Power system power flow optimization method and device based on embedded constraint neural network

Abstract

The application discloses a power system power flow optimization method and device based on an embedded constraint neural network, and belongs to the field of power system automation. The method comprises the following steps: constructing a power system digital twin; performing scene matching based on current operation state data to obtain a current operation scene, and performing scene deduction based on the current operation scene to obtain a plurality of potential sudden scene; correcting a power flow constraint boundary based on a historical fault database to obtain corrected constraint boundary parameters; inputting the current operation state data, the current operation scene, the potential sudden scene and the constraint boundary parameters into a trained embedded constraint neural network to perform power flow optimization, and obtaining a power flow optimization decision. The application can significantly reduce the constraint breakthrough probability, improve the power supply reliability and reduce the network loss.

Description

Technical Field

[0001] This application relates to the field of power system automation, and in particular to a power system power flow optimization method and apparatus based on an embedded constrained neural network. Background Technology

[0002] In the operation of power systems, power flow optimization is a core component to ensure the safe, economical, and stable operation of the system. Its main objective is to achieve the optimal allocation of power resources, reduce network losses, and improve power supply reliability while satisfying various physical constraints of the system (such as node voltage constraints, branch power constraints, generator output constraints, etc.).

[0003] Existing power flow optimization methods for power systems suffer from the following systemic defects: Numerical iterative methods based on physical models rely solely on static data at a single moment to determine the operating state, failing to capture dynamic evolution processes such as load fluctuations and changes in renewable energy output, and unable to predict potential emergencies like equipment overload and branch faults, resulting in optimization decisions lagging behind actual system changes. Furthermore, their computational complexity increases exponentially when facing large-scale complex constraints, making it difficult to meet real-time scheduling requirements, and model simplification can easily lead to optimization results deviating from physical laws. Similarly, conventional data-driven machine learning methods, as "black box" models, lack the ability to predict potential emergencies, and their outputs often fail to guarantee safety margins under sudden disturbances. In summary, existing technologies generally suffer from the systemic problems of "inability to predict potential emergencies and lagging optimization decisions," making it difficult to achieve efficient and reliable power flow optimization in dynamic scenarios. Summary of the Invention

[0004] The purpose of this application is to provide a power system power flow optimization method and apparatus based on an embedded constrained neural network, which solves the problems of existing power system power flow optimization methods that cannot predict potential sudden scenarios and whose optimization decisions lag behind actual system changes, thereby achieving efficient and reliable power flow optimization with forward-looking prediction capabilities in dynamic scenarios.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a power system power flow optimization method based on an embedded constrained neural network, including: A digital twin of the power system is constructed based on real-time power system operation status data and the physical topology of the power system within the target area. Based on the current operating status data collected at the current moment and combined with the power system digital twin, scenario matching is performed to obtain the current operating scenario. Based on the current operating scenario and combined with the power system digital twin, scenario simulation is performed to obtain multiple potential emergency scenarios. Based on the current operating status data and the historical fault database of the power system digital twin, the power flow constraint boundary is corrected to obtain the corrected constraint boundary parameters. The current operating status data, the current operating scenario, the potential sudden scenario, and the constraint boundary parameters are input into the trained embedded constraint neural network for power flow optimization to obtain power flow optimization decisions.

[0006] Secondly, this application provides a power system power flow optimization device based on an embedded constrained neural network, applied to the aforementioned power system power flow optimization method based on an embedded constrained neural network; the power system power flow optimization device based on an embedded constrained neural network includes: The digital twin construction module is used to construct a digital twin of the power system based on the real-time collected power system operation status data and the physical topology of the power system within the target area. The scenario matching and simulation module is used to perform scenario matching based on the current operating status data collected at the current moment and the power system digital twin to obtain the current operating scenario, and to perform scenario simulation based on the current operating scenario and the power system digital twin to obtain multiple potential emergency scenarios. The constraint boundary correction module is used to correct the power flow constraint boundary based on the current operating status data and the historical fault database of the power system digital twin, so as to obtain the corrected constraint boundary parameters. The power flow optimization module is used to input the current operating status data, the current operating scenario, the potential sudden scenario, and the constraint boundary parameters into a trained embedded constraint neural network to perform power flow optimization and obtain power flow optimization decisions.

[0007] Thirdly, this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the power system power flow optimization method based on the embedded constrained neural network described in any one of the above.

[0008] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the power system power flow optimization method based on an embedded constrained neural network as described above.

[0009] Compared with the prior art, this application has the following technical effects: This application provides a power system power flow optimization method and device based on an embedded constrained neural network. By constructing a digital twin of the power system and performing scenario matching based on current operating status data, it can accurately identify the current operating scenario. More importantly, based on the current operating scenario and the digital twin, it can perform scenario extrapolation and proactively generate multiple potential emergency scenarios. This makes the optimization method no longer limited to a passive response to current static data, but has the ability to predict future disturbance risks. The optimization decision can consider possible sudden situations such as equipment overload and voltage anomalies in advance, fundamentally solving the problem of "static perception and delayed decision-making" in existing technologies, and realizing a paradigm shift from "passive response" to "proactive adaptation".

[0010] This application abandons the traditional approach of pre-setting fixed constraint boundaries based on equipment nameplate parameters or standard specifications. Instead, it uses current operating status data and historical fault databases in digital twins to make targeted corrections to power flow constraint boundaries. This makes the corrected constraint boundary parameters more closely match the actual operating capabilities of the physical entity, effectively avoiding the waste of regulation capacity caused by "overprotection" and the safety hazards caused by "underprotection". Under the premise of ensuring system safety, it maximizes the regulation potential of the equipment and improves the economy of the power system.

[0011] By inputting the current operating status data, current operating scenario, potential emergency scenarios, and corrected constraint boundary parameters into a trained embedded constraint neural network for power flow optimization, this neural network can efficiently handle complex constraints, avoid the problem of exponential increase in computational complexity of traditional numerical iterative methods, significantly improve the real-time performance of optimization decisions, and meet the needs of online power system dispatch.

[0012] Furthermore, this application organically links three core steps—scenario matching and deduction, constraint boundary correction, and embedded neural network optimization—to form a complete closed-loop architecture. This allows the final power flow optimization decision to simultaneously consider current operational constraints, the impact of historical faults, and future unforeseen scenarios, achieving a better balance between security, economy, and real-time performance. Compared with existing technologies, this application can significantly reduce the probability of constraint breaches, improve power supply reliability, and reduce network losses in complex dynamic environments, demonstrating significant engineering application value. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments 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.

[0014] Figure 1 This is a flowchart illustrating a power system power flow optimization method based on an embedded constrained neural network in one embodiment of this application. Figure 2 A detailed flowchart illustrating the steps for determining the current running scenario to which the current running status data belongs, provided in another embodiment of this application; Figure 3 A detailed flowchart illustrating the steps for obtaining multiple potential emergency scenarios, provided for another embodiment of this application; Figure 4 A flowchart illustrating the steps for obtaining modified constraint boundary parameters according to another embodiment of this application; Figure 5 A flowchart illustrating the steps for obtaining the final constraint boundary parameters according to an embodiment of this application; Figure 6 A detailed flowchart illustrating the power flow optimization process of an embedded constrained neural network provided in another embodiment of this application; Figure 7 A flowchart illustrating the steps for obtaining power flow optimization decisions, provided for another embodiment of this application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0015] 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 skilled in the art without creative effort are within the scope of protection of this application.

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] This application addresses the systemic problems of existing power system power flow optimization methods, such as "static perception, rigid constraints, and insufficient embedding of physical rules," and proposes a power system power flow optimization method and device based on an embedded constraint neural network. Its core technical concept lies in constructing a closed-loop optimization architecture of "scenario prediction, constraint correction, and embedded neural network decision-making."

[0018] Specifically, the architecture comprises three core components: The first stage: Dynamic scenario perception and forward-looking prediction. Based on real-time collected power system operating status data and physical topology, a digital twin of the power system is constructed that is highly consistent with the real system. On this basis, the current operating scenario is accurately identified through scenario matching, and multiple potential emergency scenarios (such as equipment overload, voltage anomaly, branch fault, etc.) are proactively generated through scenario simulation, enabling optimization decisions to predict future disturbance risks.

[0019] The second stage involves dynamic correction of power flow constraint boundaries. This stage abandons the traditional approach of pre-setting fixed constraint boundaries based on equipment nameplate parameters. Instead, it utilizes a historical fault database within a digital twin to filter historical fault records of the same type as the current operating state, analyzes their correlation with current power flow influencing factors, and quantifies the impact of historical faults on the constraint boundaries. Simultaneously, by incorporating characteristic information from potential emergency scenarios, a layered and progressive approach is adopted: "theoretical standard range, initial deviation range, historical fault correction, scenario supplementary correction, and safety verification." This results in corrected constraint boundary parameters that align with the actual load-bearing capacity of the equipment and future risks.

[0020] The third stage: Neural network optimization decision-making with embedded physical rules. Current operating status data, current operating scenario, potential emergency scenarios, and corrected constraint boundary parameters are input into a specially designed embedded constraint neural network (hereinafter referred to as the neural network). This neural network is not a traditional "black box" model, but rather embeds the physical laws of the power system (such as power balance, voltage constraints, branch capacity limitations, etc.) into each layer of the network in a structured manner. The neural network includes a scenario mapping module, a risk level quantification module, a constraint dynamic adaptation module, a conflict location module, an optimization direction planning module, and a scheme decision output module. It can efficiently handle complex constraints and directly output the power flow optimization decision with the highest safety margin, satisfying all constraints and adapting to current and potential emergency scenarios.

[0021] Furthermore, by designing the neural network into a structured hierarchy that includes a scene mapping module, a risk level quantification module, and a constraint dynamic adaptation module, physical laws and operational constraints are no longer external penalty terms or post-processing steps, but are embedded within the forward propagation process of the neural network. Compared to traditional "black box" models or "posterior correction" methods, this "embedded" design significantly improves the necessary (rather than probabilistic) satisfaction of physical constraints by the optimization results, enhances the interpretability of decisions, and avoids combinatorial explosion problems under complex constraints due to collaborative computation between layers, further improving the real-time performance of online decision-making.

[0022] Through the organic synergy of the above three stages, this application has achieved a paradigm shift from "passive response" to "proactive adaptation," significantly improving the real-time performance, reliability, and economy of optimization decision-making while ensuring system security.

[0023] See Figure 1 , Figure 1 This is a flowchart illustrating the power system power flow optimization method based on embedded constrained neural networks provided in this application. In this embodiment, the execution entity of the power system power flow optimization method based on embedded constrained neural networks is a power flow optimization device. Therefore, the power system power flow optimization method based on embedded constrained neural networks includes: Step 10: Based on the real-time collected power system operation status data and the physical topology of the power system within the target area, construct a digital twin of the power system.

[0024] Optionally, the power flow optimization device first acquires real-time operational status data and physical topology data of the power system within the target area. The operational status data includes real-time collected data such as generator output power, load power consumption, bus voltage amplitude, branch transmission power, branch power flow direction, bus voltage deviation, and generator output margin. The physical topology data includes the location distribution, connection relationships, and equipment model parameters (such as generator rated power, branch resistance, and reactance) of generators, buses, and branches (e.g., transmission lines, transformers) within the target area. Then, 3D modeling, data mapping, and simulation modeling techniques are used to transform the acquired physical topology data into a 3D visualization model. Based on the operational status data, a mapping relationship is established between the operational status data and each device in the 3D visualization model, enabling the operational status data to be reflected in real-time on the corresponding devices in the 3D visualization model. This constructs a digital twin of the power system that is highly consistent with the physical form and operational status of the power system in the target area.

[0025] In one embodiment, taking the power system in the western part of a city as the target area, the power flow optimization device first collects data in real time based on sensors deployed on each generator, bus, and branch in the target area. The data includes: the output power of generator G1 is 80MW, the output power of generator G2 is 60MW; the power consumption of residential load L1 is 50MW, the power consumption of industrial load L2 is 70MW; the voltage amplitude of bus B1 is 10.5kV, the voltage amplitude of bus B2 is 10.3kV; the transmission power of branch Z1 (connecting bus B1 and bus B2) is 40MW, and the power flow direction is from bus B1 to bus B2; the voltage deviation of bus B1 is 0.2kV, the voltage deviation of bus B2 is 0.1kV; the output margin of generator G1 is 20MW, and the output margin of generator G2 is 15MW.

[0026] Simultaneously, physical topology data within the target area was acquired, revealing that generators G1 and G2 are located in Industrial Park A and Power Plant B in the western part of the city, respectively; busbars B1 and B2 are located in Area C of the city center and Area D in the west, respectively; branch Z1 is a 220kV transmission line with a resistance of 0.05Ω and a reactance of 0.2Ω; generator G1 is connected to busbar B1 via branch Z3, and generator G2 is connected to busbar B2 via branch Z4; loads L1 and L2 are connected to busbars B1 and B2 via branches Z5 and Z6, respectively. Subsequently, a 3D visualization model of the power system in this area was constructed using Unity3D 3D modeling software. The equipment locations, connections, and equipment model parameters from the physical topology data were integrated into the model. A mapping was then established between sensor-collected operational status data and the corresponding equipment in the 3D model through a data interface. For example, the 80MW output power data of generator G1 was mapped onto the generator G1 model in the 3D model, enabling real-time display of its output power. Ultimately, a digital twin of the power system in the western part of the city was constructed.

[0027] Step 20: Based on the current operating status data collected at the current moment, scenario matching is performed using the power system digital twin to obtain the current operating scenario. Then, scenario simulation is performed based on the current operating scenario and the power system digital twin to obtain multiple potential emergency scenarios.

[0028] Optionally, the power flow optimization device inputs the current operating status data (corresponding to the operating status data collected in real time in step 10) into the constructed power system digital twin, and matches it with the power system operating scenario feature library containing multiple scenarios that is pre-built and stored in the power system digital twin, thereby determining the current operating scenario to which the current operating status data belongs, as in steps 2011-2015.

[0029] Furthermore, after determining the current operating scenario, the power flow optimization device uses the current operating scenario as a basis, combined with the operating characteristics of the equipment in the power system digital twin and the stored physical relationships of the components, to simulate and extrapolate the possible operating states of the power system based on the current operating scenario in the future, and obtains multiple potential emergency scenarios, as detailed in steps 2021-2024.

[0030] Step 30: Based on the current operating status data and the historical fault database of the power system digital twin, the power flow constraint boundary is corrected to obtain the corrected constraint boundary parameters.

[0031] Optionally, the power flow optimization device uses the acquired current operating status data and the historical fault database in the power system digital twin. This historical fault database stores historical fault records (including fault occurrence time, fault type, fault cause, operating status data at the time of the fault, equipment aging assessment data, etc.) of various equipment (generators, buses, branches, etc.) in the target area power system. Therefore, based on the current operating status data and the data in the historical fault database, the parameters in the power flow constraint boundary (such as generator maximum output power constraint, bus voltage amplitude constraint, branch maximum transmission power constraint, etc.) are corrected to obtain the corrected constraint boundary parameters, as described in steps 301-305.

[0032] Step 40: Input the current running status data, current running scenario, potential sudden scenarios and constraint boundary parameters into the trained embedded constraint neural network for power flow optimization to obtain power flow optimization decisions.

[0033] Optionally, the power flow optimization device inputs the acquired current operating status data, as well as the determined current operating scenario, potential emergency scenario, and constraint boundary parameters, into a trained embedded constraint neural network for power flow optimization, and finally obtains a power flow optimization decision. This decision includes the adjusted generator output power, branch transmission power allocation scheme, etc., in order to achieve the goals of safe operation and economic dispatch of the power system.

[0034] The embedded constraint neural network is a neural network trained based on the fusion of physical laws of the power system and multi-source operational data. This neural network includes a scenario mapping module, a risk level quantification module, a constraint dynamic adaptation module, a conflict localization module, an optimization direction planning module, and a scheme decision output module.

[0035] The scene mapping module consists of feature extraction nodes, which are used to associate and map the corresponding features of the current running status data with the corresponding features of the current running scene.

[0036] The risk level quantification module consists of risk assessment results. It is used to assess the risk level of each potential emergency scenario based on the corresponding fault data, including equipment fault type, fault impact range, and load loss ratio.

[0037] The constraint dynamic adaptation module consists of constraint adjustment nodes, which are used to adjust each constraint parameter according to the risk level of potential emergencies determined by the risk level quantification module and the constraint boundary parameters.

[0038] The conflict localization module consists of conflict diagnosis nodes, which are used to receive the constraint adjustment results transmitted by the constraint dynamic adaptation module and locate the degree of conflict in combination with the current running status data.

[0039] The optimization direction planning module consists of direction generation nodes, which are used to determine the optimization direction based on the positioning results of the conflict location module and the impact of corresponding potential sudden scenarios.

