing the risk of line overload, avoiding the chain failure caused by local overload; second, after the fault element is effectively isolated, by quickly reconfiguring the network topology, the power supply in the non-fault area is maximized, and the system power supply reliability is significantly improved; finally, by optimizing the grid connection path of distributed power, promoting the efficient local consumption of new energy, and effectively inhibiting the impact of reverse power flow on the main network operation.

[0024] In an embodiment of the present application, the control measure of implementing network reconfiguration for the power distribution network specifically includes the following steps S21-S23: S21, an optimization function is established to minimize network loss and maximize distributed power consumption.

[0025] Specifically, the following formula can be used to establish the optimization function: ; Wherein, , are the line self-inductance and mutual inductance, respectively; , are the active and reactive power of node , , are the active and reactive power of node Active and reactive power; This is the status variable for the tie switch, and its value can be 0 (open) or 1 (closed). This is the distributed power generation absorption penalty coefficient, used to balance absorption level and economic efficiency.

[0026] S22, based on the sensitivity matrix, establishes the relationship between line power adjustment and node injected power to assess power flow changes.

[0027] Specifically, the relationship between line power adjustment and node injected power can be expressed as:

[0028] in, Inject power into the node, This is the line power adjustment amount. This is a sensitivity matrix, reflecting the node... The effect of injection power changes on the line The linear influence of active power provides a theoretical basis for the rapid generation of reconfiguration strategies.

[0029] S23. Based on the optimization function and power flow changes, network reconfiguration is implemented for the distribution network based on key operational constraints, including: radial topology preservation, node voltage safety constraints, line thermal stability constraints, and robustness assurance in extreme scenarios.

[0030] It is important to emphasize that the formulation and implementation of distribution network reconfiguration strategies must meet strict operational constraints. Besides the power balance and equipment operation limits implicit in the aforementioned objective function, it is also essential to ensure that the reconfigured network maintains a radial topology, i.e., no closed-loop operation. Simultaneously, all node voltages must be maintained within permissible safe ranges to avoid voltage exceedances. Furthermore, the power flow of all lines after reconfiguration must not exceed their thermal stability limits to prevent equipment overload. In scenarios where extreme disasters lead to communication outages or partial switch failures, the reconfiguration model must also possess robustness to cope with incomplete topology information or partial control command failures. This typically requires introducing stochastic optimization or robust optimization methods to model the uncertainty of switch states, thereby ensuring the effectiveness of the reconfiguration strategy in harsh real-world environments.

[0031] Specifically, network reconfiguration must ensure that the following key operational constraints are met: Radiation topology preservation: the reconstructed network must maintain a strict radiation topology, i.e. there is no closed loop running path. This is achieved by the following mechanisms: a graph theory based spanning tree constraint modeling method is adopted, and a node-branch association matrix is constructed to describe the network connectivity. By introducing binary switch state variables and imposing connectivity constraints, it is ensured that there is only one connected path between any node and the root node (substation). The system is configured with a dynamic loop network detection module to calculate the network's fundamental loop matrix in real time, and when a non-zero loop vector is detected, the topology correction mechanism is automatically triggered to preferentially open the redundant tie switch to eliminate the closed loop. This mechanism cooperates with the switch position SOE message uploaded on the edge layer to ensure the real-time accuracy of the topology model, providing a reliable network structure basis for reconstruction decision-making.

[0032] Node voltage safety constraint: all node voltages must be strictly maintained within the allowed safe operating range.

[0033] Line thermal stability constraint: all line powers must be below their thermal stability limits. Based on the line power transmission distribution factor, a power flow safety assessment model is established:

[0034] where the thermal stability limit power is determined by the conductor type, environmental temperature and cooling conditions, and the current line power. The system constructs a "pre-check-real-time monitoring-dynamic correction" triple protection mechanism: before optimization, sensitivity analysis is used to predict overload risk; during operation, the FTU (Feeder Terminal Unit, distribution switch monitoring terminal) is used to monitor the line current in real time, and based on the actual power, the line voltage, the line current, and the phase, when the overload probability exceeds the threshold, the power flow reconstruction is automatically started to eliminate the overload by load transfer or DG output adjustment.

