Power distribution network self-healing control method and device based on information geometry, equipment and medium

Abstract

This application relates to the field of smart grid and power system automation technology, and discloses a distribution network self-healing control method, device, equipment, and medium based on information geometry. The method includes: modeling the distribution network's operating state into a parameterized probability distribution based on real-time operating state data to construct a statistical manifold; calculating the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network; comparing the Ricci scalar curvature with a preset negative threshold to determine whether there are signs of instability in the distribution network; when signs of instability exist, determining a geodesic line on the statistical manifold connecting the current state point to a preset target stable state point as the optimal recovery path to generate the optimal control law, and mapping it to the target control command of the distribution network. This method not only captures signs of instability but also provides a theoretically most efficient and stable recovery trajectory, and exhibits stronger adaptability and robustness to random disturbances.

Description

Technical Field

[0001] This application relates to the field of smart grid and power system automation technology, and in particular to a method, device, equipment and medium for self-healing control of distribution networks based on information geometry. Background Technology

[0002] With the high proportion of distributed generation (DG) integration, the safe operation of distribution networks faces significant challenges. Traditional fault handling and self-healing strategies for distribution networks are mostly based on rule-based logic or conventional optimization algorithms, which have two inherent limitations: First, response lag: most are "post-event" responses, meaning the recovery process can only be initiated after a fault occurs, making it difficult to accurately detect fault precursors and implement preventative control. Second, model dependence: they heavily rely on accurate physical models, and their robustness and adaptability are insufficient in the face of the high uncertainty brought about by high-proportion DG integration.

[0003] In recent years, although some studies have attempted to introduce machine learning (especially reinforcement learning) for data-driven intelligent control, it is generally regarded as a "black box" model, lacking interpretable physical meaning and exhibiting unstable learning processes. Therefore, there is an urgent need for a new self-healing control method that can address the problems of traditional self-healing control, such as the difficulty in accurately capturing and preventing fault precursors, and its insufficient robustness and adaptability in the face of high uncertainty brought about by high-proportion distributed generation (DG) access. Summary of the Invention

[0004] The main objective of this application is to provide a distribution network self-healing control method, device, equipment, and medium based on information geometry, which can solve the problems of traditional self-healing control in the prior art, such as difficulty in accurately capturing and preventing fault precursors, and insufficient robustness and adaptability in the face of high uncertainty brought about by high proportion of distributed generation (DG) access.

[0005] To achieve the above objectives, the first aspect of this application provides a self-healing control method for distribution networks based on information geometry, the method comprising:

[0006] Obtain real-time operating status data of the power distribution network;

[0007] Based on the real-time operating status data, the operating status model of the distribution network is transformed into a parameterized probability distribution, and a statistical manifold is constructed based on the parameters of the probability distribution.

[0008] Calculate the Ricci scalar curvature at the current state point on the statistical manifold that corresponds to the current operating state of the distribution network;

[0009] The Ricci scalar curvature is compared with a preset negative threshold to determine whether there are signs of instability in the power distribution network.

[0010] When it is determined that there are signs of instability, a geodesic line connecting the current state point to a preset target stable state point is determined on the statistical manifold as the optimal recovery path.

[0011] An optimal control law is generated based on the optimal recovery path, and the optimal control law is mapped to a target control command for the distribution network. The target control command is then executed. The optimal control law is used to drive the operating state of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network. The target control command is used to control one or more control resources of the distribution network.

[0012] In one feasible implementation, the parameterized probability distribution is the joint probability distribution of key electrical quantities in the distribution network;

[0013] The key electrical quantities include at least one of the following: node voltage amplitude, phase angle, active power of distributed generation, reactive power of distributed generation, and branch current.

[0014] The parameters of the probability distribution are composed of the statistical moments of the key electrical quantities.

[0015] In one feasible implementation, the geometric metric of the statistical manifold is the Fisher information matrix. Then, calculating the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network includes:

[0016] Based on the Fisher information matrix of the current state point and its derivative, calculate the Krzy symbol, and then calculate the Ricci scalar curvature of the current state point based on the Krzy symbol.

[0017] In one feasible implementation, mapping the optimal control law to the target control command of the distribution network includes:

[0018] The target control command is obtained using a pre-established sensitivity matrix and the optimal control law; the sensitivity matrix is ​​used to reflect the mapping relationship between the parameter changes of the probability distribution corresponding to the control law and the physical control quantity corresponding to the control command of the control resource.

[0019] In one feasible implementation, the optimal control law includes the following mathematical expression:

[0020] ;

[0021] In the formula: Let θ(t) be the optimal control law for the current state point θ(t), where the state point is represented by the parameters of the probability distribution. For control gain matrix; It measures the difference between the current state point θ(t) and the optimal recovery path θ.geo Lyapunov function of the deviation between; It represents the covariant derivative on the statistical manifold.

[0022] In one feasible implementation, the target control instruction includes the following mathematical expression:

[0023] ;

[0024] In the formula, It is a sensitivity matrix The false reversal; The target control command is a vector containing specific control quantities; Let θ(t) be the optimal control law for the current state point θ(t).

[0025] In one feasible implementation, the control resources include at least one of distributed power sources, energy storage systems, flexible loads, or electric vehicle charging station clusters.

[0026] To achieve the above objectives, a second aspect of this application provides a power distribution network self-healing control device based on information geometry, the device comprising:

[0027] Data acquisition module: used to acquire real-time operating status data of the power distribution network;

[0028] Manifold determination module: used to model the operating state of the distribution network into a parameterized probability distribution based on the real-time operating state data, and to construct a statistical manifold based on the parameters of the probability distribution;

[0029] Curvature determination module: used to calculate the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network;

[0030] Instability detection module: used to compare the Ricci scalar curvature with a preset negative threshold to determine whether there are signs of instability in the distribution network;

[0031] Path determination module: When it is determined that there are signs of instability, a geodesic line connecting the current state point to a preset target stable state point is determined on the statistical manifold as the optimal recovery path;

[0032] The self-healing control module is used to generate an optimal control law based on the optimal recovery path, map the optimal control law to a target control command for the distribution network, and execute the target control command. The optimal control law is used to drive the operating state of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network. The target control command is used to control one or more control resources of the distribution network.

[0033] To achieve the above objectives, a third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method shown in the first aspect.

[0034] To achieve the above objectives, a fourth aspect of this application provides a computer device including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method shown in the first aspect.