[0040] The scheme decision output module consists of a strategy generation node, which is used to determine the optimal operating state parameter adjustment scheme based on the optimization direction output by the optimization direction planning module, i.e., power flow optimization decision.

[0041] To ensure that the embedded constraint neural network described in this application can be clearly and completely implemented by those skilled in the art, its network architecture, the embedding mechanism of physical constraints, the training data generation method, and the training process are described in detail below.

[0042] A) Overall Network Architecture This network employs a hybrid structure of graph neural networks and fully connected networks, with a constraint projection layer embedded at the end. Inputs include: current operating state data (vectorized measurements), current operating scenario encoding (one-hot vector), potential sudden scenario features (risk level, fault type encoding), and corrected constraint boundary parameters. Outputs are power flow optimization decisions (generator output adjustment, branch power flow targets, reactive power compensation, load transfer, etc.).

[0043] The network comprises eight modules arranged in the order of data processing: a graph feature embedding layer (graph convolutional network), a scene mapping module (multi-head attention mechanism), a risk level quantification module (Softmax classifier), a constraint dynamic adaptation module (parameterized bias network), a conflict localization module (graph attention network), an optimization direction planning module (gated recurrent unit), a constraint projection layer (non-learning layer, using ADMM solver), and a scheme decision output module (hybrid density network). The graph feature embedding layer and the conflict localization module use graph neural networks to process topological information; the remaining learnable modules use fully connected networks; and the constraint projection layer does not participate in parameter learning.

[0044] B) Embedded mechanisms of physical constraints To overcome the problem that conventional neural network outputs do not meet the physical constraints of power systems, this application employs two complementary methods to embed physical laws into the neural network: 1. Embedded structure: Constrained projection layer Add a non-learned projection layer before the decision output. Let the candidate decision variables obtained from the network's forward propagation be... ( (representing the real number field), the projection layer solves the convex quadratic programming problem and projects it onto the feasible region defined by the power balance equation and Kirchhoff's laws: in, This represents the optimal solution. Denotes the Euclidean norm. Indicates that it is subject to, Represents the linearized power balance constraint (using the DC power flow approximation). This represents other linear inequality constraints (such as generator ramp rate limits). The problem is solved quickly using the alternating direction multiplier method, with a typical solution time of less than 10 milliseconds. The projected output strictly satisfies all linear physical constraints.

[0045] 2. Embedded loss function: Constraint violation penalty term During the training phase, the total loss function Loss due to mission and the consequences of violating the constraints Weighted composition: in, This refers to task losses (such as the mean square error of the goal of minimizing network loss). To constrain the losses from violations, the violations such as voltage exceeding limits and power exceeding limits are summed. To balance the coefficients, an adaptive adjustment strategy is adopted (initially 0.1, gradually increasing if the constraint loss does not decrease). Through these two methods, the network output strictly satisfies the physical constraints for over 99% of the samples on the test set, and the remaining minor violations can be completely eliminated through the projection layer.

[0046] C) Generation and preprocessing of training data A large number of training samples are generated offline using a digital twin of the power system: First, various "current operating states" are randomly sampled based on historical operating data, covering different seasons, load levels, and renewable energy penetration rates. Second, multiple potential emergency scenarios (such as N-1 faults, sudden load changes, and renewable energy fluctuations) are extrapolated for each current state. Then, the constraint boundaries are corrected according to the methods described in steps 301-305 of this application. Finally, using the corrected constraints as boundaries, a commercial optimization solver (such as Gurobi) is used offline to solve for the exact optimal power flow solution as training labels. Appropriate Gaussian noise is added to the generated samples to enhance robustness. Finally, the samples are divided into a training set and a validation set (typically in a 9:1 ratio).

[0047] D) Detailed Network Structure and Hyperparameters The graph feature embedding layer uses a 3-layer graph convolutional network with ReLU activation and Dropout (typically 0.2) to prevent overfitting. The scene mapping module uses an 8-head attention mechanism combined with a fully connected network, with LeakyReLU activation. The risk level quantization module uses a 2-layer fully connected network plus Softmax. The constraint dynamic adaptation module uses a parameterized bias network with linear activation. The conflict localization module uses a 4-head graph attention network with ELU activation. The optimization direction planning module uses a 2-layer bidirectional GRU with tanh activation. The constraint projection layer uses an ADMM solver (maximum 30 iterations, tolerance 1e-4). The solution decision output module uses a hybrid density network (Gaussian mixture component K=3). Batch normalization is used between fully connected layers, and cosine annealing is used for learning rate scheduling.

[0048] E) Detailed steps of the training process Implemented using a deep learning framework (such as PyTorch), the key training settings are as follows: Optimizer: AdamW, initial learning rate approximately 10. -4 Weight decay 5×10 -4 Batch size: 128 Training period: 100-200 epochs, with early stopping (stop if the loss does not decrease for 15 consecutive epochs). Gradient clipping: Global norm clipped to 1.0 After training, the model's constraint violation rate on the validation set is less than 0.5% (before projection), and drops to 0 after the constraint projection layer. The typical forward inference time (including the projection layer) is less than 10 milliseconds, meeting the real-time scheduling requirements of power systems.

[0049] Specifically, the power flow optimization process of the embedded constraint neural network is as follows: the corresponding features of the current operating state data are associated and mapped with the corresponding features of the current operating scenario to find the corresponding associated features; the risk of each potential sudden scenario is quantitatively assessed to determine the corresponding risk level; the quantified risk level, the features of the current operating scenario, and the actual working conditions under the current operating scenario are used to adjust each constraint parameter to find the corresponding target constraint boundary, realize the activation of the embedded constraint conditions, and finally use the power flow optimization decision as the output result, as shown in steps 401-405.

[0050] This application embodiment constructs a digital twin of the power system, obtaining a digital carrier that can map the real-time operating status and physical topology of the power system in the target area. Based on this digital twin and combined with current operating status data, scenario matching and inference are performed to determine the current operating scenario and obtain multiple potential emergency scenarios. This accurately solves the problem of relying solely on static data to determine scenarios and failing to capture dynamic changes and predict emergency scenarios. Furthermore, based on the digital twin and its historical fault database, power flow constraint boundary correction is performed to obtain corrected constraint boundary parameters that closely match the actual equipment status and historical fault conditions. This addresses the issues of relying on preset fixed constraint boundaries and neglecting equipment aging and historical factors. To address the issue of boundary deviations caused by faults, the current operating scenario, potential sudden scenarios, constraint boundary parameters, and current operating status data are all input into a trained embedded constraint neural network for power flow optimization. Leveraging the characteristics of the embedded constraint neural network, the problem of exponentially increasing computational complexity under multiple complex constraints is avoided. At the same time, there is no need to oversimplify the physical model of the power system. The resulting power flow optimization decision not only solves the problems of existing methods failing to meet real-time scheduling requirements and the optimization results being out of touch with reality, but also balances the goals of safe operation and economic dispatch of the power system, achieving efficient and reliable power flow optimization under dynamic scenarios and precise constraints.

[0051] In one embodiment, such as Figure 2 As shown, the process of steps 2011-2015 includes: Step 2011: Based on the multi-dimensional features in the current operating status data and the power system operating scenario feature library constructed by the power system digital twin, perform similarity processing to obtain the similarity values ​​between the multi-dimensional features and the features of each historical scenario.

[0052] Optionally, the power flow optimization device first extracts multi-dimensional features from the current operating status data. These multi-dimensional features must fully cover the core operating parameters of the power system, specifically including generator output power characteristics (including the real-time output power of all generators), load power consumption characteristics (including the real-time power consumption of all residential and industrial loads), bus voltage amplitude characteristics (including the real-time voltage amplitude of all buses), branch transmission power characteristics (including the real-time transmission power of all branches), branch power flow direction characteristics (including the real-time power flow direction of all branches, represented by numerical codes, such as "1" representing the direction from bus A to bus B, and "0" representing the opposite direction), bus voltage deviation characteristics (including the real-time voltage deviation of all buses), and generator output margin characteristics (including the real-time output margin of all generators).

[0053] Subsequently, based on the pre-built power system operation scenario feature library in the power system digital twin, which stores multi-dimensional feature samples corresponding to different scenarios (such as off-peak, peak, off-peak, and fault recovery scenarios) during historical operation, each scenario sample contains a feature vector that corresponds one-to-one with the current multi-dimensional features. Therefore, an improved cosine similarity algorithm is used to calculate the similarity between the current multi-dimensional feature vector and the feature vectors of each historical scenario in the scenario feature library, obtaining the similarity value between each historical scenario and the current operating state. The similarity formula is: ; in, For vectors sum vector The similarity between them, with a value range of [0,1]. For the current number Class feature vectors , For historical scenes Class feature vectors For the first Class feature weights, where · represents the vector dot product. The representative vector is the 2-norm. This formula, based on traditional cosine similarity, introduces feature weight coefficients to assign higher weights to features that are more critical to the operating state of the power system (such as generator output power and bus voltage amplitude), thereby improving the accuracy of similarity calculation.

[0054] In one embodiment, taking the current operating status data of the power system in the western region of a city as an example, the multi-dimensional feature vectors extracted by the power flow optimization device are as follows: generator output power feature vector (Unit: MW, corresponding to generators G1 and G2 respectively); Load power consumption characteristic vector (Unit: MW, corresponding to residential load L1 and industrial load L2 respectively); Bus voltage amplitude characteristic vector (Unit: kV, corresponding to bus B1 and B2 respectively); Branch transmission power characteristic vector (Unit: MW, corresponding to branch Z1); Branch power flow characteristic vector (Code "1" represents the direction from bus B1 to bus B2, corresponding to branch Z1); Bus voltage deviation characteristic vector (Unit: kV, corresponding to busbars B1 and B2 respectively); Generator output margin characteristic vector (Unit: MW, corresponding to generators G1 and G2 respectively).

[0055] The power system operation scenario feature library stores feature vectors for three typical historical scenarios: 1. Feature vector of off-peak scene ; 2. Peak Scene Feature Vector ; 3. Feature vector of a low-level scene ; Define the feature weight coefficient vector (The weights corresponding to generator output power, load power consumption, bus voltage amplitude, branch transmission power, branch power flow direction, bus voltage deviation, and generator output margin characteristics are calculated in sequence.) (Similarity between current features and off-peak scene features); (Similarity between current features and peak scene features); (Similarity between current features and features in low-temperature scenarios).

[0056] Step 2012: Based on the similarity value and a preset similarity threshold, a selection of candidate scene groups is obtained.

[0057] Optionally, the power flow optimization device presets a similarity threshold. This threshold is determined based on the stability requirements of the power system operation scenarios. Specifically, by statistically analyzing the correspondence between similarity values ​​and scenario matching accuracy in historical operations, the minimum similarity value that achieves a scenario matching accuracy of over 90% is selected as the preset threshold (e.g., the preset similarity threshold is 0.8). The obtained similarity values ​​of each historical scenario are compared with the preset similarity threshold, and historical scenarios with similarity values ​​greater than or equal to the preset threshold are selected and grouped into a candidate scenario group. If no scenario meets the criteria after selection (i.e., all historical scenario similarity values ​​are less than the preset threshold), the power flow optimization device includes the two historical scenarios with the highest similarity values ​​into the candidate scenario group to ensure that there are enough scenarios for subsequent analysis. If the number of scenarios that meet the criteria after selection exceeds five, the five scenarios with the highest similarity values ​​are selected into the candidate scenario group to avoid excessive computational complexity in subsequent steps.

[0058] Continuing with the above embodiments, the preset similarity threshold is 0.8. Comparing the similarity values ​​of each historical scene with the threshold reveals that: Off-peak scene similarity value (0.93 ≥ 0.8) meets the condition; peak scene similarity value (0.68 < 0.8) does not meet the condition; and low-peak scene similarity value (0.72 < 0.8) does not meet the condition. Therefore, the power flow optimization device selects the off-peak scene as the only scene that meets the condition, forming a candidate scene group. ={Off-peak scenario}.

[0059] Step 2013: Based on the candidate scenario group and the power system digital twin, the constraint conditions are extracted to obtain the power operation constraints corresponding to each candidate scenario. The correlation between the power operation constraints of each candidate scenario and the current power system operation constraints is analyzed to obtain the constraint correlation value of each candidate scenario.

[0060] Optionally, the power flow optimization device extracts constraints for each candidate scenario in the candidate scenario group based on the power system digital twin, and obtains the power operation constraints corresponding to each candidate scenario. These power operation constraints refer to the parameter restrictions that must be met to maintain the safe and stable operation of the power system under the candidate scenario. Specifically, they include generator constraints (such as the maximum output power, minimum output power, and output change rate limit of the generator), bus constraints (such as the upper limit of bus voltage amplitude and the lower limit of bus voltage amplitude), branch constraints (such as the maximum transmission power of the branch and the upper limit of branch power loss), and load constraints (such as the minimum power supply limit corresponding to the load power supply reliability requirements).

[0061] Next, based on the extracted power operation constraints for each candidate scenario and the current power system operation constraints (i.e., the initial constraints before correction, specifically including the current generator initial maximum output power, bus initial voltage amplitude range, branch initial maximum transmission power, etc.), the correlation value between the power operation constraints of each candidate scenario and the current constraints is calculated. The formula is as follows: ; in, To constrain the correlation degree value; These represent the generator, busbar, branch, and load constraint types, respectively. For the first The number of parameters in a class constraint; For candidate scenarios Class constraint Each parameter value (if it is a range constraint, take the average of the upper and lower limits); For the current system Class constraint The parameter value (if it is a range constraint, the average of the upper and lower limits is taken). This correlation value is used to quantify the degree of overlap between the two in terms of constraint parameter range and constraint priority. The higher the correlation value, the better the constraint conditions of the candidate scenario match the current system constraint conditions.

[0062] Continuing with the above embodiments, candidate scenario group ={Peak Scenario}, the power operation constraints for the peak scenario extracted from the power system digital twin are as follows: Generator constraints: G1 maximum output power 86.45MW, minimum output power 30MW, G2 maximum output power 73.72MW, minimum output power 20MW, output change rate ≤5MW / min; Bus constraints: B1 voltage amplitude 9.682-10.51kV, B2 voltage amplitude 9.5-10.5kV; Branch constraints: Z1 maximum transmission power 59.78MW, power loss ≤2MW; Load constraints: L1 and L2 power supply power ≥95% of rated load (L1 rated 50MW, L2 rated 70MW). The current power system operating constraints are as follows: Generator constraints: G1 maximum output power 86.45MW, minimum output power 30MW; G2 maximum output power 73.72MW, minimum output power 20MW; output change rate ≤ 5MW / min; Bus constraints: B1 voltage amplitude 9.682-10.51kV; B2 voltage amplitude 9.5-10.5kV; Branch constraints: Z1 maximum transmission power 59.78MW; power loss ≤ 2MW; Load constraints: L1 and L2 supply power ≥ 95% of rated load (L1 rated 50MW, L2 rated 70MW). The calculated correlation value between the off-peak scenario and the current constraints is Corr = 4.0 (out of 4.0, representing a perfect match between the two constraints).

[0063] Step 2014: Based on the constraint correlation value of each candidate scenario and the trend analysis of the power system digital twin, obtain the power flow change trend data of each candidate scenario in the corresponding operating period, and perform power flow consistency analysis based on the power flow change trend data and the current real-time power flow change data of the power system to obtain the consistency coefficient between the power flow trend of each candidate scenario and the current power flow trend.

[0064] Optionally, the power flow optimization device simulates and calculates the power flow trend of each candidate scenario within the corresponding operating period (e.g., the next hour is divided into 4 time nodes with 15-minute intervals) based on the power operation constraints and the power system digital twin of each candidate scenario, and obtains power flow trend data. This power flow trend data includes the predicted value of generator output power, the predicted value of load power consumption, the predicted value of bus voltage amplitude, and the predicted value of branch transmission power at each time node.

[0065] Subsequently, based on the acquired real-time power flow data, which consists of historical power flow data recorded at 15-minute intervals over the past hour (consistent with the time interval of trend analysis, totaling four time points), including the actual values ​​of generator output power, load power consumption, bus voltage amplitude, and branch transmission power at each time point, the consistency coefficient between the power flow trend data and the current real-time power flow data for each candidate scenario is calculated using the power flow consistency formula. The power flow consistency formula is as follows: ; in, The consistency coefficient; This is the time node number, with a value ranging from 1 to 4; The attenuation coefficient is set to 2.0. The values ​​range from 1 to 4, representing the changes in the generator's total output power, total load power consumption, average bus voltage, and branch transmission power, respectively. For candidate scenarios Time Node The change in class parameters (difference from the previous node); For the current system Time Node The change in class parameters (difference from the previous node, when...) When the contribution is zero, it means that the direction of change is consistent and the magnitude is the same. The consistency coefficient is used to quantify the degree of consistency between the two in terms of the direction and magnitude of change. The higher the consistency coefficient, the more consistent the trend of the candidate scenario is with the trend of the current system.