[0035] Robustness guarantee for extreme scenarios: to address complex scenarios such as communication interruption and switch failure under extreme disasters, the system establishes a multi-level robust response mechanism to ensure the reliability of the reconstruction strategy. This mechanism first builds a switch failure probability model based on historical failure data, and uses scenario analysis to generate a typical scenario set covering more than 90% of possible failure modes, providing a data basis for uncertain decision-making. On this basis, a distributed robust optimization framework is established, with the core being a min-max optimization model, and the objective function being the performance guarantee under the worst scenario:

[0036] where X is a set of contingency switching, is a constraint violation penalty term, is a distributed power consumption penalty coefficient, by minimizing the system performance loss in the worst case and the constraint violation penalty term, to ensure the effectiveness of the strategy in the uncertainty set When communication interruption occurs, the edge computing node can autonomously enable the pre-stored typical reconstruction scheme library and execute the preset bottom-line recovery strategy to maintain basic power supply capability. At the same time, the system strengthens the design of the network topology through k-edge connectivity analysis in graph theory, ensures that the network still has the necessary reconstruction capability under the condition of partial switch failure, establishes a resilient architecture that meets the N-1 safety criterion, and thus forms a comprehensive robustness guarantee system from prediction optimization to emergency execution, from algorithm design to topology strengthening.

[0037] In the present application, the load shedding measure refers to a key control means for maintaining system stable operation by actively cutting off part of non-critical load when extreme disaster leads to insufficient power supply capacity of the power grid. Its core objectives include: maintaining system voltage stability, preventing node voltage from exceeding the limit; prioritizing critical load power supply to ensure continuous operation of important loads such as hospitals, communication base stations, and emergency facilities.

[0038] To achieve the above objectives, the present application constructs an optimization function with the objective of minimizing the load shedding cost, and its mathematical expression is as follows:

[0039] wherein, represents the unit reduction cost of load , which reflects its power supply priority; and respectively represent the active power values of load before and after reduction.

[0040] The optimization process needs to satisfy the following constraint conditions: power balance, voltage constraint and line capacity constraint.

[0041] The load shedding and mobile energy storage pre-layout are optimized cooperatively, a reasonable load control strategy is formulated in the pre-disaster stage, thereby effectively reducing the total system operation cost and significantly improving the elastic recovery capability of the power system under extreme disaster, on the premise of ensuring system voltage stability and critical load power supply.

[0042] Step S3: using a graph representation learning method to model the distribution network and abstracting the distribution network into a graph structure.

[0043] In an embodiment of the present application, a graph representation learning method is used to model the distribution network and abstract the distribution network into a graph structure, which specifically includes the following steps S31-S32: S31, abstract the power distribution network as a graph structure G=(N, E), wherein: the node set N includes all bus nodes, load nodes, distributed energy access nodes and zero power injection nodes in the power distribution network; the edge set E includes: the physical connection of the distribution line and the inner link transformer, which fully reflects the connection relationship of the power distribution network in space.

[0044] S32, divide the edge set E into two functional subsets: controllable lines with switching capability, including sectionalizing switch lines and tie switch lines, whose on-off state is adjustable; fixed connection lines, whose connection state is unchangeable.

[0045] In the power distribution network with switching control capability, the core of the interruption management mechanism can be formalized as a dynamic learning and active reconstruction process of the network topology. Based on this, the present application further divides the edge set E into two functional subsets: controllable lines with switching capability, including sectionalizing switch lines and tie switch lines, whose on-off state is adjustable, which constitutes the operation basis of network reconstruction; fixed connection lines, whose connection state is unchangeable, which forms the topological skeleton of the network.

[0046] The node set N adopts three types of data fusion mechanism: predicted variables: including the predicted values of active / reactive power demand or power generation of each node; estimated variables: power-related parameters obtained based on state estimation; measured variables: real-time measurement data of three-phase voltage of each node.

[0047] In terms of the edge set E, by introducing a power flow simulator as a simulation tool, the branch power flow measured value consistent with the real system operation characteristics is obtained, which provides data support for system state analysis.

[0048] From the perspective of graph theory, the power distribution network reconstruction problem can be transformed into an optimization decision problem of controllable edge state. In the main power distribution network level, the emergency load shedding control strategy is mathematically characterized by binary variables associated with switchable load nodes, realizing the rapid recovery and stable operation of system supply and demand balance under fault or extreme conditions.