[0035] The following benefits can be obtained by adopting this application:

[0036] This application provides a self-healing control method for distribution networks based on information geometry. The method includes: acquiring real-time operating status data of the distribution network; modeling the operating status of the distribution network into a parameterized probability distribution based on the real-time operating status data, and constructing a statistical manifold based on the parameters of the probability distribution; calculating the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating status of the distribution network; comparing the Ricci scalar curvature with a preset negative threshold to determine whether there are signs of instability in the distribution network; when it is determined that there are signs of instability, determining a geodesic line on the statistical manifold connecting the current state point to a preset target stable state point as the optimal recovery path; generating an optimal control law based on the optimal recovery path, mapping the optimal control law to a target control command for the distribution network, and executing the target control command; the optimal control law is used to drive the operating status of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network; the target control command is used to control one or more control resources of the distribution network.

[0037] The above method achieves several advantages. First, it allows for the assessment of system stability by calculating the Ricci scalar curvature in the information geometry space. Comparing the Ricci scalar curvature with a preset negative threshold determines the presence of instability precursors in the distribution network. This approach captures instability precursors at the level of abrupt changes in the inherent geometric properties of the system's state space. Compared to traditional criteria relying on electrical quantity limits, this method predicts critical states earlier and more fundamentally, achieving proactive preventative control and providing a forward-looking early warning advantage. Second, it transforms the optimal recovery problem into finding geodesics on a statistical manifold, providing a theoretically most efficient and stable recovery trajectory for self-healing control. This avoids the potential for traditional optimization algorithms to get stuck in local optima or converge slowly, achieving optimal recovery. Third, the entire framework of this application is built upon probability distributions and statistical manifolds. Uncertainties in the power grid are naturally inherent in the geometric structure of the manifold, thus giving this method greater adaptability and robustness to random disturbances. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] in:

[0040] Figure 1 This is a flowchart of a power distribution network self-healing control method based on information geometry in an embodiment of this application;

[0041] Figure 2 This is a schematic diagram of the structure of a power distribution network self-healing control system based on information geometry in an embodiment of this application;

[0042] Figure 3 This is a structural block diagram of a power distribution network self-healing control device based on information geometry, as described in an embodiment of this application.

[0043] Figure 4 This is a structural block diagram of the computer device in the embodiments of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0045] It should be noted that this application describes a self-healing control method for distribution networks based on information geometry. This method aims to perform in-depth analysis and control of the operating state of the distribution network from a geometric perspective, thereby achieving proactive early warning of system instability risks and optimized self-healing recovery. It can be understood that this method as a whole constitutes a continuously operating closed-loop monitoring and control process.

[0046] This paper presents a self-healing control method and system for distribution network resilience based on information geometry and dissipative structure theory, providing a novel theoretical paradigm for solving the resilience problem of complex power grids. This method innovatively abstracts the dynamic evolution process of the distribution network as the movement of system state points on a curved geometric space—a statistical manifold—composed of probability distributions.

[0047] This method first constructs a corresponding statistical manifold by probabilistically modeling the power grid state, and uses the Fisher information matrix as its metric. By calculating the Ricci scalar curvature defined by Riemannian geometry on this manifold in real time, this application can capture abrupt changes in the intrinsic geometric properties of the system's state space. A significantly negative curvature value is identified as a key precursor to an impending phase transition or instability in the system, thus achieving early warning of faults.

[0048] In the self-healing control phase, this application treats the distribution network as an open system far from equilibrium. The essence of resilient self-healing is interpreted as the system actively controlling itself (e.g., utilizing distributed energy sources) to absorb negative entropy flow from the outside to resist internal entropy increase, thereby self-organizing to form a new ordered and stable structure—a dissipative structure. This method equates the optimal self-healing control path to guiding the system state point along the shortest path from the fault point to the target stable point on the statistical manifold—the geodesic. By solving the geodesic equations and generating the corresponding control law, this application can guide the power grid to complete post-fault reconstruction and recovery in the most efficient and stable manner.

[0049] Understandably, this application aims to address the fundamental theoretical bottlenecks of existing distribution network resilience self-healing control methods. Current methods either rely excessively on precise physical models, resulting in poor adaptability to unknown disturbances, or fall into the dilemma of data-driven black-box decision-making, lacking interpretability and safety guarantees. Especially when dealing with unknown black swan events caused by the coupling of new energy uncertainties and extreme disasters, existing technologies struggle to achieve effective early warning and optimal recovery for system instability.

[0050] The purpose of this application is to overcome the shortcomings of existing technologies and provide a self-healing control method for distribution network resilience based on information geometry and dissipative structure theory. This method aims to achieve forward-looking prediction of the system's instability critical point and guide the system to evolve along the optimal path on the probabilistic manifold, thereby achieving rapid and efficient self-healing and significantly improving the resilience of the distribution network.

[0051] The following section details a self-healing control method for power distribution networks based on information geometry, as proposed in this application.

[0052] Please see Figure 1 , Figure 1 This is a flowchart illustrating a power distribution network self-healing control method based on information geometry, as described in this application. This method can be applied to either a terminal or a server. The terminal can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet, or laptop. The server can be a standalone server or a server cluster composed of multiple servers. This embodiment uses server application as an example. Figure 1 The method shown includes the following steps:

[0053] 101. Obtain the real-time operating status data of the power distribution network;

[0054] Understandably, to achieve self-healing control of the distribution network, it is necessary to monitor the network's status and implement timely self-healing control. Therefore, the first step is to acquire real-time operational status data of the distribution network. This real-time operational status data includes several key electrical quantities, i.e., acquiring real-time data characterizing the distribution network's operational status. (See also...) Figure 2 , Figure 2 This is a schematic diagram of a distribution network self-healing control system based on information geometry, as described in an embodiment of this application. In modern distribution networks 40, a large number of synchronous phasor measurement units (PMUs) are typically deployed to achieve high-precision state awareness. In a specific embodiment of this application, real-time data acquisition is accomplished through PMUs deployed at key nodes and branches of the distribution network. These sensors 41 can synchronously collect dynamic information of the power grid with extremely high temporal resolution (e.g., 50 or 100 samples per second). The collected real-time data constitutes a high-dimensional microscopic state vector, which contains multiple key electrical quantities. As an optional implementation, these key electrical quantities may include, but are not limited to: voltage amplitude and phase angle at each node, active and reactive power of distributed power sources (such as photovoltaic and wind power) connected to the grid, and branch currents on key lines. These high-frequency, synchronous measurement data provide rich and dynamic raw information for subsequent probabilistic modeling, forming the basis for accurately grasping the random fluctuation characteristics of the power grid.