[0066] Continuing with the above embodiments, taking the off-peak scenario as an example, when the power flow optimization device performs trend analysis through the power system digital twin, it simulates the next hour (time node) under the off-peak scenario. , , , The power flow trend data represents the generator output power trend: , Load power consumption trend: , Bus voltage amplitude trend: kV, kV; Branch transmission power trend: .

[0067] Real-time power flow data of the current power system (past 1 hour, time node) , , , Actual generator output power: , Actual power consumption of the load: , Actual value of bus voltage amplitude: kV, kV; Actual value of branch transmission power: .

[0068] The consistency coefficient for off-peak scenarios is then calculated using the power flow consistency formula. This indicates that its trend is highly consistent with the current trend.

[0069] Step 2015: Prioritize the candidate scenarios based on the consistency coefficient and the constraint correlation value, and determine the candidate scenario with the highest priority as the current running scenario.

[0070] Optionally, the power flow optimization device uses the obtained constraint correlation values ​​of each candidate scenario. And obtain the consistency coefficient A non-linear weighted algorithm is used to calculate the priority score of each candidate scene, and the formula is as follows: when and hour, ; when or hour, ; in, Score the priority of candidate scenarios; The maximum constraint correlation value among all candidate scenarios; This represents the maximum consistency coefficient among all candidate scenarios. This formula uses an exponential function to enhance the scoring advantage of scenarios with high relevance and high consistency, avoiding the insufficient score discrimination problem caused by traditional linear weighting.

[0071] Continuing with the above embodiments, candidate scenario group ={off-peak scenario}, its constraint correlation value ( Consistency coefficient ( Substitute into the priority scoring formula: .

[0072] Since the candidate scenario group only contains off-peak scenarios, it is directly identified as the current running scenario.

[0073] This application embodiment uses a complete scenario matching process, including feature similarity calculation, candidate scenario screening, constraint correlation verification, power flow consistency matching, and priority ranking, to ensure that the determined current operating scenario accurately matches the real-time operating status of the power system, thereby improving the accuracy and reliability of the current operating scenario determination.

[0074] In one embodiment, such as Figure 3 As shown, the process of steps 2021-2024 includes: Step 2021: Based on the operating parameters of power system components in the current operating scenario and the physical relationships of components stored in the power system digital twin, determine the key operating constraints, and construct a constraint relationship graph with each key operating constraint as a node and the influence intensity between different key operating constraints as edges.

[0075] Optionally, the power flow optimization device first extracts the operating parameters of power system components from the current operating scenario, specifically including generator operating parameters (generator output power, generator output margin, generator speed), bus operating parameters (bus voltage amplitude, bus voltage deviation, bus load factor), and branch operating parameters (branch transmission power, branch power flow direction, branch temperature). Then, based on the physical relationships of components stored in the power system digital twin, which refers to the actual connection logic between generators, buses, and branches (e.g., generators are connected to the corresponding bus through specific branches, and buses are connected to other buses or loads through branches) and parameter influence relationships (e.g., changes in generator output power will cause changes in the voltage amplitude of the connected bus, and changes in branch transmission power will affect the branch temperature), it determines the key operating constraints. These key operating constraints refer to the constraint parameters that directly affect the safe and stable operation of the power system, specifically including the maximum generator output power constraint, the minimum generator output power constraint, the upper limit constraint of the bus voltage amplitude, the lower limit constraint of the bus voltage amplitude, the maximum branch transmission power constraint, and the maximum branch temperature constraint.

[0076] Then, using each key operational constraint as an independent node, the weights of the edges are determined by calculating the influence strength between different key operational constraints, thus constructing a constraint association graph. The influence strength is calculated using a mutual information algorithm to quantify the degree to which a change in the parameter of one constraint affects the parameter of another constraint. The influence strength ranges from [0, 1], with values ​​closer to 1 indicating a more significant influence.

[0077] In one embodiment, taking the off-peak scenario of a power system in the western region of a city as an example, the operating parameters of the power system components extracted by the power flow optimization device are as follows: Generator operating parameters: G1 output power 85MW, output margin 15MW, speed 3000r / min; G2 output power 65MW, output margin 10MW, speed 3000r / min; Bus operating parameters: B1 voltage amplitude 10.4kV, voltage deviation 0.15kV, load rate 75%; B2 voltage amplitude 10.2kV, voltage deviation 0.12kV, load rate 70%; Branch operating parameters: Z1 transmission power 45MW, power flow direction B1→B2, temperature 45℃; Z3 (G1-B1) transmission power 85MW, temperature 42℃; Z4 (G2-B2) transmission power 65MW, temperature 40℃.

[0078] The physical relationships between components in a digital twin include: G1 is connected to B1 via Z3, G2 is connected to B2 via Z4, and B1 and B2 are connected via Z1; changes in the output power of G1 will affect the transmission power of Z3 and the voltage amplitude of B1, and changes in the transmission power of Z1 will affect the voltage amplitude of B2 and the temperature of Z1.

[0079] Based on this, the key operational constraints are determined as follows: N1: Maximum output power constraint of G1 (86.45MW); N2: Maximum output power constraint of G2 (73.72MW); N3: Upper limit constraint of voltage amplitude of B1 (10.51kV); N4: Upper limit constraint of voltage amplitude of B2 (10.5kV); N5: Maximum transmission power constraint of Z1 (59.78MW); N6: Maximum temperature constraint of Z1 (80℃, equipment rated value).

[0080] The influence strengths were calculated using the mutual information algorithm: W(N1-N3) = 0.85 (G1 output power constraint affects B1 voltage constraint); W(N1-N5) = 0.70 (G1 output power constraint affects Z1 transmission power constraint); W(N5-N6) = 0.90 (Z1 transmission power constraint affects Z1 temperature constraint); W(N5-N4) = 0.80 (Z1 transmission power constraint affects B2 voltage constraint); W(N2-N4) = 0.82 (G2 output power constraint affects B2 voltage constraint). The influence strengths between other nodes were all less than 0.5 (e.g., W(N1-N6) = 0.32), which were considered weak associations and edges were not constructed for the time being. The final constrained association graph is constructed as follows: nodes are N1-N6, and edges and weights are N1-N3 (0.85), N1-N5 (0.70), N5-N6 (0.90), N5-N4 (0.80), and N2-N4 (0.82).

[0081] Step 2022: Based on the influence coefficients of each node in the constraint correlation graph and the operational stability data of each component in the current operating scenario, the potential disturbance source types are analyzed to obtain the potential disturbance source types. Based on the potential disturbance source types and the characteristics of the triggering probability and triggering time interval of different disturbance sources in the power system operation rules, the disturbance triggering sequence is determined. The disturbance triggering sequence includes the expected triggering time, disturbance intensity level and influence range of each potential disturbance source.

[0082] Optionally, the power flow optimization device extracts the influence coefficient of each node in the constraint correlation graph. This influence coefficient indicates the importance of the node in the graph, and its formula is: ; in, For nodes The PageRank value represents its importance; the higher the value, the more important it is. For nodes PageRank value; The damping coefficient is 0.85. Pointing to a node The set of nodes; For nodes The set of nodes it points to; For nodes arrive Edge weights; in the denominator Traversal All nodes in the system are considered. Nodes with higher impact coefficients correspond to more critical constraints, and their abnormal parameters are more likely to cause system disturbances. Therefore, by combining the operational stability data of each component in the current operating scenario (including the component's historical failure frequency, the percentage difference between the current operating parameters and the constraint boundary, and the component's aging assessment score), potential disturbance source types are screened. The screening rule is as follows: if the impact coefficient of a critical constraint node corresponding to a component is ≥0.7, and the component's operational stability data meets the following conditions: historical failure frequency ≥1 time / year, current operating parameters ≤10% of the difference between the current operating parameters and the constraint boundary, or aging assessment score ≤0.8, then the abnormal parameters of that component are defined as potential disturbance source types.

[0083] Furthermore, after identifying the types of potential disturbance sources, the power flow optimization device determines the expected trigger time, disturbance intensity level, and impact range for each potential disturbance source based on the power system operation data stored in the power system digital twin (including historical trigger probabilities and trigger time interval statistics for different disturbance sources). The expected trigger time is calculated using a Poisson process model; the disturbance intensity level is categorized as slight (parameters exceeding constraint boundaries by less than 5%), moderate (5%-10%), and severe (above 10%); and the impact range is determined based on the node impact range corresponding to the disturbance source in the constraint correlation graph. Finally, the above information is integrated into a disturbance trigger sequence.

[0084] Continuing with the above embodiments, based on the constructed constraint correlation graph, the influence coefficients of each node are calculated as follows: PR(N1) = 0.82 (node ​​corresponding to G1, affecting N3 and N5); PR(N2) = 0.78 (node ​​corresponding to G2, affecting N4); PR(N3) = 0.75 (node ​​corresponding to B1 voltage, affected by N1); PR(N4) = 0.80 (node ​​corresponding to B2 voltage, affected by N5 and N2); PR(N5) = 0.88 (node ​​corresponding to Z1 transmission power, affected by N1, affecting N6 and N4); PR(N6) = 0.72 (node ​​corresponding to Z1 temperature, affected by N5). The operational stability data for each component in the current operating scenario are as follows: G1: Historical fault frequency 2 times / year (winding aging), current output power difference from maximum constraint percentage 1.68% ((86.45-85) / 86.45), aging score 0.7 (≤0.8); G2: Historical fault frequency 0 times / year, current output power difference from maximum constraint percentage 11.83% ((73.72-65) / 73.72), aging score 0.9 (>0.8); Z1: Historical fault frequency 1 time / year (lightning strike). The current transmission power is 24.7% of the difference between the maximum constraint and the current power ((59.78-45) / 59.78), and the aging score is 0.8 (=0.8); B1: historical fault frequency 0 times / year, current voltage is 1.05% of the difference between the upper limit constraint and the current voltage ((10.51-10.4) / 10.51), and there is no aging score (no obvious aging of the bus); B2: historical fault frequency 0 times / year, current voltage is 2.86% of the difference between the upper limit constraint and the current voltage ((10.5-10.2) / 10.5), and there is no aging score.

[0085] Therefore, according to the screening rules, the potential disturbance source types are: sudden increase in G1 output power disturbance (G1 output power exceeds the maximum constraint); sudden increase in Z1 transmission power disturbance (Z1 transmission power exceeds the maximum constraint); sudden increase in B1 voltage amplitude disturbance (B1 voltage exceeds the upper limit constraint).

[0086] Based on the power system operation data, the disturbance trigger sequence is determined as follows: G1 Output Sudden Increase Disturbance: The expected trigger time is calculated using a Poisson process model. The historical average trigger interval is 180 days, and the expected trigger time is 24 hours after the current time. Disturbance intensity level: moderate (8% beyond the constraint boundary, i.e., the output increases to 86.45×1.08≈93.36MW). Affected range: B1 voltage amplitude (N3), Z1 transmission power (N5).

[0087] Z1 power surge disturbance: The expected trigger time is 48 hours after the current time; Disturbance intensity level: slight (3% above the constraint boundary, i.e., the power surges to 59.78×1.03≈61.57MW); Affected range: B2 voltage amplitude (N4), Z1 temperature (N6).

[0088] B1 voltage amplitude surge disturbance: The expected trigger time is synchronized with the G1 disturbance (t=24h); Disturbance intensity level: moderate (6% above the constraint boundary, i.e., the voltage rises to 10.51×1.06≈11.14kV); Affected range: no other related nodes (B1 has no downstream constraint nodes).

[0089] Step 2023: Based on the expected triggering time, disturbance intensity level and impact range of each disturbance source in the disturbance triggering sequence, and combined with the dynamic characteristics of the components in the power system digital twin, the disturbance propagation process in the power system is simulated to obtain the disturbance propagation process. Based on the disturbance propagation process and the changes in the operating parameters of each component in the power system at different times, the dynamic response data of each component is obtained.

[0090] Optionally, the power flow optimization device combines the dynamic characteristics of components in the power system digital twin (i.e., component dynamic characteristic models) based on the determined disturbance triggering sequence. These models include generator dynamic models (describing the relationship between generator output changes and speed and excitation current), bus dynamic models (describing the relationship between bus voltage changes and input / output power), and branch dynamic models (describing the relationship between branch transmission power changes and temperature and impedance). For each disturbance source, according to its expected triggering time, the parameter changes corresponding to the disturbance intensity level are input into the component dynamic characteristic model to simulate the disturbance propagation process in the power system. The propagation process simulation uses a time-domain simulation method with a time step of 0.01 seconds and a simulation duration of 10 seconds after disturbance triggering (covering the complete process of disturbance occurrence, propagation, and transient stabilization). During the simulation, the operating parameters of each component in the power system are recorded in real time at different times (recorded once every 0.1 seconds), including instantaneous values ​​of parameters such as generator output power, bus voltage amplitude, branch transmission power, and branch temperature. Finally, the recorded parameter change data is organized to obtain the dynamic response data of each component. The dynamic response data of the component includes the correspondence between "time and parameter value", as well as key characteristics such as the peak value, stable value, and response time (the time from disturbance triggering to the parameter reaching a stable value).

[0091] Continuing with the above embodiments, taking the disturbance triggering sequence as an example of "G1 output surge disturbance (t=24h, output rises to 93.36MW)," the power flow optimization device calls the component dynamic characteristic model to simulate disturbance propagation. For the generator dynamic model: the increase in G1 excitation current causes the output to surge from 85MW to 93.36MW (triggered at t=24h), and the speed briefly rises to 3015r / min (t=24h+0.2s), and then stabilizes to 3000r / min through the speed regulation system (t=24h+2s). For the bus dynamic model: the increase in G1 output causes the Z3 transmission power to increase from... The voltage increases from 85MW to 93.36MW (t=24h+0.1s), the input power of B1 increases, and the voltage amplitude rises from 10.4kV to 11.14kV (t=24h+0.3s, corresponding to the sudden voltage rise disturbance of B1), and then stabilizes at 10.50kV (t=24h+3s) through the reactive power compensation device. For the branch dynamic model: the increase in B1 voltage causes the transmission power of Z1 to increase from 45MW to 52MW (t=24h+0.5s), and the increase in Z1 current causes the temperature to rise from 45℃ to 58℃ (t=24h+1s), and then stabilizes at 55℃ (t=24h+4s).

[0092] The changes in the operating parameters of each component at critical moments were recorded, and the dynamic response data of the components were compiled as shown in Table 1 below: Table 1 Component Dynamic Response Data

[0093] Similarly, by simulating the "Z1 transmission power surge disturbance", dynamic response data of Z1, B2 and G2 are obtained, such as the peak transmission power of Z1 being 61.57MW and the peak voltage of B2 being 10.82kV.

[0094] Step 2024: Based on the analysis of the number of components exceeding the critical operational constraint deviation threshold in the dynamic response data of each component and the number of affected constraints in the constraint correlation graph, the scenario types of multiple potential emergency scenarios are obtained.

[0095] Optionally, the power flow optimization device pre-sets deviation thresholds for each key operating constraint. These deviation thresholds refer to the allowable range of operating parameters exceeding the constraint boundary, specifically: a slight deviation threshold (within 5% of the constraint boundary), a moderate deviation threshold (5%-10% of the constraint boundary), and a severe deviation threshold (more than 10% of the constraint boundary). The deviation thresholds correspond to the disturbance intensity level in step 2022. Then, the dynamic response data of each component is analyzed, and the number of components exceeding the deviation thresholds of the key operating constraints is counted: for the dynamic response data corresponding to each disturbance source, it is checked whether the peak value of the operating parameter of each component exceeds the deviation threshold of the corresponding constraint. If it does, it is recorded as an "affected component," and the total number of affected components is counted (if multiple parameters of the same component exceed the threshold, it is counted as one component). Simultaneously, based on the constraint association graph, the number of affected constraints is counted: according to the key constraint nodes corresponding to the affected components, combined with the edge association relationships in the graph, all constraint nodes indirectly affected by the disturbance are identified (e.g., disturbance G1 affects N1, indirectly affecting N3 and N5 through edges N1-N3 and N1-N5), and the total number of these constraint nodes (including directly affected nodes) is counted. Finally, based on the number of components exceeding the deviation threshold and the number of affected constraints, potential outbreak scenarios are classified as follows: Local outbreak scenario: number of affected components ≤ 2, number of affected constraints ≤ 2; Regional outbreak scenario: number of affected components 3-5, number of affected constraints 3-5; Global outbreak scenario: number of affected components ≥ 6, number of affected constraints ≥ 6.

[0096] Continuing with the above embodiments, taking the component dynamic response data from step 2023, the key operating constraint deviation thresholds are set as: slight (≤5%), moderate (5%-10%), and severe (>10%).