[0049] The present application builds a graph representation model of the power distribution system, establishes a classification system of nodes and edges and a variable representation mechanism, forms an analysis framework with clear structure, data-driven and good scalability. This method fully excavates the internal correlation between the topological characteristics and operation data of the power distribution system, provides a model basis for the integration of intelligent algorithms such as graph neural network in the subsequent stage, and significantly improves the optimization efficiency and decision accuracy of the power distribution network reconstruction and emergency control strategy.

[0050] From the graph perspective, the reconfiguration of distribution network is essentially to solve the state (closed or open) problem of switchable edges (i.e. lines with on-off control capability). At the level of main distribution network, the emergency load shedding control strategy can be mathematically characterized by the binary variables associated with the switchable load nodes.

[0051] Step S4, based on the Markov decision process model, the load of the micro-grid cluster is quickly restored in the power distribution network fault through cooperative control.

[0052] In an embodiment of the present application, the fault management problem of the power distribution network is modeled as a Markov decision process model based on the graph structure, specifically including the following steps S41-S45: S41, generating a state K according to the node set N and the edge set E of the graph structure.

[0053] (1) Specifically, the state K is composed of the power distribution network related observations reflecting the current operation state of the network, including the node set N, the edge set E and the network topology.

[0054]

[0055] Wherein, PNd , QNd represent the estimated or predicted active and reactive demand of the node, PNg , QNg correspond to the active and reactive power generation of the node. The three-phase voltage measured at the bus is represented as V N , V viol represent the voltage violation in the network. The edge nodes include: the power flow through the network branch, denoted as l The running topology of the network is t , and the total energy supplied in the network is represented by E supp The variable O in the state K encapsulates the outage scenario, i.e. multi-line fault in the network, including the outage of the switch. The inoperability of the outage switch is solved by using a shielding mechanism that suppresses the corresponding switching action, denoted by μ.

[0056] S42, designing all possible combinations of switch and load operation in the power distribution network to generate action B.

[0057] Specifically, the action B is the control action of emergency response, including switch operation and load shedding, including the number of switchable lines (including sectional lines and tie lines), the number of switchable loads in the network. The line switch state is represented by a binary variable (0 for open, 1 for closed).

[0058] S43, learning the transition probability P of the power distribution network state in a data-driven mannertr .

[0059] Transition probability P tr Capture the dynamic nature of the network with emergency response, denoted as: , represents the transition from the network state at time step t to the network state at step (t+1), , represents the network state at time step (t+1), , represents the network state at time step t, , represents the action at time step t. The transition probability is learned by the agent from its interaction with the environment.

[0060] S44, with the objective of maximizing power supply and minimizing voltage constraint violations, the reward function O is set.

[0061] Specifically, the reward function O adopts a GRL (Graph Representation Learning) algorithm to take optimal control actions to alleviate power distribution network failures, which takes the form of: , represents the power supply, V viol , represents the voltage violation in the network, and the reward reflects the objective of improving the resilience of the power distribution network by maximizing power supply and minimizing voltage constraint violations. Wherein is a correction term considering the network state under specific fault conditions and switch operation.

[0062] S45, according to Generate Markov decision process model M.

[0063] Step S5, based on the Markov decision process model, realize the rapid load recovery of microgrid cluster in power distribution network failure through cooperative control.

[0064] In one embodiment of the present application, based on the Markov decision process model, the rapid load recovery of microgrid cluster in power distribution network failure is realized through cooperative control, which specifically includes the following steps S51-S53: S51, real-time acquisition of fault information to determine the fault partition, and control system operation to disconnect the electrical connection on both sides of the fault area through the tie switch, forming physical isolation.

[0065] Specifically, the fault information is obtained by using the power distribution network monitoring system, and the fault area is determined by combining the graph theory model. Let the set of power distribution network nodes be N, and the set of lines be E. The fault area can be represented as a subgraph , wherein, represents the fault node, representing the faulted line, , The fault isolation is achieved by the operation of tie switches, and the isolation time needs to be satisfied , is the critical outage time.