[0055] 102. Based on the real-time operating status data, the operating status model of the distribution network is transformed into a parameterized probability distribution, and a statistical manifold is constructed based on the parameters of the probability distribution;

[0056] Furthermore, a statistical manifold is constructed: specifically, based on the real-time operating status data, the operating status of the distribution network is modeled as a parameterized probability distribution, and a statistical manifold is constructed based on the parameters of the probability distribution.

[0057] In step 102, a statistical manifold is constructed based on the acquired real-time data. This step forms the theoretical basis of the technical solution of this application. Its core idea is to map the uncertain power grid operating state into a definite, analyzable geometric object. Specifically, the system models the high-dimensional microstate vector (i.e., time series data of key electrical quantities) collected in step 101, which varies over time and contains random disturbances, into a parameterized probability distribution. For example, it can be assumed that within a certain short time window, the fluctuations of these electrical quantities follow a joint probability distribution function p(x;θ), where x is the microstate vector and θ is a parameter vector composed of several parameters. Here, the parameter vector θ is no longer the instantaneous electrical quantity itself, but a macroscopic description of the statistical characteristics of these electrical quantities over a period of time. In a specific implementation, the parameter vector θ can be composed of the statistical moments of the selected key electrical quantities, such as the first moment (mean), the second central moment (variance), and the covariance between different electrical quantities.

[0058] The set of all possible parameter vectors θ constitutes a high-dimensional parameter space. Due to the nature of probability distributions, this parameter space is not a flat Euclidean space, but a geometric space with an inherent curved structure, which is called a statistical manifold M in information geometry theory. To perform geometric analysis on this manifold M, its metric needs to be defined, that is, the way to measure the "distance" between two points on the manifold. In a preferred embodiment of this application, the geometric metric of the statistical manifold M is defined by the Fisher information matrix (FIM) g(θ), which, as a Riemannian metric tensor, determines the distance and angle between points on the manifold. FIM matrix elements The calculation formula is:

[0059] ;

[0060] In the formula: This represents the mathematical expectation of a probability distribution; and These are the two components of the parameter vector θ; Quantitatively characterizing macroscopic parameters Distribution of the system's microstate caused by minute changes The degree of difference This is the symbol for partial differentials.

[0061] In this process, the Fisher information matrix plays the role of a metric tensor in geometry. It defines the local geometry of each point on the manifold M, enabling the calculation of geometric quantities such as length, angle, and curvature. Through this step, the dynamic evolution of the power grid is abstracted as the trajectory of the state point θ(t) on the statistical manifold M.

[0062] In one feasible implementation, the parameterized probability distribution is the joint probability distribution of key electrical quantities in the distribution network;

[0063] The key electrical quantities include at least one of the following: node voltage amplitude, phase angle, active power of distributed generation, reactive power of distributed generation, and branch current.

[0064] The parameters of the probability distribution are composed of the statistical moments of the key electrical quantities.

[0065] For example, steps 101 and 102 are for constructing the probabilistic manifold (i.e., statistical manifold) of the distribution network state, including two key steps, as follows:

[0066] 1) Define the system state probability distribution:

[0067] The distribution network at time The operational status is described by high-frequency measurement data collected by devices such as phasor measurement units (PMUs) and smart meters deployed in the network. A definition is made of... The microstate vector is composed of key electrical quantities. ,in, , For nodes The voltage amplitude and phase angle, , For distributed power sources Contributions, whether meritorious or not, branch road The current.

[0068] Due to uncertainties such as fluctuations in new energy sources and randomness in load, It is a random vector. Its statistical properties are determined by a parameterized joint probability distribution function. Description. Among them, This is a set of macroscopic state parameter vectors, which are composed of the statistical moments (such as mean, variance, and covariance) of these key electrical quantities:

[0069] ;

[0070] In the formula: This constitutes the coordinates of the system in the low-dimensional macroscopic parameter space, with each component being a corresponding electrical quantity.

[0071] 2) Constructing the statistical manifold and defining its metric:

[0072] All possible parameter vectors The set constitutes a Statistical manifold of dimension The intrinsic geometry of this manifold space is determined by the Fisher Information Matrix (FIM). Defined as a Riemannian metric tensor, it determines the distances and angles between points on a manifold. FIM matrix element. The calculation formula is:

[0073] ;

[0074] In the formula: This represents the mathematical expectation of a probability distribution; Quantitatively characterizing macroscopic parameters Distribution of the system's microstate caused by minute changes The degree of difference.

[0075] 103. Calculate the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network;

[0076] Furthermore, after constructing the statistical manifold, the process proceeds to step 103: calculating the Ricci scalar curvature R. After obtaining the statistical manifold M with the metric g(θ), the system can calculate the geometric properties of the manifold. This application uses the Ricci scalar curvature R as a key indicator for measuring the stability of the distribution network system. The Ricci scalar curvature is a core physical quantity in differential geometry describing the intrinsic curvature of space. In the physical interpretation of this application, a significantly negative Ricci scalar curvature characterizes a sharp increase in the nonlinear coupling between state variables within the system, leading to a highly divergent trajectory of state points in the manifold space. This constitutes a profound intrinsic geometric precursor to an impending phase transition or dynamic instability (such as voltage collapse or frequency oscillation).

[0077] In one feasible implementation, the geometric metric of the statistical manifold is the Fisher information matrix. Then, calculating the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network includes: calculating the Krzy symbol based on the Fisher information matrix of the current state point and the derivative of the Fisher information matrix, and calculating the Ricci scalar curvature of the current state point based on the Krzy symbol.

[0078] In other words, the specific calculation process of Ricci scalar curvature is as follows: First, based on the Fisher information matrix g(θ) and its first-order partial derivative with respect to θ, the connection coefficient of the manifold, i.e., the Krzy symbol, is calculated. Then, using the Kirchhoff symbol and its derivative, the Riemann curvature tensor is calculated; by performing a contraction operation on the Riemann curvature tensor, the Ricci tensor is obtained, and then contracted again to finally obtain the Ricci scalar curvature R(θ) of the current state point θ(t). Although this calculation process is complex, it is entirely based on the established manifold geometry and has a clear mathematical path.

[0079] 104. Compare the Ritchie scalar curvature with a preset negative threshold to determine whether there are signs of instability in the power distribution network;

[0080] Once the Ricci scalar curvature at the current state point is obtained, the process can proceed to step 104. The system will then compare the real-time calculated Ricci scalar curvature R(θ) with a preset negative threshold R. th Compare the threshold R. th This can be obtained through offline simulation analysis or statistical analysis based on historical operating data, representing the lower limit of the acceptable stability margin of the system. If R(θ) is greater than or equal to R... th If the current operating status is within a safe range, the process will return to continue the next round of monitoring and calculation, thus forming a continuous scan of the "geometric health status" of the power grid.