[0097] In the G1 power surge disturbance (moderate disturbance), the number of components exceeding the deviation threshold is counted as follows: G1: peak power output 93.36MW, exceeding the maximum constraint of 86.45MW by 8% (within the moderate deviation threshold), recorded as an affected component; B1: peak voltage 11.14kV, exceeding the upper limit constraint of 10.51kV by 6% (within the moderate deviation threshold), recorded as an affected component; Z1: peak transmission power 52MW, not exceeding the maximum constraint of 59.78MW (deviation approximately 13%, but actually not exceeding the constraint, and the deviation threshold corresponding to the disturbance is moderate 10%, so it is not recorded as an affected component); number of affected components = 2. The number of affected constraints is counted as follows: directly affected node: N1 (G1 constraint); indirectly affected nodes: N3 (B1 voltage constraint, through the N1-N3 edge), N5 (Z1 transmission power constraint, through the N1-N5 edge); number of affected constraints = 3. The process for determining the scenario type is as follows: if the number of affected components is 2≤2 and the number of affected constraints is 3 in the range of 3-5, the higher impact level is taken when combining the two, and the scenario is determined to be a regional sudden scenario (i.e., a regional voltage anomaly scenario caused by a sudden increase in G1 output).

[0098] In the Z1 transmission power surge disturbance (minor disturbance), the number of components exceeding the deviation threshold is statistically analyzed as follows: Z1: peak transmission power 61.57MW, exceeding the maximum constraint of 59.78MW by 3% (within the minor deviation threshold), is recorded as an affected component; B2: peak voltage 10.82kV, exceeding the upper limit constraint of 10.5kV by approximately 3.05% (not exceeding the minor deviation threshold of 5%), is not recorded as an affected component; number of affected components = 1. The number of affected constraints is statistically analyzed as follows: directly affected node: N5 (Z1 transmission power constraint); indirectly affected node: N6 (Z1 temperature constraint, through the N5-N6 edge); number of affected constraints = 2. The scenario type determination process is as follows: if the number of affected components 1≤2 and the number of affected constraints 2≤2, it is determined to be a local sudden scenario (i.e., a local temperature rise scenario caused by a sudden increase in Z1 transmission power).

[0099] In the B1 voltage amplitude surge disturbance (moderate disturbance), the statistical process of the number of components exceeding the deviation threshold is as follows: only the B1 voltage exceeds the threshold (peak value 11.14kV, exceeding 10.51kV by 6%), the number of affected components = 1; the statistical process of the number of affected constraints is as follows: only N3 is affected, the number = 1; the scenario type determination process is as follows: determined to be a local sudden scenario (i.e., B1 voltage surge scenario).

[0100] Three potential emergency scenarios were ultimately identified: regional emergency scenario (sudden increase in G1 output causing regional voltage anomaly), local emergency scenario (sudden increase in Z1 transmission power causing local temperature rise), and local emergency scenario (sudden voltage rise in B1).

[0101] This application embodiment uses a complete test case generation process of constraint association modeling, disturbance source screening, propagation process simulation, and scenario classification to enable the generated potential sudden scenario types to accurately match the disturbance risks that the power system may face in actual operation, covering different impact ranges from local to global, and solving the problems of existing methods being unable to quantify the degree of disturbance impact and lacking specificity in scenario generation.

[0102] In one embodiment, such as Figure 4 As shown, the process of steps 301-305 includes: Step 301: Based on the historical fault database of the power system digital twin, select historical fault records that are consistent with the power system type of the current operating status data to obtain a historical fault record group, and extract the fault association feature information related to the power flow constraint boundary in each historical fault record in the historical fault record group to obtain a historical fault association feature group.

[0103] Optionally, the power flow optimization device uses the historical fault database stored in the power system digital twin. Since this historical fault database contains fault records of different power system types, each historical fault record covers the fault occurrence time, fault equipment type (generator, bus, branch, etc.), fault cause (aging, lightning strike, overload, etc.), system operating status data at the time of the fault occurrence (such as the operating parameters of the fault equipment and the operating parameters of related equipment), and power flow constraint boundary deviation data caused by the fault (such as the deviation between the actual value of the generator's maximum output power and the initial constraint value, and the deviation between the actual value of the branch's maximum transmission power and the initial constraint value). Therefore, based on the power system type to which the current operating status data belongs, the device selects historical fault records of the same type from the historical fault database to form a historical fault record group. Subsequently, for each record in the historical fault record group, fault association feature information related to the power flow constraint boundary is extracted. This fault association feature information refers to the fault attributes that directly cause changes in the power flow constraint boundary parameters. Specifically, it includes fault equipment type characteristics, fault cause characteristics, equipment operating parameter characteristics at the time of the fault (such as the duration of continuous high load operation of the equipment before the fault, the ratio of the equipment operating parameters before the fault to the initial constraint value), and constraint boundary deviation characteristics caused by the fault (such as deviation direction, deviation amplitude, and deviation duration). Finally, a historical fault association feature group is formed.

[0104] In one embodiment, taking the power system in the western region of a city (type "urban distribution network-industrial park hybrid power grid") as an example, the historical fault database is accessed to filter out historical fault record groups of this type, which contain a total of 3 key records. Among them, record 1: the faulty equipment is a generator (model is the same as G1), the cause of the fault is "winding aging", the operating parameters at the time of the fault are "output power accounts for 90% of the initial maximum constraint for 72 consecutive hours", and the constraint boundary deviation is "maximum output power drops from 100MW to 88MW (deviation 12MW, deviation magnitude 12%)"; record 2: the faulty equipment is a branch (model is the same as Z1). Record 1: The fault was caused by lightning strike. The operating parameters at the time of the fault were: "transmission power accounted for 85% of the initial maximum constraint 1 hour before the fault" and "constraint boundary deviation: maximum transmission power decreased from 60MW to 55MW (deviation 5MW, deviation magnitude 8.3%)". Record 2: The faulty equipment was a busbar (model same as B1). The fault was caused by insulation aging. The operating parameters at the time of the fault were: "voltage amplitude was close to the initial upper limit constraint of 10.7kV for 24 hours before the fault" and "constraint boundary deviation: upper limit of voltage amplitude decreased from 10.7kV to 10.5kV (deviation 0.2kV, deviation magnitude 1.9%)".

[0105] Therefore, the fault association features of each record are extracted to form the historical fault association feature group as follows: Feature F1 (corresponding to record 1): Fault equipment type = generator, fault cause = aging, high load duration before fault = 72h, parameter percentage before fault = 90%, deviation amplitude = 12%, deviation direction = constraint value decrease; Feature F2 (corresponding to record 2): Fault equipment type = branch, fault cause = lightning strike, high load duration before fault = 1h, parameter percentage before fault = 85%, deviation amplitude = 8.3%, deviation direction = constraint value decrease; Feature F3 (corresponding to record 3): Fault equipment type = busbar, fault cause = insulation aging, high load duration before fault = 24h, parameter percentage before fault = 98% (voltage amplitude 10.6kV / 10.7kV), deviation amplitude = 1.9%, deviation direction = constraint value decrease.

[0106] Finally, a historical fault association feature group is formed, in which F1, F2, and F3 correspond to the feature information of record 1 (generator aging fault), record 2 (branch lightning strike fault), and record 3 (bus insulation aging fault) respectively.

[0107] Step 302: Identify the operating parameters that affect the power flow constraint boundary in the current operating status data to obtain the current power flow influence factor group, and determine the attribute information of each power flow influence factor based on the current power flow influence factor group.

[0108] Optionally, the power flow optimization device identifies operating parameters that affect the power flow constraint boundary based on the determined current operating status data. These parameters are called current power flow influencing factors. The identification rule is as follows: if a change in an operating parameter causes a change in at least one parameter in the power flow constraint boundary (such as the maximum output power of a generator or the maximum transmission power of a branch), then that operating parameter is included in the current power flow influencing factor group. Common current power flow influencing factors include the current output power of a generator, the operating years of a generator, the current transmission power of a branch, the operating years of a branch, the current voltage amplitude of the bus, and the equipment aging assessment score. Then, based on the current power flow influencing factor group, the attribute information of each power flow influencing factor is determined. This attribute information includes factor type (such as "equipment parameter type", "time type", "assessment type"), factor-related constraint type (such as the current output power of a generator being associated with "maximum output power constraint of a generator"), current factor value, normal operating value range of the factor (obtained from the equipment parameter library of the power system digital twin), and factor change sensitivity coefficient (quantifying the impact of each 1% change in the factor on the constraint boundary, obtained through historical data regression analysis). Finally, a list containing the correspondence between "influencing factors and attribute information" is formed.

[0109] Continuing with the above embodiments, based on the current operating status data and the historical fault association characteristics of step 301, the power flow optimization device identifies the current power flow influencing factor group and determines the attribute information. Finally, a list containing the correspondence between "influence factors and attribute information" is formed, as shown in Table 2 below, where I1 to I7 correspond to the current output power of G1, the operating years of G1, the aging assessment score of G1, the current transmission power of Z1, the operating years of Z1, the aging assessment score of Z1, and the current voltage amplitude of B1, respectively.

[0110] Table 2 lists the correspondence between "Impact Factor - Attribute Information".

[0111] Step 303: Based on the historical fault association feature group and the current tidal current influence factor group, analyze the correlation between each historical fault association feature and each current tidal current influence factor, and construct a mapping relationship group between historical fault association features and current tidal current influence factors.

[0112] Optionally, the power flow optimization device analyzes the correlation between each feature in the historical fault association feature group and each factor in the current power flow influence factor group, based on the determined historical fault association feature group and the current power flow influence factor group. Specifically, firstly, the historical fault association features and the current power flow influence factors are aligned by attribute. For example, "fault equipment type = generator" and "current power flow influence factor = G1 current output power" are both associated with generator equipment, and "fault cause = aging" and "current power flow influence factor = G1 aging degree assessment score" are both associated with aging attributes.

[0113] Next, the correlation strength between each historical fault correlation feature and each current power flow influence factor is calculated. Its formula is: ; in, For the first Historical fault association characteristics; For the first One of the current trend influencing factors; To contain simultaneously and The number of historical samples; For inclusion But not included The number of historical samples; For inclusion But not included The number of historical samples; For which it does not contain It does not include The number of historical samples.

[0114] Finally, based on the pre-set correlation strength threshold (e.g., CS≥3.84, corresponding to 95% confidence level), if the correlation strength between a historical fault correlation feature and the current trend influence factor exceeds the threshold, a mapping relationship between the two is established. The mapping relationship includes the correlation feature ID, influence factor ID, correlation strength value, and correlation logic (e.g., "fault cause = aging" → "G1 aging degree assessment score": aging faults are directly related to aging scores), ultimately forming a mapping relationship group.

[0115] Continuing with the above embodiments, taking the historical fault association characteristics (F1-F3) from step 301 and the current power flow influence factors (I1-I7, corresponding to the 7 factors in the table) from step 302, when calculating the association strength, for F1 (generator aging fault) and I1 (G1 current output power): A=5 (high output power before 5 generator aging faults in history), B=3 (3 high output power did not trigger aging faults), C=2 (no high output power in 2 aging faults), D=10 (no high output power or aging fault in 10 faults), substituting into the formula, we get CS=4.2≥3.84, thus establishing a mapping relationship; F1 and I2 (G1 operating years): A=6, B=2, C=1, D=1, D=2, B=3, C=1, D=1 ... =11, CS=5.8≥3.84, establish mapping relationship; F1 and I3 (G1 aging score): A=7, B=1, C=0, D=12, CS=7.3≥3.84, establish mapping relationship; F2 (branch lightning fault) and I4 (Z1 current transmission power): A=4, B=4, C=3, D=9, CS=3.1<3.84, no mapping established; F2 and I5 (Z1 operating years): A=3, B=5, C=4, D=8, CS=2.5<3.84, no mapping established; F3 (bus insulation aging) and I7 (B1 current voltage amplitude): A=4, B=3, C=2, D=11, CS=3.9>3.84, establish mapping.

[0116] The final mapping relationship groups are: F1→I1 (association strength 4.2, logic: generator aging fault is related to current high output power); F1→I2 (association strength 5.8, logic: generator aging fault is related to long service life); F1→I3 (association strength 7.3, logic: generator aging fault is related to low aging score); F3→I7 (association strength 3.9, logic: bus insulation aging is related to current voltage approaching the upper limit).

[0117] Step 304: Based on the historical fault association characteristics and current power flow influence factor corresponding to each mapping relationship in the mapping relationship group, calculate the deviation amount caused by the corresponding historical fault to the power flow constraint boundary when it occurs, obtain the historical fault deviation amount, and based on the historical fault deviation amount and the current power flow influence factor, calculate the influence degree value of the historical fault on the current power flow constraint boundary, and obtain the historical fault influence degree group.

[0118] Optionally, the power flow optimization device, based on the determined mapping relationship group, extracts the "fault-induced constraint boundary deviation feature" (e.g., the deviation amplitude of F1 is 12%) and the "current factor value" and "normal factor value range" of the current power flow influence factor from the corresponding historical fault association features for each mapping relationship in the mapping relationship group, and calculates the historical fault deviation amount. This historical fault deviation amount refers to the actual deviation value of the power flow constraint boundary parameter when the historical fault occurred, and its formula is: ;in For the first The historical fault deviation corresponding to each mapping relationship; The initial values ​​for the constraints associated with this mapping relationship; This represents the deviation magnitude (percentage) in the historical fault association characteristics.

[0119] Next, the impact of historical faults on the current power flow constraint boundary is calculated using the following formula: ; in, For the first The degree of influence of each mapping relationship; The influence coefficient is set to 0.5. The similarity between the current trend influence factor and the correlation characteristics of historical failures (e.g., the similarity between the current G1 operating age of 12 years and the historical failure operating age of 10 years is 83.3%). The standard similarity is set to 100%. If multiple mapping relationships exist that are associated with the same constraint boundary parameter, the maximum value of each influence degree value is taken as the final influence degree of the constraint, forming a historical failure influence degree group.

[0120] Continuing with the above embodiment, based on the mapping relationship group obtained in step 303, taking the mapping relationship (F1→I1, F1→I2, F1→I3) associated with the maximum output power constraint of G1 as an example, when calculating the historical fault deviation, F1 is associated with the initial constraint of the maximum output power of G1. Deviation range ,but When calculating similarity, F1→I1: the output power ratio before the historical fault was 90%, and the current output power ratio of G1 is 85% (85 / 100), with a similarity of... F1→I2: Historical fault operating years: 10 years; Current G1 operating years: 12 years; Similarity: F1→I3: Historical fault aging score 0.6, current G1 aging score 0.7, similarity (The higher the score, the higher the similarity); when calculating the degree of influence, ; ; ; Finally, determine the ultimate impact of the G1 constraint: take the maximum value. Similarly, calculate the maximum transmission power constraint of Z1 (no effective mapping relationship, impact level 0MW) and the voltage amplitude constraint of B1 (F3→I7, impact level...). , The final impact of historical faults is as follows: G1 maximum output power constraint impact: 13.04MW; Z1 maximum transmission power constraint impact: 0MW; B1 voltage amplitude upper limit constraint impact: 0.18kV.

[0121] It should be noted that the ratio of the historical fault aging score to the current (G1) aging score is used to quantify the degree of similarity between the current state and the historical fault state. A ratio greater than 100% indicates that the current aging level is lower than the aging level at the time of the historical fault.

[0122] Step 305: Based on the historical fault impact level group, power flow impact factor attribute information and power system digital twin, deviation correction is performed to obtain the constraint boundary parameters.

[0123] Optionally, the power flow optimization device corrects the deviation of the power flow constraint boundary based on the determined historical fault impact level group, power flow impact factor attribute information and power system digital twin, and finally obtains the constraint boundary parameters, as in steps 3051-3055.

[0124] This application's embodiments solve the problem of discrepancies between the preset fixed constraint boundaries and the actual capabilities of the equipment in existing methods by filtering fault data, identifying influencing factors, constructing correlation relationships, and quantifying the degree of influence. This allows the corrected constraint boundary parameters to not only consider static factors such as equipment aging and historical faults, but also incorporate the dynamic influence of the current operating state, thus improving accuracy.

[0125] In one embodiment, such as Figure 5 As shown, the process of steps 3051-3055 includes: Step 3051: Based on the digital twin of the power system, determine the theoretical standard range of the power flow constraint boundary under the current operating scenario.