[0066] S52, generating the restoration capability S of the power distribution network based on the state K of the Markov decision process model.

[0067] Specifically, the schedulable capacity of each microgrid is calculated : where is the power generation of the microgrid i, is the local load. The total recoverable capacity of the cluster is (n is the number of microgrids in the cluster), and the restoration capability S of the power distribution network is generated according to the total recoverable capacity of the cluster.

[0068] The restoration capability S is actually a state processing function: S = f(K). It converts the original, multi-dimensional state observation data K into a highly condensed index that is crucial for decision-making.

[0069] S53, obtaining a load restoration optimization model, solving the load restoration optimization model based on the action B of the Markov decision process model, and generating a load restoration strategy.

[0070] Specifically, the load restoration optimization model is obtained according to the following formula: ; where is the load priority weight, is the load power, is the restoration state variable.

[0071] The constraint conditions include power balance and power flow constraints , is the line current, is the rated current.

[0072] The model is solved by a distributed algorithm to achieve orderly load restoration.

[0073] The generator of the load restoration strategy based on the action B of the Markov decision process model is an optimization model and a solver. It connects the state K and the action B of the Markov decision process model, and the specific process is: running the load restoration optimization model, which solves a specific, optimal load restoration scheme, i.e., determining which restoration and which non-restoration. The solution of this optimization scheme directly corresponds to a specific action in the action B.

[0074] In summary, according to the load recovery method of the power distribution network fault area, the “graph representation learning + Markov decision process” is deeply integrated with the micro-grid cluster to form a closed-loop scheme of “topology dynamic reconstruction-source network load storage coordination-important load zero power failure”, without the need for new hardware, which can be embedded in the existing power distribution automation terminal, and significantly improves the resilience, new energy consumption rate and key load power supply reliability of the power distribution network under extreme disasters.

[0075] Corresponding to the above-mentioned load recovery method of the power distribution network fault area, the present application also provides a load recovery system for the power distribution network fault area. Since the system embodiment of the present application corresponds to the above-mentioned method embodiment, for details not disclosed in the system embodiment, please refer to the above-mentioned method embodiment, which will not be described in detail in the present application.

[0076] Figure 2 is a block schematic diagram of the load recovery system for the power distribution network fault area according to an embodiment of the present application, as shown in Figure 2 The system comprises an analysis module, a control module, a learning module, a modeling module and a coordination module.

[0077] The analysis module is used to receive and analyze meteorological information from the meteorological center; the control module is used to implement network reconstruction and load reduction control measures on the power distribution network after obtaining disaster information of the power distribution network according to the meteorological information; the learning module is used to model the power distribution network by using a graph representation learning method, and the power distribution network is abstracted as a graph structure; the modeling module is used to model the fault management problem of the power distribution network as a Markov decision process model based on the graph structure; and the coordination module is used to realize the rapid load recovery of the micro-grid cluster in the power distribution network fault through coordinated control based on the Markov decision process model.

[0078] According to an embodiment of the present application, the control module is specifically used for: establishing an optimization function with the goal of minimizing network loss and maximizing distributed power consumption; based on the sensitivity matrix, establishing the relationship between the line power adjustment amount and the node injection power to evaluate the power flow change; based on the optimization function and the power flow change, implementing network reconstruction on the power distribution network based on key operation constraints, wherein the key operation constraints include: radial topology structure maintenance, node voltage safety constraint, line thermal stability constraint and extreme scenario robustness guarantee.

[0079] According to one embodiment of the present application, the learning module is specifically configured to: abstract the power distribution network into a graph structure G=(N, E), wherein: the node set N includes all bus nodes, load nodes, distributed energy access nodes and zero-power injection nodes in the power distribution network; the edge set E includes: physical connections of distribution lines and interlinked transformers; the edge set E is divided into two functional subsets: controllable lines with switching capability, including sectionalizing switch lines and tie switch lines, whose on-off states are adjustable; fixed connection lines, whose connection states are immutable.

[0080] According to one embodiment of the present application, the modeling module is specifically configured to: generate a state K according to the node set N and the edge set E of the graph structure; design a combination of all possible switch and load operations in the power distribution network to generate an action B; learn the transition probability P of the power distribution network state in a data-driven manner tr ; set a reward function O with the goal of maximizing the power supply and minimizing the voltage constraint overrun; and generate a Markov decision process model M according to .