[0081] Conversely, if we determine that R(θ) is less than R th The system then determines that the distribution network has shown significant signs of instability. At this point, even if traditional electrical quantities such as voltage and frequency have not yet exceeded alarm limits, this method can still issue an early warning based on dramatic changes in geometric properties. This is equivalent to the current state point θ(t) having entered or approached a negative curvature region R<0.

[0082] For example, steps 103 and 104 implement fault precursor detection based on manifold curvature, including two key steps as follows:

[0083] 1) Calculate the Ricci scalar curvature of the information manifold:

[0084] manifold The degree of local curvature is precisely characterized by its curvature. This invention introduces Ricci Scalar Curvature. As a core geometric indicator for measuring system stability. The calculation requires first solving for the Christoffel symbols derived from the metric tensor. :

[0085] ;

[0086] In the formula: It is the Fisher information matrix The inverse matrix elements, , , To measure tensor components. Furthermore, Ricci scalar curvature. It is obtained from the Riemann curvature tensor contraction:

[0087] ;

[0088] In information geometry, negative Ricci curvature is closely related to critical behaviors of a system, such as chaos, bifurcation, and phase transition.

[0089] 2) Establish precursor criteria:

[0090] When the system state Evolved to manifold When the value is significantly negative, it indicates that the interactions between the subsystems within the system become extremely unstable, reaching a critical point of instability. This corresponds to a critical state where the system's Jacobian matrix approaches a singularity. For example, when... A sustained decrease and exceeding the threshold may indicate a sensitivity between the voltage of certain nodes in the system and their reactive power injection (i.e., A sharp increase in voltage () is a typical precursor to voltage instability. Therefore, the precursor criterion is set as follows:

[0091]

[0092] In the formula: It is a negative threshold value determined through historical data analysis or extensive simulation.

[0093] 105. When it is determined that there are signs of instability, a geodesic line connecting the current state point to a preset target stable state point is determined on the statistical manifold as the optimal recovery path;

[0094] Furthermore, once an indication of impending instability is detected, the system immediately enters the self-healing control decision-making phase, namely step 105: determining the optimal recovery path (geodesic). The goal at this point is to bring the system back from its current unstable state. (That is, θ(t) = (), guide to a preset target stable state point. The target point represents a known safe and reliable operating state of the power grid, whose parameters can be pre-stored in the system. Within the geometric framework of the statistical manifold M, the optimal recovery problem is transformed into finding a connection between two points. and The problem of finding the shortest path between two points. The shortest path in a manifold space is called a geodesic (optimal reconstructed path). Therefore, the system determines the optimal recovery path by solving the geodesic equations. The geodesic equations are a system of second-order ordinary differential equations:

[0095] ;

[0096] Where s is the arc length parameter of the geodesic. Solving this system of equations yields a line from... and parameter evolution path This path represents the trajectory by which the system recovers to a steady state with minimal "internal resistance" and maximum efficiency, avoiding unnecessary detours and oscillations in the state space.

[0097] 106. Generate an optimal control law based on the optimal recovery path, map the optimal control law to a target control command for the distribution network, and execute the target control command; the optimal control law is used to drive the operating state of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network; the target control command is used to control one or more control resources of the distribution network.

[0098] For example, control resources include at least one of distributed power sources, energy storage systems, flexible loads, or electric vehicle charging station clusters.

[0099] In one feasible implementation, mapping the optimal control law to a target control command for the distribution network includes: obtaining the target control command using a pre-established sensitivity matrix and the optimal control law; the sensitivity matrix is ​​used to reflect the mapping relationship between the parameter changes of the probability distribution corresponding to the control law and the physical control quantity corresponding to the control command of the control resource.

[0100] For example, the optimal control law includes the following mathematical expression:

[0101] ;

[0102] In the formula: Let θ(t) be the optimal control law for the current state point θ(t), where the state point is represented by the parameters of the probability distribution. For control gain matrix; It measures the difference between the current state point θ(t) and the optimal recovery path θ. geo Lyapunov function of the deviation between; It represents the covariant derivative on the statistical manifold.

[0103] For example, the target control command includes the following mathematical expression:

[0104] ;

[0105] In the formula, It is a sensitivity matrix The false reversal; The target control command is a vector containing specific control quantities; Let θ(t) be the optimal control law for the current state point θ(t).

[0106] It should be noted that, in determining the optimal recovery path After (i.e., the target geodesic), step 106 is executed: generating and mapping control commands. This step aims to transform the abstract geometric path into concrete, executable actions for the physical devices. First, the system needs to generate a path that drives the actual operating state θ(t) of the distribution network along the optimal recovery path determined above. The control law of evolution. This can usually be achieved by designing a tracking controller whose output is to make θ(t) follow the curve. The required rate of change of the parameter, i.e. .

[0107] The next crucial step is mapping. This dθ / dt, located in the parameter space, needs to be converted into adjustment commands for the physical control quantities of one or more control resources 42 in the distribution network. These control resources 42 can be distributed photovoltaic inverters, energy storage systems, flexible loads, etc. To achieve this mapping, this embodiment utilizes a pre-established or online identified system sensitivity matrix J. This matrix describes the impact of small changes in the physical control quantity u of the control resource 42 (e.g., the reactive power output of the photovoltaic inverter, the charging and discharging power of the energy storage system, etc.) on the parameter vector θ (e.g., the mean node voltage, the phase angle variance, etc.), and the relationship can be linearized as: dθ / dt = J* du / dt. Accordingly, with dθ / dt and the sensitivity matrix J, the system can obtain the required rate of change of the physical control quantity du / dt by solving this linear equation system in reverse, and then generate specific control commands u (i.e., For example, the pseudoinverse J can be calculated. + To solve for: du / dt = J + * (dθ / dt).

[0108] Finally, the generated specific control commands are sent to the corresponding control resources 42 via the communication network. Control resources 42 execute these commands; for example, the photovoltaic inverter increases reactive power output, or the energy storage system switches from charging to discharging. These physical actions collectively generate a generalized "control force," precisely guiding the actual operating state point θ(t) of the entire distribution network along the planned optimal recovery path. Towards the target stable state point The process of moving the device ultimately enables rapid, stable, and efficient self-healing recovery.