[0126] Optionally, the power flow optimization device determines the theoretical standard range of each power flow constraint boundary based on the equipment parameter library and system operation standard library stored in the power system digital twin, as well as the current operating scenario, combined with equipment design parameters and industry standards. The equipment parameter library contains the design parameters (such as rated power, rated voltage, rated temperature), factory test parameters, and constraint ranges from equipment manuals for all equipment (generators, buses, branches, etc.) in the current power system. The system operation standard library stores the requirements for power flow constraint boundaries in national and industry-issued power system operation standards (such as "GB / T15543-2019 Power Quality Three-Phase Voltage Imbalance" and "DL / T1870-2018 Power System Power Flow Calculation Technical Specification"). The theoretical standard range refers to the reasonable value range of the power flow constraint boundary when the equipment is in ideal operating conditions (no aging, no faults, no external interference) and meets industry standard requirements. Specifically, this includes the theoretical range of generator maximum / minimum output power, the theoretical range of upper / lower limits of bus voltage amplitude, the theoretical range of maximum branch transmission power, and the theoretical range of highest branch temperature. The upper limit of the theoretical standard range shall not exceed 105% of the equipment's rated parameters (to ensure the equipment's safety margin), and the lower limit shall not be less than 30% of the equipment's rated parameters (to meet basic operating requirements).

[0127] Continuing with the above embodiments, taking the power system in the western region of a city from step 30, the power flow optimization device extracts equipment parameters and industry standards from the digital twin, and determines the theoretical standard range of the power flow constraint boundary as shown in Table 3 below: Table 3 Theoretical Standard Range of Power Flow Constraint Boundaries

[0128] Among them, G1 has a rated output power of 100MW, with the theoretical standard range upper limit of 105MW (100×105%) and the lower limit of 30MW (the minimum stable operating power of the equipment); B1 has a rated voltage of 10kV (the industry standard allows a deviation of ±7% for the 10kV bus voltage), so the theoretical upper limit is 10×107%=10.7kV (the rated upper limit of the equipment), which is extended to 11.45kV (10.7×107%) to cover the standard allowable range.

[0129] Step 3052: Based on the current operating status data and the current power flow influencing factor attribute information, analyze the difference between the current operating status and the theoretical standard status to obtain the current operating status deviation, and determine the initial deviation range of the current power flow constraint boundary based on the current operating status deviation and the theoretical standard range of the power flow constraint boundary.

[0130] Optionally, the power flow optimization device extracts key operating parameters related to the power flow constraint boundary (e.g., current output power of G1 85MW, current transmission power of Z1 45MW, current voltage amplitude of B1 10.4kV) based on the current operating status data, and combines this with the current power flow influencing factor attribute information (e.g., normal operating range of factors, sensitivity coefficient of change) to construct a current operating status vector and a theoretical standard state vector. The parameter values ​​of the theoretical standard state vector are taken as the midpoint of the theoretical standard range (e.g., the theoretical standard range of G1 output power is 30-105MW, with a midpoint of 67.5MW; the theoretical range of B1 voltage amplitude is 9.5-11.45kV, with a midpoint of 10.475kV). The deviation of the current operating status is calculated using a percentage deviation formula, which is: ; in, This represents the current operating status deviation (percentage). The current running parameter value. This represents the theoretical standard state parameter value (midpoint of the range).

[0131] Next, based on the current operating state deviation and the theoretical standard range, the initial deviation range of the current power flow constraint boundary is determined. The calculation logic is as follows: subtract the theoretical standard value × current operating state deviation × change sensitivity coefficient from the upper and lower limits of the theoretical standard range, respectively. The formula is: ; ; in, This represents the upper limit of the initial deviation range. This represents the upper limit of the theoretical standard range. This is the lower limit of the initial deviation range. This represents the lower limit of the theoretical standard range. This is the sensitivity coefficient to changes in the current trend influence factors (take the average of the sensitivity coefficients of all factors associated with this constraint).

[0132] Continuing with the above embodiments, based on the theoretical standard range of step 3051 and the attribute information of step 302, we take the maximum output power constraint of G1, the maximum transmission power constraint of Z1, and the upper limit constraint of voltage amplitude of B1 as examples.

[0133] During the calculation of the maximum output power constraint for G1, the theoretical standard range is... 30-105MW, midpoint value Current running parameters (Current output power of G1); Current operating status deviation Average sensitivity coefficient of related influencing factors (Sensitivity coefficients of I1, I2, and I3); Upper limit of initial deviation range Initial deviation range lower limit The initial deviation range is .

[0134] In the calculation of the maximum transmission power constraint for Z1, the theoretical standard range is... 10-63MW, midpoint value Current running parameters (Current transmission power of Z1); Deviation Average sensitivity coefficient (Sensitivity coefficients of I4, I5, and I6); Initial deviation range: ; The range is 12.26MW-48.7MW.

[0135] In the calculation of the upper limit constraint of voltage amplitude B1, the theoretical standard range is... 9.5-11.45kV, midpoint value Current running parameters (Current voltage of B1); Deviation (Take the absolute value of 0.72%); Sensitivity coefficient (Sensitivity coefficient of I7); Initial deviation range: ; The range is 9.5205kV-11.425kV.

[0136] Step 3053: Based on the historical fault impact degree group, determine the comprehensive impact degree value of all relevant historical faults on the current power flow constraint boundary, and adjust the initial deviation range based on the comprehensive impact degree value to obtain the first power flow constraint boundary range after preliminary correction.

[0137] Optionally, the power flow optimization device extracts historical fault impact values ​​related to the current power flow constraint boundary based on the historical fault impact degree group (e.g., G1 maximum output power constraint impact degree 13.10MW, B1 voltage amplitude upper limit constraint impact degree 0.18kV). If multiple historical faults are associated with the same constraint boundary, a weighted summation formula is used to calculate the comprehensive impact degree value, the formula of which is: ; in, This represents the overall degree of impact. For the first The impact level of each historical fault; Weights are assigned based on the proximity of historical fault occurrence time to the present: faults within the last year have a weight of 0.6, faults within the last 1-3 years have a weight of 0.3, and faults older than 3 years have a weight of 0.1.

[0138] Then, the initial deviation range is adjusted based on the comprehensive impact value. The conditions are as follows: for upper limit constraints (such as maximum output power and upper voltage amplitude limit), the upper limit of the initial deviation range is subtracted from the comprehensive impact value; for lower limit constraints (such as minimum output power and lower voltage amplitude limit), the lower limit of the initial deviation range is added to the comprehensive impact value to obtain the first power flow constraint boundary range. That is, the adjustment formula is: (Upper limit constraint); (Lower limit constraint).

[0139] If the adjusted range exceeds the theoretical standard range, the intersection of the theoretical standard range and the adjusted range shall be taken as the final first range.

[0140] Continuing with the above embodiments, the initial deviation range obtained in step 3052 and the historical fault impact level obtained in step 304 are grouped together.

[0141] When correcting for the maximum output power constraint of G1, the initial deviation range is 41.67MW-64.16MW; the impact of historical faults. (Only 1 related fault, overall impact) First range upper limit); ; Lower limit of the first range (Lower limit > upper limit, take the minimum of the theoretical standard range lower limit of 30MW and the adjusted lower limit, and at the same time ensure that the lower limit ≤ upper limit, so it is corrected to 30MW-51.12MW); then the first power flow constraint boundary range is 30MW-51.12MW (actually take 30-51.12MW, satisfying the lower limit ≤ upper limit and covering the minimum operating power of the equipment).

[0142] When correcting for the maximum transmission power constraint of Z1, the initial deviation range is 12.26MW-48.7MW; the impact of historical faults. (No associated faults) First range upper limit); ; Lower limit of the first range Therefore, the first power flow constraint boundary range is 12.26MW-48.7MW.

[0143] When correcting for the upper limit constraint of voltage amplitude B1, the initial deviation range is 9.5205kV-11.425kV; the impact of historical faults. (Comprehensive impact level) First range upper limit); ; Lower limit of the first range Therefore, the first power flow constraint boundary range is 9.7005kV-11.245kV.

[0144] Step 3054: Based on the scenario feature information that affects the power flow constraint boundary in potential sudden scenarios, calculate the supplementary impact of each potential sudden scenario on the range of the first power flow constraint boundary, obtain the supplementary impact deviation value, and adjust the range of the first power flow constraint boundary again based on the supplementary impact deviation value to obtain the second power flow constraint boundary range after secondary correction.

[0145] Optionally, the power flow optimization device, based on the identified potential sudden scenarios, selects scenarios that affect the current power flow constraint boundary from the potential sudden scenario types (such as "regional voltage anomaly scenario caused by a sudden increase in G1 output" and "local temperature rise scenario caused by a sudden increase in Z1 transmission power"). It then extracts the scenario feature information of these scenarios, including the constraint boundary type affected by the scenario (such as maximum output power of G1 and maximum transmission power of Z1), the scenario disturbance intensity level (slight / moderate / severe), and the duration of the scenario impact (such as 5 minutes / 10 minutes). For each potential sudden scenario affecting the power flow constraint boundary, a supplementary impact deviation value is calculated using the scenario impact coefficient formula, the formula of which is: ,in To supplement the factors affecting the deviation value, This represents the upper limit of the first current flow constraint boundary range. The scene disturbance intensity coefficients are (0.05 for slight, 0.1 for moderate, and 0.2 for severe). The duration coefficients for scenario impact are (0.5 for 5 minutes, 0.8 for 10 minutes, and 1.0 for 15 minutes). If multiple potential sudden scenarios are associated with the same constraint boundary, the maximum value of the supplementary impact deviation (the most unfavorable scenario) is taken. Then, the supplementary impact deviation value is subtracted from the upper limit of the first power flow constraint boundary range, while the lower limit remains unchanged (potential sudden scenarios mainly affect the upper limit of the constraint, with a smaller impact on the lower limit), to obtain the second power flow constraint boundary range. If the adjusted upper limit is lower than the lower limit, the average of the upper and lower limits of the first range is taken as the upper limit of the second range.

[0146] Continuing with the above embodiments, based on the first range obtained in step 3053 and the obtained potential sudden scenarios (regional sudden scenario: sudden increase in G1 output; local sudden scenario: sudden increase in Z1 transmission power).

[0147] For the secondary correction process of the maximum output power constraint of G1, the first range is 30MW-51.12MW; associated with potential sudden scenarios: a sudden increase in G1 output (moderate disturbance). The effect lasts for 10 minutes. ); Supplement the influence deviation value Second range upper limit Second range lower limit (Unchanged); then the second power flow constraint boundary range: 30MW- MW.

[0148] For the secondary correction process of the maximum transmission power constraint of Z1, the first range is 12.26MW-48.7MW; associated with potential sudden scenarios: a sudden increase in Z1 transmission power (slight disturbance). The effect lasts for 5 minutes. ); Supplement the influence deviation value Second range upper limit Second range lower limit (If it remains unchanged); then the range of the second power flow constraint boundary is: 12.26MW-47.4825MW.

[0149] For the secondary correction process of the upper limit constraint of voltage amplitude of B1, the first range is 9.7005kV-11.245kV; related potential sudden scenario: a sudden increase in output of G1 causing voltage anomaly (moderate disturbance). The effect lasts for 10 minutes. ); Supplement the influence deviation value Second range upper limit Second range lower limit (If it remains unchanged); then the range of the second power flow constraint boundary is: 9.7005kV-10.345kV.

[0150] Step 3055: Based on the second power flow constraint boundary range, combined with the safety requirements of power system operation and actual operating specifications, a rationality verification is performed, and the parameters corresponding to the second power flow constraint boundary range that pass the verification are determined as constraint boundary parameters.

[0151] Optionally, the power flow optimization device performs a rationality check on the second power flow constraint boundary range based on the safety requirements of power system operation (such as the safety threshold requirements for frequency, voltage, and power in the "Guidelines for Power System Safety and Stability") and actual operating procedures (such as the restrictions on equipment operation in the dispatching operation procedures). The specific check process includes: Safety threshold verification: The upper limit of the second range shall not exceed the industry safety threshold (e.g., the upper limit of 10kV bus voltage shall not exceed 11kV), and the lower limit shall not be lower than the safety threshold (e.g., the minimum output power of the generator shall not be lower than 20MW). Equipment capacity verification: The second range must be within the actual operating capacity of the equipment (e.g., if the actual maximum output of G1 has been reduced to 86.43MW due to aging, the upper limit of the second range must be ≤86.43MW). Run compatibility checks: Different constraint boundaries must be compatible (e.g., the upper limit of the G1 maximum output power constraint must be greater than the current output power of 85MW). If not, then the current output power must be within the constraint range, so if the upper limit of the second range is not met... (MW < current 85MW, adjustment needed); Historical data verification: The second range must deviate from the constraint boundary parameters of similar historical scenarios by ≤10% (e.g., in historical off-peak scenarios, the average upper limit of G1 constraint is 50MW, and the upper limit of the second range is ≤10%). The MW deviation is 3%, which meets the requirements.

[0152] If the verification fails (e.g., the upper limit of the second voltage range of B1, 10.3454kV, is less than the current voltage of 10.4kV), then return to step 3054 to adjust the scene influence coefficient. or If the verification passes, the upper and lower limits of the second power flow constraint boundary range are taken as the final constraint boundary parameters (for parameters without lower limit constraints, only the upper limit is taken).

[0153] Continuing with the above embodiments, the verification is performed based on the second scope of step 3054, combined with safety requirements and operating procedures.

[0154] When verifying the maximum output power constraint of G1, the second range is 30MW-47.0304MW; Safety threshold verification: Upper limit 47.0304MW < industry safety threshold 105MW, lower limit 30MW > safety threshold 20MW, passed; Equipment capability verification: G1's actual maximum output is 86.43MW (result from step 305), 47.0304MW < 86.43MW, passed; Operational compatibility verification: Current G1 output power 85MW > 47.0304MW, failed; Adjustment: Return to step 3054, reduce the scenario impact coefficient α to 0.05 (moderate disturbance changed to slight), and recalculate the supplementary impact deviation value. =51.12×0.05×0.8=2.0448MW, the upper limit of the second range is 51.12-2.0448=49.0752MW; still not meeting 85MW, further adjust α=0 (ignore the scene influence), the upper limit of the second range is 51.12MW; finally, combined with the result of step 305 86.43MW, take 86.43MW as the upper limit (compatible with the current operation), and the lower limit is 30MW; the constraint boundary parameters after verification are obtained: 30MW-86.43MW (upper limit 86.43MW, lower limit 30MW).

[0155] When verifying the maximum transmission power constraint of Z1, the second range is 12.26MW-47.4825MW; safety threshold verification: upper limit 47.4825MW < 63MW, lower limit 12.26MW > 10MW, passed; equipment capability verification: actual maximum transmission power of Z1 is 59.78MW (result of step 305), 47.4825MW < 59.78MW, passed; operation compatibility verification: current transmission power 45MW < 47.4825MW, passed; after verification, the constraint boundary parameters are 12.26MW-47.4825MW (finally, the result of step 305, 59.78MW, is taken as the upper limit, and adjusted to 12.26MW-59.78MW).

[0156] For the B1 voltage amplitude upper limit constraint verification, the second range is 9.7005kV-10.3454kV; safety threshold verification: upper limit 10.3454kV < 11kV, lower limit 9.7005kV > 8.55kV, passed; equipment capability verification: actual B1 voltage upper limit 10.51kV (result of step 305), 10.3454kV < 10.51kV, passed; operational compatibility verification: current voltage 10.4kV > 10.3454kV, failed; adjustment: reduce α. To 0.03, supplement the influence deviation value 11.245×0.03×0.8=0.26988kV, the upper limit of the second range is 11.245-0.26988=10.97512kV (approximately 10.975kV); the verification is passed (10.4kV<10.975kV); after the verification is passed, the constraint boundary parameters are: 9.7005kV-10.975kV (finally take the result of step 305, 10.51kV, as the upper limit, and adjust it to 9.7005kV-10.51kV).

[0157] The final constraint boundary parameters are as follows: G1 maximum output power: 86.43MW, minimum output power: 30MW; Z1 maximum transmission power: 59.78MW, minimum transmission power: 12.26MW; B1 voltage amplitude upper limit: 10.51kV, lower limit: 9.7005kV.

[0158] This application's embodiments, through theoretical benchmark determination, current deviation analysis, historical fault adjustment, supplementation of contingency scenarios, and safety verification confirmation, ultimately obtain corrected and verified constraint boundary parameters. This solves the problem of existing methods that only consider a single factor (such as historical faults) causing a disconnect between constraint boundaries and actual operating conditions. The corrected constraint boundary parameters simultaneously cover theoretical standards, current operating conditions, historical faults, and future disturbances, thus improving accuracy compared to traditional methods.

[0159] In one embodiment, such as Figure 6 As shown, the process of steps 401-405 includes: Step 401: In the scene mapping module, based on the generator output power, load consumption power, bus voltage amplitude and branch transmission power in the current operating status data, and combined with the topological connectivity relationship and equipment operating mode characteristics corresponding to the current operating scenario, determine the scene status association feature group; and based on the scene status association feature group, construct a one-to-one mapping relationship between the current operating status and the current operating scenario.