[0081] According to one embodiment of the present application, the coordination module is specifically configured to: acquire fault information in real time to determine a fault partition, and control the system to operate the tie switch to disconnect the electrical connection on both sides of the fault area, forming physical isolation; generate a recovery capability evaluation S of the power distribution network based on the state K of the Markov decision process model; acquire a load recovery optimization model, solve the load recovery optimization model, and generate a load recovery strategy.

[0082] The load recovery system for the fault area of the power distribution network according to the embodiment of the present application deeply integrates “graph representation learning + Markov decision process” and microgrid cluster, forms a closed-loop scheme of “topology dynamic reconstruction-source network load storage coordination-important load zero power outage”, does not need to add new hardware, can be embedded in the existing power distribution automation terminal, and significantly improves the resilience, new energy consumption rate and key load power supply reliability of the power distribution network under extreme disasters.

[0083] In the description of the present application, the terms “first”, “second” are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with “first”, “second” can explicitly or implicitly include one or more of the features. The meaning of “plurality” is two or more, unless otherwise specifically limited.

[0084] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the description of the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples, without contradiction. In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the description of the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples, without contradiction.

[0085] Any process or method descriptions or descriptions of the flow diagrams in the flow charts described herein, or otherwise described in this specification, can be understood as representing the steps of a method or process, including one or more steps for implementing custom logic functions or processes, and the scope of the preferred embodiments of the present application includes additional implementation in which the steps are performed in a different order, including an order that is substantially simultaneous, or in reverse order, depending on the functionality involved, as will be understood by those skilled in the art.

[0086] The logic and / or steps represented in flow diagrams or otherwise described herein, for example, can be considered as a sequence of instructions to implement logic functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this specification, a "computer-readable medium" can be any means that can contain, store, communicate, propagate or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be a machine-readable storage device (e.g., magnetic, optical or other) a machine-readable storage diskette (e.g., floppy disk, optical disk, etc.), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), optical fibers, and a portable compact disc read-only memory (CDROM). Further, the computer-readable medium can even be paper or other suitable medium upon which the program is printed, as the program can be electronically captured, for example via the optical scanner of a device or other electronic capture device, and then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and stored in a computer memory.

[0087] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. As such, in some embodiments, the steps or methods can be implemented in a combination of hardware and software. If implemented in hardware, as in another embodiment, any of the above techniques can be implemented with or without the use of a programmable data processing apparatus, using any of the following technologies: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0088] Those of skill in the art would understand that information and signals can be represented using any of a variety of technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that can be referenced throughout the above description can be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0089] In addition, each function unit in each embodiment of the present application can be integrated in one processing module, or each unit can exist physically independently, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware, or in the form of software function module. When the integrated module is realized in the form of software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0090] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

[0091] Although the embodiments of the present application have been shown and described above, it should be understood that those skilled in the art can make various changes, modifications, replacements and variations to these embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for restoring load in a fault area of ​​a distribution network, characterized in that, Includes the following steps: Receive and parse meteorological information released by the meteorological center; After obtaining disaster information of the distribution network based on the meteorological information, two control measures are implemented for the distribution network: network reconfiguration and load reduction. A graph representation learning method is used to model the distribution network, abstracting it into a graph structure; Based on the graph structure, the fault management problem of the distribution network is modeled as a Markov decision process model; Based on the Markov decision process model, the microgrid cluster can achieve rapid load recovery during distribution network faults through coordinated control.

2. The load restoration method for a faulty area in a distribution network according to claim 1, characterized in that, Control measures for network reconfiguration of the distribution network specifically include: Establish an optimization function with the objectives of minimizing network losses and maximizing the absorption of distributed power sources; Based on the sensitivity matrix, the relationship between line power adjustment and node injected power is established to assess power flow changes; Based on the optimization function and the power flow changes, the distribution network is reconfigured according to key operational constraints, which include: maintaining the radial topology, node voltage safety constraints, line thermal stability constraints, and robustness assurance in extreme scenarios.