[0109] For example, steps 105 and 106 achieve optimal self-healing control based on non-equilibrium thermodynamics, including three key processing steps, as follows:

[0110] 1) Define the system's entropy change and dissipation structure:

[0111] If we consider the power distribution network as an open thermodynamic system, according to non-equilibrium thermodynamics, its total entropy... The rate of change can be decomposed into:

[0112] ;

[0113] In the formula: The constant validity represents the entropy generated by irreversible processes within the system (such as Joule loss); This represents the entropy flow caused by the exchange of energy and matter between the system and its external environment. Distributed generation (DG) injects active power into the system, essentially providing a "negative entropy flow" (…). This is the fundamental driving force for the system to form a new ordered and stable structure (i.e., a dissipative structure) far from equilibrium.

[0114] 2) Determine the self-healing goal and optimal recovery path:

[0115] The physical goal of resilience and self-healing is to control movement. (Such as switching operations, DG power scheduling), maximizing the injection rate of negative entropy flow to offset the positive entropy generated by the fault as quickly as possible, mathematically expressed as:

[0116] ;

[0117] In the formula: Refers to the optimal control action; The optimization objective is the negative entropy flow injection rate, which is the control action. The function.

[0118] Combining information geometry, the optimal recovery path is defined as a statistical manifold. Connect to the current fault status point With the target steady state point Geodesic (Optimal Recovery Path) A geodesic is the shortest path on a manifold, representing the path with the least resistance and highest efficiency in state evolution. It is defined by the following differential equation:

[0119] ;

[0120] In the formula: It is the arc length parameter along the path.

[0121] 3) Generate the optimal control law and map it to electrical control commands:

[0122] The core task of optimal control is to generate a set of specific physical control actions. This generates a generalized force that drives the actual state of the system. Follow the planned geodesic path This involves two steps:

[0123] 3a) Generation of geometric space control laws:

[0124] First, a tracking control law is designed on the statistical manifold, and the formula for achieving this is calculated. follow The required "parameter space driving force". An implementable control law can be expressed as:

[0125] ;

[0126] In the formula: For the control gain matrix, It is a Lyapunov function that measures the deviation between the current state and the optimal recovery path. This represents the covariant derivative on the manifold.

[0127] 3b) Physical control command mapping:

[0128] Next, we need to transform this abstract parameter's rate of change. Transformed into specific control commands for physical devices This control method, which incorporates a metric tensor and a sensitivity matrix, can adaptively adjust the control intensity based on the system's operating state (the local geometry of the manifold), achieving precise and efficient regulation. This module utilizes the system's sensitivity matrix. To establish the relationship between physical control quantities and statistical parameters:

[0129] ;

[0130] Therefore, the final physical control command is:

[0131] ;

[0132] In the formula: It is a sensitivity matrix The false reversal; It is a vector that contains specific control quantities.

[0133] In summary, compared with existing technologies, the technical solution provided by this invention has the following significant advantages: a) Foresight in fault early warning: The criterion based on information curvature captures the precursors of instability from the abrupt change level of the inherent geometric properties of the system state space. Compared with any criterion based on electrical quantity thresholds (such as voltage and frequency exceeding limits), it can predict the critical state earlier and more fundamentally, realizing true "pre-event" preventive control and providing a new approach to solving "black swan" event early warning. b) Inherent robustness to uncertainty: The entire framework is built on the basis of probability distribution and statistical manifold. The uncertainty of the system (such as new energy fluctuations and load changes) is naturally embedded in the geometric structure of the manifold, making the control strategy inherently adaptable and robust to random disturbances.

[0134] It is understood that this application can implement a closed-loop control system. Taking the signaling interaction timing of the method in this embodiment during a self-healing event as an example, in the normal monitoring cycle, the sensor continuously sends measurement data to the control system. At time t1, the control system detects that the Ricci scalar curvature R(θ) falls below the threshold R. th This triggers an early warning. Within the time window t1 to t2 (t2 is greater than t1), the control system quickly executes steps 105 and 106, namely calculating the geodesic and generating control commands. At time t2, the specific control commands are sent to the control resources. By time t3 (t3 is greater than t2), the control resources begin to respond to the commands and adjust their output. This effect changes the power grid state, and this change is captured by sensors and fed back to the control system, thus forming a closed loop.

[0135] Through the above steps, the method provided in this application can not only predict risks in advance, but also provide a theoretically optimal recovery path and transform it into precise control of physical equipment, thereby greatly improving the resilience and self-healing capability of the power distribution network under uncertain disturbances.

[0136] In one feasible implementation, a specific and optimized variant of this application is introduced, the main difference of which is that a specific selection is made for the probability distribution model of the distribution network operation status, which aims to significantly improve computational efficiency and response speed in specific application scenarios.

[0137] In many practical power distribution network scenarios, especially when facing numerous subtle disturbances caused by renewable energy and random loads, the fluctuation characteristics of the system's key electrical quantities can be well approximated by a multivariate Gaussian distribution. Based on this observation, this embodiment explicitly assumes that the system's microstate vector x follows a multivariate Gaussian distribution when performing step 102 (constructing the statistical manifold).

[0138] Specifically, a k-dimensional multivariate Gaussian distribution is entirely determined by its mean vector μ (a k-dimensional vector) and covariance matrix Σ (a k x k symmetric positive definite matrix). Therefore, in this case, the parameter θ constituting the statistical manifold M is composed of independent elements of μ and Σ. For example, the parameter vector θ can be written as θ = (μ, vech(Σ)), where vech(Σ) represents straightening the lower triangular portion (including the diagonal) of the covariance matrix Σ into a vector.

[0139] Choosing the multivariate Gaussian distribution as the parameterized probability distribution model has a core advantage: the statistical manifold (i.e., the Gaussian manifold) formed by all multivariate Gaussian distributions possesses many excellent mathematical properties, and its geometric quantities typically have analytical expressions. A significant advantage is that its Fisher information matrix g(θ) no longer needs to be calculated through complex numerical integration of p(x; θ), but can be directly expressed in analytical form. For example, for the metric tensor of the mean parameter μ, it is equal to the inverse Σ of the covariance matrix. -1 For the metric tensor of the covariance parameter part, its expression is also the same as Σ. -1 Related.

[0140] This analytical property brings significant computational convenience. In the aforementioned embodiment, calculating the Fisher information matrix g(θ) is perhaps one of the most time-consuming steps in online computation. However, in this embodiment, since g(θ) has an analytical expression, the system only needs to estimate the mean vector μ and covariance matrix Σ based on the collected data to immediately obtain the Fisher information matrix at the current state point θ(t).

[0141] The advantages of this simplification will extend to subsequent computational steps. For example, calculating the Kirchhoff symbol and Ricci scalar curvature R(θ) in step 103, and solving the geodesic equation in step 105, become more direct and faster due to the analytical knowability of the metric tensor g(θ) and its derivative. This significantly reduces the online computational burden of the entire self-healing control method and shortens the response time.