[0160] Optionally, the power flow optimization device extracts core parameters from the current operating status data in the scene mapping module, including generator output power (such as G185MW, G265MW), load power consumption (such as L155MW, L275MW), bus voltage amplitude (such as B110.4kV, B210.2kV), and branch transmission power (such as Z145MW, Z385MW, Z465MW).

[0161] Subsequently, combining the topological connectivity relationships (e.g., G1 connected to B1 via Z3, G2 connected to B2 via Z4, and B1-B2 connected via Z1) and equipment operation mode characteristics (e.g., generators using "economic operation mode", branches using "stable transmission mode", and buses using "voltage stability control mode") corresponding to the current operating scenario (off-peak scenario) in the power system digital twin, a scenario state association feature group is determined. This feature group contains three-dimensional features of "parameter-topology-mode", specifically including: the association features between generator output power and topological nodes (e.g., the proportion of load supplied to node B1 by G1 power), the association features between load power and topological paths (e.g., the proportion of L1 load supplied by G1 via the Z3-Z5 path), the association features between bus voltage and equipment operation mode (e.g., voltage regulation sensitivity under B1 voltage stability control mode), and the association features between branch transmission power and topological connectivity (e.g., the proportion of Z1 power to the total transmission capacity of B1-B2).

[0162] At the same time, based on the determined scene state association feature group, a one-to-one mapping relationship between the current running state and the current running scene is constructed, and the mapping strength of each scene state association feature is calculated (the value range is [0, 1], and the closer the value is to 1, the tighter the mapping). The formula for the mapping strength is: ; in, The current running state is number 1 Each parameter value; For the current scene Each characteristic standard value; , These are the maximum values ​​of the corresponding parameters. The average mapping strength of all features... If so, the mapping relationship is deemed valid.

[0163] In one embodiment, taking the established off-peak scenario and current operating status data as an example, when the power flow optimization device constructs the scenario status association feature group and mapping relationship, for the scenario status association feature group: Feature 1: The ratio of G1 output power (85MW) to B1 node load supply (G1 supply to B1 load ratio = G1 power / (B1 connected load power) = 85 / (55) = 154.5%); Feature 2: The ratio of L1 load (55MW) supplied by G1 power supply path (Z3-Z5 path power supply ratio = 100%, because L1 is only connected to B1, B1 is mainly supplied by G1); Feature 3: The adjustment sensitivity of B1 voltage (10.4kV) in stable control mode (for every 0.1kV voltage fluctuation, the reactive power compensation device response time ≤ 0.5s); Feature 4: The ratio of Z1 transmission power (45MW) to the total transmission capacity of B1-B2 (Z1 power / corrected maximum Z1 transmission power = 45 / 59.78 ≈ 75.3%).

[0164] When constructing the mapping relationship, the standard values ​​for features in off-peak scenarios are: Feature 1 supply ratio 140%-160%, Feature 2 path ratio ≥95%, Feature 3 sensitivity ≤0.5s, and Feature 4 capacity ratio 70%-80%; the calculated mapping strength is: (Current supply ratio is 154.5%, average scenario standard is 150%, maximum is 160%) (Current path percentage is 100%, scenario standard average is 97.5%) (Current response time: 0.5s; average response time for the scenario: 0.4s). (Currently accounting for 78.9%, the average of the scenario standard is 75%); Average mapping intensity = (0.91 + 0.98 + 0.8 + 0.882) / 4 ≈ 0.893 ≥ 0.85, the mapping relationship is valid, that is, the current running status is mapped one-to-one with the off-peak scenario.

[0165] Step 402: In the risk level quantification module, based on the equipment failure type, failure impact range and load loss ratio corresponding to each potential emergency scenario, determine the scenario risk assessment factor, and based on the scenario risk assessment factor and the preset risk level classification standard, determine the risk level corresponding to each potential emergency scenario; the risk level includes low level, medium level and high level.

[0166] Optionally, in the risk level quantification module, the power flow optimization device extracts key information about potential sudden scenarios, including equipment fault type (e.g., branch Z1 failure disconnection, load L1 sudden increase), fault impact range (e.g., affecting B2 bus power supply, affecting G1 output regulation), and load loss ratio (e.g., fault causing 20% ​​load loss in B2 area, load sudden increase causing 5% load unmet). Based on the above information, scenario risk assessment factors are determined, with a total of 3 core factors, namely Factor 1: Fault Severity factor (S), assigned according to equipment failure type (branch failure S=0.6, load surge S=0.4, generator failure S=0.8); Factor 2: Scope of impact factor (R), assigned according to the number of affected equipment (affecting 1-2 equipment R=0.3, 3-5 equipment R=0.6, ≥6 equipment R=0.9); Factor 3: Loss degree factor (L), assigned according to the proportion of load loss (loss ≤5% L=0.2, 5%-10% L=0.5, >10% L=0.8).

[0167] Next, the risk value for each potential emergency scenario is calculated using a risk quantification formula, which is as follows: ;in, , , Factor weights were assigned (determined using the analytic hierarchy process, with fault severity having the highest weight). Finally, based on the preset risk level classification criteria: low risk (Risk ≤ 0.3), medium risk (0.3 < Risk ≤ 0.6), and high risk (Risk > 0.6), the risk level for each scenario was determined.

[0168] Continuing with the above embodiments, taking two potential sudden scenarios (load surge scenario and branch Z1 fault scenario) as examples, when the power flow optimization device calculates the risk level, for the load surge scenario, the scenario information is as follows: equipment fault type is "load surge" (no hardware fault, it belongs to abnormal operating parameters), the fault impact range is "G1, Z3, B1" (3 devices), and the load loss ratio is "3%" (the total load increases from 130MW to 146.5MW, the total generator output is 150MW, which can meet the requirements, and the loss is only 3%); risk assessment factors: S=0.4 (load surge assignment), R=0.6 (affecting 3 devices), L=0.2 (loss 3%≤5%); risk value. Risk level: 0.3 < 0.40 ≤ 0.6, classified as medium risk.

[0169] Step 403: In the constraint dynamic adaptation module, based on the constraint boundary parameters and the topological features of the current operating scenario and the risk level of potential sudden scenarios, the initial value of the scenario adaptation constraint is determined. Based on the initial value of the scenario adaptation constraint and the actual operating conditions of the equipment under the current operating scenario and the parameter fluctuations caused by potential sudden scenarios, the value range of each constraint parameter is adjusted to obtain the target constraint boundary.

[0170] Optionally, in the constraint dynamic adaptation module, the power flow optimization device determines the initial value of the scenario adaptation constraint based on the determined constraint boundary parameters (such as G1 maximum output power 86.45MW, Z1 maximum transmission power 59.78MW, B1 voltage 9.682-10.51kV), combined with the topology characteristics of the current operating scenario (off-peak scenario) (such as B1-B2 connected through Z1, G1 being the main power supply for B1) and the risk level of potential sudden scenarios (medium risk). The adjustment rule for the initial value of the scenario adaptation constraint is as follows: If the risk level is low: initial constraint value = corrected constraint boundary parameter × 1.0 (no adjustment); if the risk level is medium: initial constraint value = corrected constraint boundary parameter × 0.95 (5% safety margin reserved); if the risk level is high: initial constraint value = corrected constraint boundary parameter × 0.9 (10% safety margin reserved).

[0171] Subsequently, based on the initial values ​​of the scenario adaptation constraints, and considering the actual operating conditions of the equipment under the current operating scenario (e.g., G1's current output is 85MW, Z1's current power is 45MW) and parameter fluctuations caused by potential sudden scenarios (e.g., a sudden increase in load causing power fluctuations of ±5MW in G1 and ±3MW in Z1), the value ranges of each constraint parameter are adjusted. The adjustment formula constraint upper limit adjustment value is... (The upper limit value minus the maximum fluctuation value ensures that the constraint is not exceeded after fluctuation); the lower limit of the constraint is adjusted to... (The lower limit plus the maximum fluctuation value ensures that the value does not fall below the constraint after fluctuation); where To adapt the scene, constrain the initial upper limit. As the initial lower bound, This represents the maximum parameter fluctuation caused by potential sudden scenarios. The final target constraint boundary (including adjusted upper and lower limits) is obtained.

[0172] Continuing with the above embodiments, taking the constraint boundary parameters and the medium-level risk in step 402 as an example, the process by which the power flow optimization device determines the target constraint boundary is as follows: During the adjustment of the maximum output power constraint for G1, the corrected constraint boundaries are: 86.45MW (upper limit) and 30MW (lower limit); initial values ​​for scenario adaptation constraints (medium-level risk): , (Lower limit adjusted to 28.5MW, with margin reserved); Actual operating conditions of equipment: G1 current output 85MW > 82.13MW, initial value needs adjustment; Potential sudden parameter fluctuations: Sudden load increase causes G1 power fluctuation ±5MW (maximum fluctuation 5MW); Adjusted target constraints: (To ensure that even with the current 85MW fluctuations, the safety threshold is not breached, the upper limit is forcibly lowered to trigger optimization.) The final target constraint boundary is 33.5MW (lower limit) - 77.13MW (upper limit).

[0173] During the adjustment of the maximum transmission power constraint for Z1, the corrected constraint boundaries are: 59.78MW (upper limit) and 12.26MW (lower limit); initial values ​​for scenario adaptation constraints: ; Actual operating conditions of the equipment: Z1 current power 45MW < 56.79MW, which meets the operating conditions; Potential sudden scenario parameter fluctuation: Branch fault scenario causes Z1 power fluctuation ±3MW (maximum fluctuation 3MW); Adjusted target constraints: ; Therefore, the target constraint boundary is 14.65MW-53.79MW.

[0174] For the adjustment process of the B1 voltage amplitude constraint, the corrected constraint boundary is: 9.682kV (lower limit) - 10.51kV (upper limit); initial values ​​of the scenario adaptation constraint: , Actual operating conditions of the equipment: B1 current voltage 10.4kV > 9.98kV, adjustment is required; Potential sudden parameter fluctuations: sudden load increase causing voltage fluctuations of ±0.2kV (maximum fluctuation 0.2kV); Target constraints after adjustment: ; Therefore, the target constraint boundary is 9.40kV-9.78kV.

[0175] Step 404: Based on the mapping intensity index of the target constraint boundary and the scene state associated feature group, determine the constraint activation priority sequence, and check and activate the running parameters corresponding to each constraint condition based on the constraint activation priority sequence to obtain the activated constraint condition group.

[0176] Optionally, the power flow optimization device determines the evaluation index of constraint activation priority based on the determined target constraint boundary and the mapping strength index of the scene state associated feature group in step 401 (such as feature 1 mapping strength 0.91, feature 2 0.98, feature 3 0.8, feature 4 0.92). This evaluation index includes index 1: the mapping strength between constraint parameters and scene features (…). The stronger the mapping, the higher the priority; Indicator 2: Risk sensitivity of constraint parameters ( This refers to the increase in risk value triggered by the violation of this constraint, which is assigned a value according to the constraint type (generator power constraint). Branch power constraints Bus voltage constraint Indicator 3: Current deviation of constraint parameters ( The percentage deviation of the current parameter from the target constraint boundary; the smaller the deviation (the closer to the constraint), the higher the priority. .

[0177] Then, the priority calculation formula is used: ;in, This is a deviation correction term (the smaller the deviation, the closer this term is to 1). The calculated Priority values ​​are sorted from largest to smallest to form a constraint activation priority sequence. Then, based on this sequence, the operating parameters corresponding to each constraint are checked (e.g., checking whether the current power of G1, 85MW, is within the target constraint of 33.5-77.13MW). If the parameter is within the constraint range, the constraint is activated (marked as "effectively activated"); if the parameter is outside the constraint range, the operating parameters are adjusted to be within the constraint before activation (marked as "activated after adjustment"). Finally, the activated constraint group is obtained.

[0178] Continuing with the above embodiments, based on the mapping strength determined in step 401 and the target constraint boundary determined in step 403, the power flow optimization device determines the constraint activation priority as follows: First, calculate the evaluation index (taking 3 core constraints as an example). For the maximum output power constraint of G1, M: the corresponding feature 1 mapping intensity is 0.91; (Generator power constraints); Priority = 0.91 × 0.4 + 0.8 × 0.3 + 0.898 × ​​0.3 ≈ 0.364 + 0.24 + 0.269 ≈ 0.87. For the maximum transmission power constraint of Z1: M: corresponding to feature 4 mapping strength 0.92; (Branch power constraints); Priority = 0.92 × 0.4 + 0.6 × 0.3 + 0.837 × 0.3 ≈ 0.368 + 0.18 + 0.251 = 0.799. For the voltage amplitude constraint in B1, M: corresponds to a feature 3 mapping intensity of 0.8. (Bus voltage constraint); Current value 10.4kV, upper limit 9.78kV. ;Priority=0.8×0.4+0.7×0.3+0.937×0.3≈0.32+0.21+0.281=0.811.

[0179] The priority sequence for determining the constraint activation is: G1 maximum output power constraint (0.873) > B1 voltage magnitude constraint (0.811) > Z1 maximum transmission power constraint (0.799).

[0180] Finally, the constraints are checked and activated: G1 power constraint: The current 85MW is not within the range of 33.5-77.13MW, and the constraint needs to be lowered for effective activation. The constraint is "G1 output power ≤ 77.13MW"; B1 voltage constraint: The current 10.4kV is not within the range of 9.40-9.78kV, and the constraint needs to be lowered for effective activation. The constraint is "B1 voltage amplitude ≤ 9.78kV and ≥ 9.40kV"; Z1 power constraint: The current 45MW is within the range of 14.65-53.79MW, and the constraint is effective for activation. The constraint is "Z1 transmission power ≤ 53.79MW"; The activated constraint group includes the above three effective activated constraints, as well as the G2 power constraint, B2 voltage constraint, etc.

[0181] Step 405: Based on the constraint condition set, combined with the current running status data, the scene status association feature set and the target constraint boundary, power flow optimization is performed to obtain the power flow optimization decision.

[0182] Optionally, the power flow optimization device propagates forward in the embedded neural network based on the determined set of constraints, current operating state data, scene state associated feature set, and target constraint boundary, and finally obtains the power flow optimization decision, as in steps 4051-4054.

[0183] This application's embodiments, through scene mapping, risk quantification, constraint adaptation, and priority activation, ultimately determine power flow optimization decisions within an embedded constraint neural network. This results in three major advantages for the optimization decisions: first, scene adaptability, ensuring optimization fits the current operating scenario through precise mapping; second, risk foresight, reserving safety margins based on risk levels to address potential unforeseen scenarios; and third, constraint effectiveness, ensuring constraints are not violated through priority activation and dynamic adaptation. Ultimately, this improves the safety and economy of power flow optimization.

[0184] In one embodiment, such as Figure 7 As shown, the process of steps 4051-4054 ​​includes: Step 4051: In the conflict location module, conflict identification is performed based on the branch power flow direction, bus voltage deviation and generator output margin of the current operating status data in combination with the constraint condition group to obtain the power flow operation conflict characteristics. Based on the power flow operation conflict characteristics, the deviation of each activated constraint condition and the corresponding operating parameter is compared one by one to determine the conflict location result.

[0185] Optionally, in the conflict location module, the power flow optimization device identifies conflicts based on the activated constraint set (G1 output power ≤ 77.13MW, Z1 transmission power ≤ 53.79MW, B1 voltage amplitude 9.40-9.78kV, etc.), combined with the branch power flow direction (e.g., Z1 power flows from B1 to B2), bus voltage deviation (B1 current 10.4kV deviates from the target constraint upper limit of 9.78kV by 0.62kV, B2 current 10.2kV deviates from the target constraint upper limit of 10.5kV by -0.3kV), and generator output margin (G1 current 85MW, G2 current 65MW). The conflict identification process involves: matching degree between branch power flow direction and constraint direction (whether the flow direction causes the power to exceed the constraint upper limit), bus voltage deviation rate (the ratio of deviation value to constraint interval width), and generator output margin rate (the ratio of margin value to constraint upper limit). The above indicators were quantitatively analyzed to extract conflict characteristics and obtain the power flow operation conflict characteristics, which are specifically manifested as the correspondence between "indicator anomaly and constraint correlation".

[0186] Subsequently, based on the extracted power flow conflict characteristics, the deviation between each activated constraint and the corresponding operating parameter is compared one by one. That is, the deviation between the actual value of the parameter and the constraint boundary is calculated (e.g., G1 current 85MW deviates from the upper limit 77.13MW by 7.87MW, B1 current 10.4kV deviates from the upper limit 9.78kV by 0.62kV). Then, combined with the power flow conflict characteristics, the conflict location result is determined, and the type of constraint, associated equipment, degree of deviation and risk level involved in the conflict are clarified.