3. The load restoration method for a faulty area in a distribution network according to claim 1, characterized in that, The distribution network is modeled using a graph representation learning method, which abstracts the distribution network into a graph structure, specifically including: The power distribution network is abstracted as a graph structure G=(N,E), where: the node set N includes all bus nodes, load nodes, distributed energy access nodes and zero power injection nodes in the power distribution network; the edge set E includes: the physical connection between the power distribution lines and the inline transformers; Divide the edge set E into two functional subsets: controllable lines with switching capabilities, including segmented switch lines and tie switch lines, whose on / off states are adjustable; and fixed connection lines, whose connection states are immutable.

4. The load restoration method for a faulty area in a distribution network according to claim 1, characterized in that, Based on the graph structure, the fault management problem of the distribution network is modeled as a Markov decision process model, specifically including: Generate state K based on the node set N and edge set E of the graph structure; Design all possible combinations of switch and load operations in the distribution network to generate action B; Learning the state transition probability P of the distribution network in a data-driven manner tr ; To maximize power supply And set the reward function O to minimize the voltage constraint overshoot. according to Generate a Markov decision process model M.

5. The load restoration method for a faulty area in a distribution network according to claim 4, characterized in that, Based on the aforementioned Markov decision process model, rapid load recovery of microgrid clusters during distribution network faults is achieved through coordinated control, specifically including: Real-time fault information is acquired to determine the fault zone, and the system operates the interconnection switch to disconnect the electrical connection on both sides of the fault zone, forming physical isolation; The state K is generated based on the Markov decision process model to assess the resilience S of the distribution network. Obtain the load recovery optimization model, solve the load recovery optimization model, and generate a load recovery strategy.

6. A load restoration system for a fault area in a distribution network, characterized in that, include: The parsing module is used to receive and parse meteorological information released by the meteorological center; The control module is used to implement two control measures on the distribution network: network reconfiguration and load reduction, after obtaining disaster information of the distribution network based on the meteorological information. The learning module is used to model the distribution network using a graph representation learning method, abstracting the distribution network into a graph structure. A modeling module is used to model the fault management problem of the distribution network as a Markov decision process model based on the graph structure. The coordination module is used to achieve rapid load recovery of the microgrid cluster in the event of a distribution network fault through coordinated control based on the Markov decision process model.

7. The load restoration system for a faulty area in a distribution network according to claim 6, characterized in that, The control module is specifically used for: Establish an optimization function with the objectives of minimizing network losses and maximizing the absorption of distributed power sources; Based on the sensitivity matrix, the relationship between line power adjustment and node injected power is established to assess power flow changes; Based on the optimization function and the power flow changes, the distribution network is reconfigured according to key operational constraints, which include: maintaining the radial topology, node voltage safety constraints, line thermal stability constraints, and robustness assurance in extreme scenarios.

8. The load restoration system for a faulty area in a distribution network according to claim 6, characterized in that, The learning module is specifically used for: The power distribution network is abstracted as a graph structure G=(N,E), where: the node set N includes all bus nodes, load nodes, distributed energy access nodes and zero power injection nodes in the power distribution network; the edge set E includes: the physical connection between the power distribution lines and the inline transformers; Divide the edge set E into two functional subsets: controllable lines with switching capabilities, including segmented switch lines and tie switch lines, whose on / off states are adjustable; and fixed connection lines, whose connection states are immutable.

9. The load restoration system for a faulty area in a distribution network according to claim 6, characterized in that, The modeling module is specifically used for: Generate state K based on the node set N and edge set E of the graph structure; Design all possible combinations of switch and load operations in the distribution network to generate action B; Learning the state transition probability P of the distribution network in a data-driven manner tr ; To maximize power supply And set the reward function O to minimize the voltage constraint overshoot. according to Generate a Markov decision process model M.

10. The load restoration system for a faulty area in a distribution network according to claim 9, characterized in that, The collaboration module is specifically used for: Real-time fault information is acquired to determine the fault zone, and the system operates the interconnection switch to disconnect the electrical connection on both sides of the fault zone, forming physical isolation; The state K is generated based on the Markov decision process model to assess the resilience S of the distribution network. Obtain the load recovery optimization model, solve the load recovery optimization model, and generate a load recovery strategy.

Images