[0142] For example, in a microgrid dominated by wind and solar power, the short-term fluctuations in power output closely resemble a Gaussian distribution. When a sudden strong wind causes a drastic change in wind turbine output, the elements of the system's covariance matrix Σ change abruptly. This is reflected in the geometry of the statistical manifold M, causing a rapid decrease in the Ricci scalar curvature R. Using the method in this embodiment, the control system can leverage the analytical properties of the Gaussian manifold to complete the entire process—from detecting curvature changes to planning geodesic recovery paths and generating control commands for the energy storage system or solar inverter—in milliseconds. Compared to general models that require numerical integration, the method in this embodiment demonstrates significant advantages in applications with extremely high real-time requirements.

[0143] Therefore, this embodiment demonstrates the flexibility of the technical solution of this application. By selecting a specific model (such as a multivariate Gaussian distribution) that conforms to the physical characteristics of a specific scenario for the "parameterized probability distribution", the practicality and engineering application value of the method can be effectively improved without sacrificing theoretical completeness.

[0144] In one feasible implementation, this embodiment aims to further illustrate the universality and collaborative optimization capability of the method of this application at the control execution level, specifically demonstrating how to map abstract geometric control laws onto diverse and heterogeneous control resources in the distribution network to achieve global collaborative self-healing.

[0145] In this embodiment, the preliminary steps, including steps 101 (acquiring real-time data), 102 (constructing a statistical manifold), 103 / 104 (detecting early signs of instability), and 105 (determining the optimal recovery path), are basically the same in principle and execution process as in the aforementioned embodiments. The features of this embodiment are mainly reflected in the implementation details of step 106 (generating and mapping control commands).

[0146] In modern and future active distribution networks, the types of resources available for regulation and control are becoming increasingly diverse. Besides traditional generator sets and reactive power compensation devices, a large number of power electronic devices and demand-side resources are emerging. In the distribution network 40 considered in this embodiment, the set u of control resources 42 is a multi-dimensional, heterogeneous vector. Specifically, it includes at least the following categories: 1. Distributed power sources: for example, inverters in photovoltaic power plants, which can regulate their output active and reactive power. 2. Energy storage systems: for example, battery energy storage systems deployed at critical locations in the grid, which can rapidly charge and discharge, achieving bidirectional power throughput. 3. Flexible loads: for example, through agreements with industrial users, some interruptible loads in their production processes can be regulated. 4. Electric vehicle charging pile clusters: managed uniformly by a load aggregator, their total charging power can be regulated as a whole.

[0147] When the control system calculates the optimal recovery path in step 105 After generating the target parameter change rate dθ / dt required for the driving state evolution in the initial stage of step 106, the core problem to be solved is how to reasonably allocate this "total task" to the aforementioned control resources 42 with different performances.

[0148] Therefore, the sensitivity matrix J used in this embodiment has a broader dimension and richer content. It is no longer the sensitivity of a single type of resource, but a generalized sensitivity matrix describing the impact of all available heterogeneous resources on statistical parameters. Each column of matrix J corresponds to a specific physical control quantity; for example, the first column might correspond to the reactive power regulation of a photovoltaic inverter, the second column to the charging and discharging power of an energy storage system, the third column to the total power reduction of an electric vehicle charging cluster, and so on. Each row of the matrix corresponds to a component of θ. The elements J in the matrix... ij This means that a unit change in the j-th physical control quantity can cause a change in the i-th statistical parameter.

[0149] When solving for the control command du / dt, the system needs to solve the equation dθ / dt = J * du / dt. Since the number of control resources is typically greater than the dimension of statistical parameters, this is an underdetermined system of equations, meaning there are infinitely many solutions. This provides space for collaborative optimization. The system can find an optimal solution du / dt based on one or more optimization objectives, while satisfying dθ / dt = J * du / dt. These optimization objectives can include: minimizing total control cost, minimizing energy storage battery losses, and minimizing the impact on the user side, etc.

[0150] The following will illustrate this with a specific working scenario. Suppose that the system detects an early sign of instability (manifested as the Ricci scalar curvature R(θ) falling below a threshold) due to excessive power flow on a certain transmission line. The control system calculates the optimal recovery path. The physical meaning of dθ / dt is "the power flowing through this line needs to be rapidly reduced while maintaining voltage stability at critical nodes." At this point, the instruction generation module (such as...) Figure 2 14) Solving using the extended sensitivity matrix J. It will be found that there are multiple resource combinations to achieve this goal. Through internal optimization algorithms, the system determines the most economical and fastest collaborative control strategy: Instruction 1: Issued to a 5 MW energy storage system located downstream of the line, instructing it to immediately switch from a light-load charging state to a full-power discharging state to supply power to the local load. Instruction 2: Issued to the electric vehicle charging load aggregator in the area, instructing it to temporarily reduce the total charging load by 30% within the next 5 minutes.

[0151] These two commands were issued simultaneously. The discharge of the energy storage system directly reduced the power drawn from the line, while the reduction in electric vehicle charging load further alleviated the line's burden. Together, they rapidly reduced the line power flow to a safe range, effectively "pushing" the entire power grid's operating state point θ(t) back to the planned value. It will eventually return to stability.

[0152] This embodiment fully demonstrates the good compatibility and scalability of the method framework of this application. It provides a unified, top-level control objective (evolving along geodesics) and can decompose and map this objective onto diverse execution units at the physical level to achieve multi-resource collaborative optimization. This is of vital significance for fully tapping the resilience potential of active distribution networks and addressing the complex challenges of future power grids.

[0153] The core technical solution of this application lies in establishing a novel distribution network resilience self-healing control framework that integrates information geometry and non-equilibrium physics. This framework elevates the analysis of the power grid from the traditional Euclidean space to a deeper geometric space that can endogenously describe the uncertainties of the system.

[0154] Specifically, this technical solution first treats the instantaneous operating state of the distribution network as no longer a deterministic vector of electrical quantities, but rather models it as a joint probability distribution of key electrical quantities using probabilistic methods. All these possible probability distributions together constitute a high-dimensional, inherently curved mathematical space, i.e., a statistical manifold. To describe the geometric properties of this space, this solution introduces the Fisher information matrix as its metric tensor, which precisely defines the "information distance" between different system states (i.e., different probability distributions).