[0187] Continuing with the above embodiments, based on the constraint set in step 404 and the current operating data, during the conflict identification and localization process, the power flow optimization device first calculates the conflict identification index: Branch power flow direction matching degree with constraint direction: Z1 power flows from B1 to B2, constraint upper limit 53.79MW, current 45MW, matching degree = 1-(53.79-45) / 53.79≈83.7%; Bus voltage deviation rate: B1 deviation rate = |10.4-9.78| / (9.78-9.40) = 0.62 / 0.38 ≈ 163.2% (severely exceeding the limit); B2 deviation rate = |10.2-10.5| / (10.5-9.5) = 0.3 / 1.0 = 30%; Generator output margin ratio: G1 output margin ratio = (77.13-85) / 77.13≈-10.2% (exceeding the limit); G2 output margin ratio = (73.72-65) / 73.72≈11.83%.

[0188] Further extracting the power flow operation conflict characteristics, characteristic 1: the output margin rate of G1 is negative, which has exceeded the target constraint limit of 77.13MW, and there is an emergency power over-constraint conflict; characteristic 2: the voltage deviation rate of B1 is 163.2%, which far exceeds the target constraint limit of 9.78kV, and there is an emergency voltage over-limit conflict; characteristic 3: the power matching degree of Z1 is 83.7%, which is not over-limit but close to the constraint boundary. In potential scenarios, power fluctuations of ±3MW are likely to trigger the risk of over-constraint.

[0189] The final conflict location result was determined to be a conflict: Conflict 1: G1 power emergency over-constraint conflict, associated constraint "G1≤77.13MW", associated equipment G1, B1, L1, deviation amount +7.87MW, the degree of deviation is urgent, the risk level is high; Conflict 2: Emergency over-limit conflict of B1 voltage, associated constraint "B1 voltage ≤ 9.78kV", associated equipment B1, G1, Z3, deviation +0.62kV, the degree of deviation is urgent, and the risk level is high. Conflict 3: Potential over-constraint risk of Z1 power, associated constraint "Z1≤53.79MW", associated equipment Z1, B1, B2, deviation amount -8.79MW (margin), deviation degree is moderate, risk level is medium; Conflict 4: Warning of deviation between B2 voltage and upper limit, associated constraint "B2 voltage ≤ 10.5kV", associated devices B2, G2, Z4, deviation amount -0.3kV (from the upper limit), risk level low.

[0190] Step 4052: In the optimization direction planning module, based on the severity of the power flow operation conflict characteristics and the conflict location results, combined with the fault impact range of potential sudden scenarios with a high risk level, the optimization demand priority is determined, and the solution path of each conflict problem is analyzed based on the optimization demand priority to obtain the power flow optimization direction; the power flow optimization direction includes generator output adjustment, load transfer distribution, branch power flow reconfiguration and reactive power compensation configuration.

[0191] Optionally, in the power flow optimization module, the power flow optimization device first quantifies the severity of power flow conflict based on the obtained power flow operation conflict characteristics. To avoid comparison distortion caused by different physical dimensions, a normalized severity formula is used: ; in, This refers to the deviation between the operating parameters and the constraint boundaries. To constrain boundary values, The risk level coefficients for conflict association are (low level 1.0, medium level 1.5, high level 2.0). Constraint type weights (generator power 0.4, bus voltage 0.3, branch power 0.3) are assigned, and the constraints are sorted from largest to smallest according to severity values. Combined with conflict location results (number of associated devices, scope of impact), a preliminary basic sequence of optimization requirements priority is determined.

[0192] Then, considering the impact range of high-risk potential emergencies (such as the "G1 failure and shutdown scenario", with a risk value of 0.83), the scenario impact superposition coefficient is calculated: ; in, The number of conflicts within the scope of the scene's influence; The total number of conflicts is given. The conflict severity is then multiplied by the scenario impact factor to obtain the corrected severity, which is used to determine the priority of optimization requirements (the higher the corrected severity, the higher the priority).

[0193] Finally, based on the determined optimization requirements priority, the solutions to each conflict problem are analyzed, including: for power over-constraint conflicts, solutions include adjusting generator output (increasing margin) and reconfiguring branch power flow (changing power flow direction); for low voltage conflicts, solutions include configuring reactive power compensation (boosting voltage) and shifting loads (reducing voltage drop). Next, considering the operating characteristics of the power system (e.g., reactive power compensation is more direct for voltage regulation, while generator output adjustment is more effective for power balance), the direction of power flow optimization is determined. Each optimization direction needs to clearly define the applicable conflict type, implementation methods, and expected effects (e.g., "generator output adjustment" applies to the G1 power conflict; by increasing the output of G2, the burden on G1 is reduced, and the expected increase in G1 margin to over 5MW).

[0194] Continuing with the above embodiments, based on the conflict characteristics and high-risk potential sudden scenarios in step 4051, the power flow optimization device determines the priority and optimization direction, first calculating the conflict severity: Conflict 1 (G1 power): (7.87 / 77.13)×2.0×0.4≈0.0817; Conflict 2 (B1 voltage): (0.62 / 9.78)×2.0×0.3≈0.0380; Conflict 3 (Z1 power): (8.79 / 53.79)×1.5×0.3≈0.0735; Conflict 4 (B2 voltage): (0.3 / 10.5)×1.0×0.3≈0.0086.

[0195] Recalculate the scene impact superposition coefficient: The high-risk scenario (G1 failure shutdown) affects Conflict 1, Conflict 2, and Conflict 3. , ; Conflict 4 is not directly affected by this scenario. .

[0196] Next, determine the severity and priority of the optimization requirements after the correction: Conflict 1: 0.0817×1.75≈0.1430; Conflict 3: 0.0735×1.75≈0.1286; Conflict 2: 0.0380×1.75≈0.0665; Conflict 4: 0.0086×1.0≈0.0086.

[0197] The priority order is: Conflict 1 (0.1430) > Conflict 3 (0.1286) > Conflict 2 (0.0665) > Conflict 4 (0.0086).

[0198] Finally, the direction for trend optimization was determined: Regarding conflict 1 (G1 power emergency over-limit): the solution is to adjust the generator output (reduce G1 output to within 77.13MW and increase G2 output to share the load), and determine the optimization direction 1: generator output adjustment; Regarding conflict 3 (potential risk to Z1 power): the solution is branch power flow reconfiguration (enable Z2 shunting to reduce Z1 load rate), and the optimization direction 2 is determined to be: branch power flow reconfiguration; Regarding conflict 2 (emergency voltage overrun of B1): the solution is to configure reactive power compensation (connect reactors or disconnect capacitors on the B1 side) and adjust generator output (reduce reactive power output of G1). Optimization direction 3 is determined: reactive power compensation configuration. Regarding conflict 4 (low voltage at B2): the solution is to configure reactive power compensation (add capacitors to B2) and load transfer and distribution, and the optimization direction 4 is determined to be load transfer and distribution.

[0199] Step 4053: In the scheme decision output module, based on the power flow optimization direction and optimization requirement priority, combined with the constraint condition group and the scenario state association feature group, the optimization parameter analysis is performed to obtain the optimization parameter adjustment range. Based on the optimization parameter adjustment range, the specific adjustment strategies for each power flow optimization direction are combined to obtain the power flow optimization scheme set.

[0200] Optionally, in the scheme decision output module, the power flow optimization device performs optimization parameter analysis based on the determined power flow optimization direction and optimization requirement priority, combined with the activated constraint condition group and scenario state association feature group. Specifically, for each optimization direction, when analyzing the corresponding optimization parameter adjustment range, the dynamic compression effect of parameter fluctuations caused by potential sudden scenarios on the adjustment range needs to be considered, that is: adjusted parameter value ± maximum fluctuation value of potential scenario ≤ target constraint boundary.

[0201] Subsequently, based on the adjustment range of each optimization parameter, specific adjustment strategies for different optimization directions are combined (prioritizing the combination of strategies for high-priority optimization directions) to ensure that all parameters after combination still meet the constraints after superimposed potential scenario fluctuations, thus obtaining a set of power flow optimization schemes.

[0202] Continuing with the above embodiments, the process by which the power flow optimization device generates a set of optimization schemes based on the determined optimization direction and constraints, and scenario information, includes: Based on the determined optimization direction and constraints, and scenario information, the process by which the power flow optimization device generates a set of optimization schemes includes: First, determine the adjustment range of the optimized parameters after considering fluctuation compression: G2 output adjustment range: target constraint upper limit 73.72MW, potential scenario fluctuation ±3MW, then the safety upper limit = 73.72-3=70.72MW; currently 65MW, so the adjustment range is [65,70.72]MW; Z1 power adjustment range: target constraint upper limit 53.79MW, potential scenario fluctuation ±3MW, then the safety upper limit = 53.79-3=50.79MW; currently 45MW, so the adjustment range is [42,50.79]MW; Z2 power adjustment range: equipment rated 40MW, potential scenario fluctuation ±2MW, safety upper limit = 38MW; from power balance Z1+Z2=55MW, when Z1∈[42,50.79]MW, the safety upper limit is 38MW. When Z2∈[4.21,13]MW; B2 reactive power compensation capacity: target voltage 10.5kV, current 10.2kV, each Mvar increases by about 0.05kV, 6Mvar needs to be compensated to reach 10.5kV; considering fluctuation ±0.1kV, the safety target is set at 10.4kV, 4Mvar needs to be compensated; equipment upper limit 10Mvar, so the adjustment range is [3,8]Mvar; B2 load transfer: G2 maximum safe output 70.72MW, B2 current load 75MW, at least 4.28MW needs to be transferred; B1 can accept upper limit = G1 safe upper limit 77.13MW - current G1 supply to B1 load 55MW = 22.13MW; therefore, the transfer range is [4.28,10]MW.

[0203] Then, determine the optimal strategy combination to generate a set of schemes (all satisfying post-fluctuation constraints): Scheme 1 (prioritizing conflict 1+3): G2=70MW (≤70.72), Z1=46MW (≤50.79), Z2=9MW, compensation=5Mvar, transfer=6MW; Scheme 2 (balanced): G2=69MW, Z1=45MW, Z2=10MW, compensation=6Mvar, transfer=7MW; Scheme 3 (conservative, reserving greater margin): G2=68MW, Z1=44MW, Z2=11MW, compensation=7Mvar, transfer=8MW; Scheme 4 (aggressive, approaching the safety boundary): G2=70.5MW, Z1=48MW, Z2=7MW, compensation=4Mvar, transfer=5MW. Finally, the schemes were verified by fluctuation superposition: taking Scheme 1 as an example, G2=70MW+fluctuation 3MW=73MW≤73.72MW; Z1=46MW+fluctuation 3MW=49MW≤53.79MW; B2 voltage=10.2+0.05×5=10.45kV, after fluctuation -0.1kV=10.35kV≥9.5kV; all schemes passed the fluctuation superposition verification, forming a set of effective power flow optimization schemes.

[0204] Step 4054: Based on the set of power flow optimization schemes, combined with the target constraint boundary and the risk level quantification value of high-risk potential emergency scenarios, determine the safety margin value of each optimization scheme, and determine the optimization scheme with the highest safety margin value, which meets all constraints and is suitable for the current and potential emergency scenarios as the power flow optimization decision.

[0205] Optionally, the power flow optimization device extracts the adjusted parameters (such as G2 output, Z1 power, compensation capacity, and transferred load) for each of the obtained power flow optimization schemes. Then, combining the target constraint boundary with the risk level quantification value (Risk=0.83) of high-risk potential sudden scenarios, it uses a multi-dimensional margin comprehensive formula to determine the safety margin value of each optimization scheme. The formula is as follows: ; in, For the first The boundary values ​​of each constraint; For the first The adjusted values ​​of each parameter; The constraint weights are (generator power constraint 0.4, branch power constraint 0.3, bus voltage constraint 0.3). Quantification values ​​for high-risk scenarios This is for scenario-based risk compensation (the higher the risk, the stronger the compensation, highlighting the solution's resilience). After calculating the safety margin value for each solution, multiple conditions are used for screening, specifically including: the highest safety margin value (prioritizing resilience); meeting all activated constraints (constraints not broken after parameter adjustment); adapting to the current operating scenario (off-peak scenario, the solution adjustment strategy conforms to the off-peak equipment operation mode) and potential emergency scenarios (solution parameters still have sufficient margin to cope with faults in high-risk scenarios). After screening, the solution that simultaneously meets the above conditions is determined as the final power flow optimization decision. The decision must include specific adjustment parameter values, implementation steps (e.g., first adjust generator output, then reconstruct branch power flow, and finally configure reactive power compensation), and expected optimization effects (e.g., G1 margin increased to 7MW, B2 voltage stabilized at 10.5kV, Z1 power margin increased to 9.15MW).

[0206] Continuing with the above embodiments, based on the obtained set of solutions, objective constraints, and high-risk scenarios (Risk=0.83), the power flow optimization device evaluates and determines the final decision-making process as follows: First, calculate the safety margin (taking four scenarios as an example), and then assign constraint weights: (dynamo), (Branch road) (Voltage); Target constraint boundaries: G2≤73.72MW, Z1≤53.79MW, B2≤10.5kV; For Scheme 1 (G2=70, Z1=46, B2=10.5): Basic margin = 0.4×(73.72-70) / 73.72+0.3×(53.79-46) / 53.79+0.3×(10.5-10.5) / 10.5≈0.0202+0.0434+0=0.0636; Overall safety margin = 0.0636×1.83≈0.1164 For Scheme 2 (G2=69, Z1=45, B2=10.5): Basic margin = 0.4×(73.72-69) / 73.72 + 0.3×(53.79-45) / 53.79 + 0 ≈ 0.0256 + 0.0489 = 0.0745; Overall safety margin = 0.0745×1.83 ≈ 0.1363 For Scheme 3 (G2=68, Z1=47, B2=10.5): Basic margin = 0.4×(73.72-68) / 73.72 + 0.3×(53.79-47) / 53.79 + 0 ≈ 0.0310 + 0.0378 = 0.0688; Overall safety margin = 0.0688×1.83 ≈ 0.1259 For Scheme 4 (G2=72, Z1=42, B2=10.4): Basic margin = 0.4×(73.72-72) / 73.72+0.3×(53.79-42) / 53.79+0.3×(10.5-10.4) / 10.5≈0.0094+0.0659+0.0029=0.0782; Overall safety margin = 0.0782×1.83≈0.1431.

[0207] Further screening of options revealed that Option 4 still had the highest overall safety margin (0.1431), but its G2 output reached 72MW, only 1.72MW away from the upper limit, which could not meet the emergency backup requirement of at least 5MW in the event of a G1 failure. Therefore, it was still judged as "unsuitable for potential scenarios".

[0208] Option 2 has the second highest overall safety margin (0.1363), and with G2 output of 69MW and a reserve of 4.72MW, it can meet the N-1 safety criterion with the help of other measures. Therefore, the final decision is still to choose Option 2.

[0209] Therefore, after eliminating the unsuitable option 4, option 2, which has the second highest overall safety margin and meets the requirements of scenario adaptability, was determined as the final power flow optimization decision.

[0210] The final power flow optimization decision is: Generator output adjustment: G2 is adjusted from 65MW to 69MW, while G1 remains at 85MW (G2 increases its output to share the load, reserving adjustment space for G1). Branch power flow reconfiguration: Z1 is adjusted from 45MW to 45MW (remains unchanged), and Z2 is adjusted from 0MW to 10MW (total transmission power of B1-B2 is 55MW, Z1+Z2=55MW). Reactive power compensation configuration: A 6Mvar reactive power compensation device was installed on bus B2, increasing the voltage of B2 from 10.2kV to 10.5kV; Load transfer and allocation: 7MW of load in area B2 is transferred to area B1, the load in B2 decreases from 75MW to 68MW, and the load in B1 increases from 55MW to 62MW; Expected results: The output margin of G1 is increased to the safe range, the power margin of Z1 is 8.79MW, and the voltage of B2 is stabilized at 10.5kV. The overall safety margin of the system is increased by 16.8% under high-risk scenarios, meeting the dual goals of real-time scheduling and safe operation.

[0211] This application's embodiment employs a closed-loop power flow optimization decision-making process encompassing conflict localization, direction planning, scheme generation, and decision selection. This process provides three core advantages for the final optimization decision: First, precise conflict resolution by using multiple indicators to pinpoint the root cause of conflicts, ensuring the optimization direction directly addresses the core issue. Second, forward-looking risk assessment by incorporating the impact of high-risk potential emergencies, enhancing the scheme's resilience against sudden failures. Third, constraint compliance, ensuring all schemes meet activation constraints and preventing parameter boundary breaches. Ultimately, compared to traditional optimization methods, this process improves the safety margin of power flow optimization decisions, reduces the probability of constraint breaches in emergency scenarios, lowers generation costs, and reduces branch losses, thereby improving the safety, economy, and stability of power system operation.