[0155] Building upon this geometric framework, this scheme further introduces Ricci scalar curvature from differential geometry as a key indicator for diagnosing system stability. By calculating the curvature at the location of system state points in real time, this scheme can proactively identify negative curvature regions on the manifold caused by strong coupling between electrical parameters. For example, in the lead-up to voltage collapse, the sensitivity of node voltages to load disturbances increases dramatically; this strong nonlinear correlation manifests on the statistical manifold as a high degree of curvature in local space, i.e., negative Ricci curvature. These regions have been shown to be precursors to phase transitions or dynamic instability in the system, thus constructing an advanced early warning mechanism independent of specific fault modes.

[0156] Upon detecting early signs of instability, this scheme activates its self-healing control decision module. This module is theoretically based on dissipative structure theory, viewing the resilience recovery process as an active self-organizing process: by intelligently controlling resources such as distributed energy sources, it injects negative entropy into the system to counteract the internal disordering trend caused by the fault, ultimately forming a new, orderly dissipative structure. To achieve this goal, this scheme maps the optimal recovery strategy to finding the shortest path—a geodesic—on the statistical manifold connecting the current fault state to the target stable state. Finally, by designing a control law that guides the system state strictly along this geodesic evolution and translating this abstract control law into specific power regulation commands for distributed power sources, energy storage, flexible loads, and other devices, this scheme ensures the efficiency, stability, and optimality of the self-healing process.

[0157] This application offers the following advantages. First, by calculating the Ricci scalar curvature in the information geometric space to determine system stability, it can capture early signs of instability from the perspective of abrupt changes in the inherent geometric properties of the system's state space. Compared to traditional criteria relying on electrical quantity limits, it can predict critical states earlier and more fundamentally, achieving "preemptive" preventative control and providing a forward-looking early warning advantage. Second, this application transforms the optimal recovery problem into finding geodesics on a statistical manifold, providing a theoretically most efficient and stable recovery trajectory for self-healing control. This avoids the problems of traditional optimization algorithms potentially getting trapped in local optima or experiencing slow convergence, achieving optimal recovery. Third, the entire framework of this application is built upon probability distributions and statistical manifolds. Uncertainties in the power grid are naturally inherent in the geometric structure of the manifold. Therefore, this method has a natural adaptability and robustness to random disturbances, exhibiting intrinsic robustness. Finally, based on clear geometric principles, each step of this application has explicit mathematical and physical meaning, overcoming the "black box" decision-making defects of traditional machine learning methods, improving the safety and reliability of control decisions, and possessing theoretical interpretability.

[0158] Furthermore, this embodiment describes the physical system architecture for implementing the above-described information geometry-based distribution network self-healing control method. This system provides a complete hardware and software platform for the practical engineering deployment of the method.

[0159] Figure 2 This is a schematic diagram of a self-healing control system provided in an embodiment of this application. The system mainly consists of a central processing server 10, a data acquisition unit 20, and a communication unit 30, and interacts with the external power distribution network 40 environment.

[0160] The data acquisition unit 20 is the system's data input interface. It is connected to multiple sensors 41 deployed in the distribution network 40 via a high-speed, reliable communication link (e.g., fiber optic or wireless private network). As described in the previous embodiments, these sensors 41 are preferably phasor measurement units capable of providing high-precision synchronous vector measurements, but may also include other types of smart meters, fault recorders, etc. The main function of the data acquisition unit 20 is to collect massive amounts of measurement data streams from each sensor 41 in real time and continuously, perform preliminary data cleaning, time synchronization, and formatting, and then securely transmit the processed real-time data stream to the core of the system—the central processing server 10.

[0161] The central processing server 10 is the computational core that enables the entire self-healing control method. It typically consists of a high-performance computer, including a powerful central processing unit, a graphics processing unit (for parallel acceleration of matrix operations), and large-capacity memory and hard disk storage. The memory stores the computer program instructions that implement the method of this application. When the processor executes these programs, the following functional modules are logically implemented:

[0162] State Modeling Module 11: This module executes step 102. It receives real-time data from the data acquisition unit 20 and calculates the statistical moments of key electrical quantities within a specified time window based on a preset probabilistic model (such as the aforementioned multivariate Gaussian model, or a more general nonparametric model), thereby determining the parameter vector θ(t) at the current moment. This process is equivalent to mapping the physical state of the power grid to a point on the statistical manifold M in real time. This module is also responsible for maintaining the information required for the manifold geometry, such as the method for calculating the Fisher information matrix.

[0163] Stability Assessment Module 12: This module corresponds to steps 103 and 104. It is tightly coupled with State Modeling Module 11, and the stability assessment module 12 immediately initiates calculations once State Modeling Module 11 updates the current state point θ(t). It uses the manifold geometry information provided by State Modeling Module 11 to calculate the Ricci scalar curvature R(θ) at the current point θ(t). Subsequently, it compares the calculated R(θ) with the negative threshold R stored in the configuration library. th Compare them. If R(θ) is detected to be less than R... th This module will immediately generate an internal warning signal and trigger the control decision module 13.

[0164] Control Decision Module 13: This module executes step 105 and is activated upon receiving an early warning signal from the stability assessment module 12. This module first reads the preset target stable state point from the configuration library. Then, using the current state as the point (The state point at which the warning is triggered) is taken as the starting point. With the endpoint as the starting point, solve the geodesic equation. The result of the solution is the optimal recovery path. That is, an ideal trajectory in which parameters evolve over time. .

[0165] Instruction generation module 14: This module is used to execute step 106. It receives the optimal recovery path output from control decision module 13. This module internally contains a tracking controller and a sensitivity matrix J. It first... The control law (i.e., the target parameter change rate dθ / dt) required to drive the actual state to follow the path is generated. Then, using the sensitivity matrix J (as described in the previous embodiment), the specific physical control commands u for each control resource 42 are solved in reverse. These commands can be precise power setpoints, switching actions, etc.

[0166] The communication unit 30 is the system's control output interface. It receives specific control commands generated by the command generation module 14 in the central processing server 10. This unit is responsible for encoding these commands into messages that conform to specific communication protocols (such as IEC61850, DNP3, etc.) and accurately sending them to the local controllers (such as the controllers of photovoltaic inverters, the energy management systems of energy storage systems, the control platforms of load aggregators, etc.) of one or more corresponding control resources 42 in the distribution network 40 through a secure, low-latency communication network.

[0167] The entire system's workflow forms a complete closed loop of "perception-analysis-decision-execution". Data flows from the distribution network 40 through sensors 41 and data acquisition units 20 into the central processing server 10. Within the server 10, processing and decision-making occur in a pipeline manner among four modules. The resulting control commands are then returned to the control resources 42 in the distribution network 40 via the communication unit 30. By changing their operating states, these resources influence the power grid, thereby achieving closed-loop self-healing control of the power grid's state. The system architecture provided in this embodiment is clear and complete, providing a solid foundation for the engineering and productization of the advanced control method proposed in this application.