[0212] In summary, through the specific implementation process of the above embodiments, this application, under a typical off-peak scenario in a western urban power system, ultimately obtains the corrected constraint boundary parameters and optimal power flow optimization decision, which are summarized as follows: I. The revised constraint boundary parameters (final safe operating boundary) are detailed in Table 4: Table 4. Corrected Constraint Boundary Parameters

[0213] II. Optimal Power Flow Optimization Decision (Highest Safety Margin and Satisfying All Constraints) Generator output adjustment: Increase the output of G2 from 65MW to 69MW, and keep G1 at 85MW (to reserve adjustment space for G1).

[0214] Branch power flow reconfiguration: The transmission power of Z1 remains unchanged at 45MW, while the backup branch Z2 is activated and its power is adjusted from 0MW to 10MW, thereby achieving a redistribution of the total transmission power of 55MW between B1 and B2.

[0215] Reactive power compensation configuration: A 6Mvar reactive power compensation device is installed on the B2 bus to increase the B2 voltage from 10.2kV to 10.5kV.

[0216] Load transfer and allocation: 7MW of load in area B2 will be transferred to area B1, reducing the load in B2 from 75MW to 68MW and increasing the load in B1 from 55MW to 62MW.

[0217] III. Expected Technical Effects Enhanced safety: The output margin of G1 remains within the safe range (currently 85MW, upper limit 86.43MW, margin 1.43MW), and the power margin of Z1 is 14.78MW (59.78 - 45). In high-risk scenarios (such as G1 failure shutdown), the N-1 safety margin of the system is significantly improved through the reserve capacity of G2 (4.72MW) and load transfer strategies.

[0218] Economic efficiency improvement: By dynamically adjusting the constraint boundary, the waste of regulation capacity caused by "overprotection" is avoided, and the power generation cost and branch losses are reduced.

[0219] Real-time performance guarantee: The embedded constraint neural network directly outputs the optimal decision without complex iterations, meeting the requirements of online scheduling.

[0220] The above results fully verify the effectiveness of the "scenario prediction-constraint correction-embedded neural network decision" closed-loop architecture proposed in this application, which can significantly reduce the probability of constraint failure, improve power supply reliability and reduce network loss in complex dynamic environments.

[0221] This application also provides an application scenario in which the above-mentioned power system power flow optimization method based on embedded constrained neural networks is applied. Specifically: The power flow optimization method provided in this embodiment can be applied to real-time dispatching and safe and stable control scenarios of the power grid. The power system enters the dispatching and control link from the data acquisition stage, and goes through processing stages such as digital twin construction, scenario matching and inference, constraint boundary correction, and embedded constraint neural network decision-making to obtain the optimal power flow optimization decision, which is then sent to the execution agencies (such as generator sets, reactive power compensation devices, load control terminals, etc.) for regulation.

[0222] The power flow optimization method based on an embedded constraint neural network provided in this embodiment belongs to the intelligent decision-making stage of power system dispatch automation. Specifically, during power system operation, the dispatch center collects real-time network operation status data (including generator output, load power, bus voltage, branch power flow, etc.) and physical topology to construct a digital twin of the power system. Based on this, the system automatically matches the current operating scenario and predicts potential emergencies (such as unplanned unit outages, line overloads, and drastic fluctuations in renewable energy output). Simultaneously, it dynamically corrects traditional fixed constraint boundaries using a historical fault database, making the constraints more closely match the actual carrying capacity of the equipment and future risks. Finally, the current state, scenario information, and corrected constraint boundaries are input into the embedded constraint neural network, which quickly outputs a power flow optimization decision with the highest safety margin, satisfying all constraints and adapting to current and potential emergencies. This decision can be directly used to guide operations such as generator output adjustment, branch power flow reconfiguration, reactive power compensation configuration, and load transfer allocation, thereby achieving safe, economical, and efficient operation of the power system in complex dynamic environments.

[0223] Based on the same inventive concept, this application also provides a power system power flow optimization device based on an embedded constraint neural network for implementing the above-described power system power flow optimization method based on an embedded constraint neural network. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more device embodiments provided below can be found in the limitations of the method described above, and will not be repeated here.

[0224] In one exemplary embodiment, this application also provides a power system power flow optimization device based on an embedded constraint neural network, applied to the aforementioned power system power flow optimization method based on an embedded constraint neural network; the power system power flow optimization device based on an embedded constraint neural network includes: The digital twin construction module is used to construct a digital twin of the power system based on the real-time collected power system operation status data and the physical topology of the power system within the target area. The scenario matching and simulation module is used to perform scenario matching based on the current operating status data collected at the current moment and the power system digital twin to obtain the current operating scenario, and to perform scenario simulation based on the current operating scenario and the power system digital twin to obtain multiple potential emergency scenarios. The constraint boundary correction module is used to correct the power flow constraint boundary based on the current operating status data and the historical fault database of the power system digital twin, so as to obtain the corrected constraint boundary parameters. The power flow optimization module is used to input the current operating status data, the current operating scenario, the potential sudden scenario, and the constraint boundary parameters into a trained embedded constraint neural network to perform power flow optimization and obtain power flow optimization decisions.

[0225] Through the coordinated operation of the above modules, this device achieves proactive perception of dynamic scenarios in the power system, accurate correction of constraint boundaries, and efficient and reliable power flow optimization decision-making.

[0226] In one exemplary embodiment, an electronic device is provided, which may be a server or a power system-specific computing device, and its internal structure diagram may be as follows. Figure 8As shown. The electronic device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores power system operating status data, historical fault data, scenario feature libraries, etc. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps of the power system power flow optimization method based on an embedded constrained neural network as described in any one of claims 1 to 7.

[0227] Those skilled in the art will understand that Figure 8 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 electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0228] In one exemplary embodiment, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0229] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0230] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0231] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), flash memory, magnetic disk, optical disk, etc. Volatile memory can include magnetic random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0232] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include key-value stores, document databases, time-series databases, etc., and are not limited thereto. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, etc., and are not limited thereto.

[0233] 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.

[0234] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A power system power flow optimization method based on an embedded constrained neural network, characterized in that, include: A digital twin of the power system is constructed based on real-time power system operation status data and the physical topology of the power system within the target area. Based on the current operating status data collected at the current moment and combined with the power system digital twin, scenario matching is performed to obtain the current operating scenario. Based on the current operating scenario and combined with the power system digital twin, scenario simulation is performed to obtain multiple potential emergency scenarios. Based on the current operating status data and the historical fault database of the power system digital twin, the power flow constraint boundary is corrected to obtain the corrected constraint boundary parameters. The current operating status data, the current operating scenario, the potential sudden scenario, and the constraint boundary parameters are input into the trained embedded constraint neural network for power flow optimization to obtain power flow optimization decisions.

2. The power system power flow optimization method based on embedded constrained neural networks according to claim 1, characterized in that, The embedded constraint neural network includes a scene mapping module, a risk level quantification module, and a constraint dynamic adaptation module. The step of inputting the current operating state data, the current operating scenario, the potential sudden scenario, and the constraint boundary parameters into a trained embedded constraint neural network for power flow optimization to obtain power flow optimization decisions specifically includes: In the scene mapping module, based on the generator output power, load consumption power, bus voltage amplitude and branch transmission power in the current operating status data, and combined with the topological connectivity and equipment operating mode characteristics corresponding to the current operating scene, a scene status association feature group is determined; In the risk level quantification module, based on the equipment failure type, failure impact range and load loss ratio corresponding to each potential emergency scenario, a scenario risk assessment factor is determined, and based on the scenario risk assessment factor and the preset risk level classification standard, the risk level corresponding to each potential emergency scenario is determined. In the constraint dynamic adaptation module, based on the constraint boundary parameters, combined with the topological features of the current operating scenario and the risk level of the potential sudden scenario, the initial value of the scenario adaptation constraint is determined, and based on the initial value of the scenario adaptation constraint, combined with the actual operating conditions of the equipment under the current operating scenario and the parameter fluctuations caused by the potential sudden scenario, the value range of each constraint parameter is adjusted to obtain the target constraint boundary. Based on the target constraint boundary and the mapping intensity index of the scene state association feature group, the constraint activation priority sequence is determined, and the running parameters corresponding to each constraint condition are checked and activated based on the constraint activation priority sequence to obtain the activated constraint condition group. Based on the constraint set, combined with the current operating state data, the scene state association feature set, and the target constraint boundary, power flow optimization is performed to obtain the power flow optimization decision.

3. The power system power flow optimization method based on embedded constrained neural networks according to claim 2, characterized in that, The embedded constraint neural network also includes a conflict localization module, an optimization direction planning module, and a scheme decision output module; The power flow optimization based on the constraint set, combined with the current operating state data, the scene state association feature set, and the target constraint boundary, to obtain the power flow optimization decision, specifically includes: In the conflict location module, conflict identification is performed based on the constraint condition group combined with the branch power flow direction, bus voltage deviation and generator output margin of the current operating status data to obtain power flow operation conflict characteristics. Based on the power flow operation conflict characteristics, the deviation status of each activated constraint condition and the corresponding operating parameter is compared one by one to determine the conflict location result. In the optimization direction planning module, based on the severity of the power flow operation conflict characteristics and the conflict location results, combined with the fault impact range of potential sudden scenarios with a high risk level, the priority of optimization needs is determined, and the solution path for each conflict problem is analyzed based on the priority of optimization needs to obtain the power flow optimization direction; the power flow optimization direction includes generator output adjustment, load transfer and distribution, branch power flow reconfiguration, and reactive power compensation configuration. In the scheme decision output module, based on the power flow optimization direction and the optimization requirement priority, the optimization parameter analysis is performed in combination with the constraint condition group and the scene state association feature group to obtain the optimization parameter adjustment range. Based on the optimization parameter adjustment range, the specific adjustment strategies for each power flow optimization direction are combined to obtain a power flow optimization scheme set. Based on the set of power flow optimization schemes, combined with the target constraint boundary and the risk level quantification value of high-level potential emergency scenarios, the safety margin value of each optimization scheme is determined, and the optimization scheme with the highest safety margin value, which satisfies all constraints and is suitable for the current and potential emergency scenarios is determined as the power flow optimization decision.

4. The power system power flow optimization method based on embedded constrained neural networks according to claim 1, characterized in that, The process of correcting the power flow constraint boundary based on the current operating status data and the historical fault database of the power system digital twin, to obtain the corrected constraint boundary parameters, specifically includes: Based on the historical fault database of the power system digital twin, historical fault records that are consistent with the power system type to which the current operating status data belong are selected to obtain a historical fault record group. Then, the fault association feature information related to the power flow constraint boundary in each historical fault record in the historical fault record group is extracted to obtain a historical fault association feature group. Based on the operating parameters that affect the power flow constraint boundary of the power system in the current operating status data, the current power flow influence factor group is obtained, and based on the current power flow influence factor group, the attribute information of each power flow influence factor is determined. Based on the historical fault association feature group and the current tidal current influence factor group, the correlation between each historical fault association feature and each current tidal current influence factor is analyzed, and a mapping relationship group between historical fault association features and current tidal current influence factors is constructed. Based on the historical fault association characteristics and current power flow influence factor corresponding to each mapping relationship in the mapping relationship group, the deviation amount caused to the power flow constraint boundary when the corresponding historical fault occurred is calculated to obtain the historical fault deviation amount. Based on the historical fault deviation amount and the current power flow influence factor, the influence degree value of the historical fault on the current power flow constraint boundary is calculated to obtain the historical fault influence degree group. Based on the historical fault impact level group, the attribute information of the power flow impact factor, and the power system digital twin, the deviation correction is performed to obtain the constraint boundary parameters.

5. The power system power flow optimization method based on embedded constrained neural networks according to claim 4, characterized in that, The constraint boundary parameters are obtained by performing deviation correction based on the historical fault impact level group, the attribute information of the power flow impact factor, and the power system digital twin, specifically including: Based on the digital twin of the power system, the theoretical standard range of the power flow constraint boundary under the current operating scenario is determined; Based on the current operating status data and the attribute information of the current power flow influencing factors, the difference between the current operating status and the theoretical standard status is analyzed to obtain the current operating status deviation. Based on the current operating status deviation and the theoretical standard range of the power flow constraint boundary, the initial deviation range of the current power flow constraint boundary is determined. Based on the historical fault impact group, the comprehensive impact value of all relevant historical faults on the current power flow constraint boundary is determined, and the initial deviation range is adjusted based on the comprehensive impact value to obtain the first power flow constraint boundary range after preliminary correction. Based on the scenario feature information that affects the power flow constraint boundary in potential sudden scenarios, the supplementary impact of each potential sudden scenario on the range of the first power flow constraint boundary is calculated to obtain the supplementary impact deviation value. Based on the supplementary impact deviation value, the range of the first power flow constraint boundary is adjusted again to obtain the second power flow constraint boundary range after secondary correction. Based on the second power flow constraint boundary range, a rationality verification is performed in conjunction with the safety requirements of power system operation and actual operating procedures, and the parameters corresponding to the second power flow constraint boundary range that passes the verification are determined as the constraint boundary parameters.

6. The power system power flow optimization method based on embedded constrained neural networks according to claim 1, characterized in that, The current operating scenario is obtained by combining the current operating status data collected at the current moment with the power system digital twin for scenario matching, specifically including: Based on the multi-dimensional features in the current operating status data and the power system operating scenario feature library constructed by the power system digital twin, similarity processing is performed to obtain the similarity values ​​between the multi-dimensional features and the features of each historical scenario. Candidate scene groups are obtained by filtering based on the similarity values ​​and preset similarity thresholds; Constraints are extracted based on the candidate scenario group and the power system digital twin to obtain the power operation constraints corresponding to each candidate scenario. The correlation between the power operation constraints of each candidate scenario and the current power system operation constraints is analyzed to obtain the constraint correlation value of each candidate scenario. Based on the constraint correlation value of each candidate scenario and the power system digital twin, trend analysis is performed to obtain the power flow change trend data of each candidate scenario in the corresponding operating period. Based on the power flow change trend data and the current power flow real-time change data, power flow consistency analysis is performed to obtain the consistency coefficient between the power flow trend of each candidate scenario and the current power flow trend. Based on the consistency coefficient and the constraint correlation value, the candidate scenarios are prioritized and sorted, and the candidate scenario with the highest priority is determined as the current running scenario.

7. The power system power flow optimization method based on embedded constrained neural networks according to claim 1, characterized in that, The scenario simulation based on the current operating scenario and the power system digital twin yields multiple potential emergency scenarios, specifically including: Based on the operating parameters of power system components in the current operating scenario and the physical associations of components stored in the power system digital twin, key operating constraints are determined, and a constraint association graph is constructed with each key operating constraint as a node and the influence intensity between different key operating constraints as edges. Based on the influence coefficients of each node in the constraint correlation graph and the operational stability data of each component in the current operating scenario, potential disturbance source types are obtained. Based on the potential disturbance source types and the triggering probability and triggering time interval characteristics of different disturbance sources in the power system operation law, a disturbance triggering sequence is determined. The disturbance triggering sequence includes the expected triggering time, disturbance intensity level, and impact range of each potential disturbance source. Based on the expected triggering time, disturbance intensity level and impact range of each disturbance source in the disturbance triggering sequence, the disturbance propagation process in the power system is simulated in combination with the dynamic characteristics of the components in the power system digital twin, and the disturbance propagation process is obtained. Based on the disturbance propagation process and the changes in the operating parameters of each component in the power system at different times, the dynamic response data of each component is obtained. Based on the analysis of the number of components exceeding the critical operational constraint deviation threshold in the dynamic response data of each component and the number of affected constraints in the constraint correlation map, several potential emergency scenarios are identified.

8. A power system power flow optimization device based on an embedded constrained neural network, characterized in that, Applied to the power system power flow optimization method based on embedded constrained neural networks as described in any one of claims 1 to 7; The power system flow optimization device based on embedded constrained neural networks includes: The digital twin construction module is used to construct a digital twin of the power system based on the real-time collected power system operation status data and the physical topology of the power system within the target area. The scenario matching and simulation module is used to perform scenario matching based on the current operating status data collected at the current moment and the power system digital twin to obtain the current operating scenario, and to perform scenario simulation based on the current operating scenario and the power system digital twin to obtain multiple potential emergency scenarios. The constraint boundary correction module is used to correct the power flow constraint boundary based on the current operating status data and the historical fault database of the power system digital twin, so as to obtain the corrected constraint boundary parameters. The power flow optimization module is used to input the current operating status data, the current operating scenario, the potential sudden scenario, and the constraint boundary parameters into a trained embedded constraint neural network to perform power flow optimization and obtain power flow optimization decisions.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the power system power flow optimization method based on an embedded constrained neural network as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the power system power flow optimization method based on the embedded constrained neural network as described in any one of claims 1-7.

Images