[0168] Please see Figure 3 , Figure 3 This is a structural block diagram of a power distribution network self-healing control device based on information geometry, as described in an embodiment of this application. Figure 3 The apparatus shown includes:

[0169] Data acquisition module 301: used to acquire real-time operating status data of the power distribution network;

[0170] Manifold determination module 302: used to model the operating state of the distribution network into a parameterized probability distribution based on the real-time operating state data, and to construct a statistical manifold based on the parameters of the probability distribution;

[0171] Curvature determination module 303: used to calculate the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network;

[0172] Instability judgment module 304: used to compare the Ricci scalar curvature with a preset negative threshold to determine whether there are signs of instability in the power distribution network;

[0173] Path determination module 305: When it is determined that there is a precursor to instability, it determines a geodesic line on the statistical manifold that connects the current state point to a preset target stable state point as the optimal recovery path;

[0174] Self-healing control module 306: used to generate an optimal control law based on the optimal recovery path, map the optimal control law to a target control command of the distribution network, and execute the target control command; the optimal control law is used to drive the operating state of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network; the target control command is used to control one or more control resources of the distribution network.

[0175] It should be noted that, Figure 3 The functions of each module in the device shown are as follows: Figure 1 The steps in the method shown are similar, and will not be repeated here to avoid repetition. For details, please refer to [reference needed]. Figure 1 The content of each step in the method shown.

[0176] This application offers the following advantages. First, by calculating the Ricci scalar curvature in the information geometric space to determine system stability, it can capture early signs of instability from the perspective of abrupt changes in the inherent geometric properties of the system's state space. Compared to traditional criteria relying on electrical quantity limits, it can predict critical states earlier and more fundamentally, achieving "preemptive" preventative control and providing a forward-looking early warning advantage. Second, this application transforms the optimal recovery problem into finding geodesics on a statistical manifold, providing a theoretically most efficient and stable recovery trajectory for self-healing control. This avoids the problems of traditional optimization algorithms potentially getting trapped in local optima or experiencing slow convergence, achieving optimal recovery. Third, the entire framework of this application is built upon probability distributions and statistical manifolds. Uncertainties in the power grid are naturally inherent in the geometric structure of the manifold. Therefore, this method has a natural adaptability and robustness to random disturbances, exhibiting intrinsic robustness. Finally, based on clear geometric principles, each step of this application has explicit mathematical and physical meaning, overcoming the "black box" decision-making defects of traditional machine learning methods, improving the safety and reliability of control decisions, and possessing theoretical interpretability.

[0177] Figure 4 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 4As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. The internal memory may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. Those skilled in the art will understand that… Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0178] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform actions such as... Figure 1 The steps of the method shown.

[0179] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following actions: Figure 1 The steps of the method shown.

[0180] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0181] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0182] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A self-healing control method for distribution networks based on information geometry, characterized in that, The method includes: Obtain real-time operating status data of the power distribution network; Based on the real-time operating status data, the operating status model of the distribution network is transformed into a parameterized probability distribution, and a statistical manifold is constructed based on the parameters of the probability distribution. Calculate the Ricci scalar curvature at the current state point on the statistical manifold that corresponds to the current operating state of the distribution network; The Ricci scalar curvature is compared with a preset negative threshold to determine whether there are signs of instability in the power distribution network. When it is determined that there are signs of instability, a geodesic line connecting the current state point to a preset target stable state point is determined on the statistical manifold as the optimal recovery path. An optimal control law is generated based on the optimal recovery path, and the optimal control law is mapped to a target control command for the distribution network. The target control command is then executed. The optimal control law is used to drive the operating state of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network. The target control command is used to control one or more control resources of the distribution network. Wherein, the geometric metric of the statistical manifold is the Fisher information matrix, then calculating the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network includes: Based on the Fisher information matrix of the current state point and its derivative, calculate the Krzy symbol, and calculate the Ricci scalar curvature of the current state point based on the Krzy symbol. The optimal control law includes the following mathematical expression: ； In the formula: Let θ(t) be the optimal control law for the current state point θ(t), where the state point is represented by the parameters of the probability distribution. For control gain matrix; It measures the difference between the current state point θ(t) and the optimal recovery path θ. geo Lyapunov function of the deviation between; Represents the covariant derivative on the statistical manifold; The target control command includes the following mathematical expression: ； In the formula, It is a sensitivity matrix The false reversal; The target control command is a vector containing specific control quantities; Let θ(t) be the optimal control law for the current state point θ(t).

2. The method according to claim 1, characterized in that, The parameterized probability distribution is the joint probability distribution of key electrical quantities in the distribution network; The key electrical quantities include at least one of the following: node voltage amplitude, phase angle, active power of distributed generation, reactive power of distributed generation, and branch current. The parameters of the probability distribution are composed of the statistical moments of the key electrical quantities.

3. The method according to claim 1, characterized in that, The process of mapping the optimal control law to the target control command of the distribution network includes: The target control command is obtained using a pre-established sensitivity matrix and the optimal control law; the sensitivity matrix is ​​used to reflect the mapping relationship between the parameter changes of the probability distribution corresponding to the control law and the physical control quantity corresponding to the control command of the control resource.

4. The method according to claim 1, characterized in that, The control resources include at least one of distributed power sources, energy storage systems, flexible loads, or electric vehicle charging pile clusters.

5. A power distribution network self-healing control device based on information geometry, characterized in that, The apparatus is used to implement the method as described in any one of claims 1-4, the apparatus comprising: Data acquisition module: used to acquire real-time operating status data of the power distribution network; Manifold determination module: used to model the operating state of the distribution network into a parameterized probability distribution based on the real-time operating state data, and to construct a statistical manifold based on the parameters of the probability distribution; Curvature determination module: used to calculate the Ricci scalar curvature at the current state point on the statistical manifold corresponding to the current operating state of the distribution network; Instability detection module: used to compare the Ricci scalar curvature with a preset negative threshold to determine whether there are signs of instability in the distribution network; Path determination module: When it is determined that there are signs of instability, a geodesic line connecting the current state point to a preset target stable state point is determined on the statistical manifold as the optimal recovery path; The self-healing control module is used to generate an optimal control law based on the optimal recovery path, map the optimal control law to a target control command for the distribution network, and execute the target control command. The optimal control law is used to drive the operating state of the distribution network to evolve along the optimal recovery path to achieve self-healing control of the distribution network. The target control command is used to control one or more control resources of the distribution network.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 4.

7. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 4.

Images