Power distribution network power flow analysis method, device, equipment, medium and product

Abstract

A power distribution network power flow analysis method, device, equipment, medium and product are disclosed. Historical operation data and topological structure data of each node in the power distribution network to be analyzed are obtained, the nodes including new energy access nodes and new energy non-access nodes; based on the historical operation data of each new energy access node, new energy output prediction is performed to obtain the output prediction result of each target time; for each node, based on the historical operation data and topological structure data of the current node, the time sequence dynamic characteristics and electrical space coupling characteristics of the node are determined; through attention gate mechanism, dynamic weighted fusion is performed to obtain the fusion characteristics of the node; based on the fusion characteristics of each node, a network-wide fusion feature matrix is determined; based on the output prediction result, the network-wide fusion feature matrix and the topological structure data, power flow analysis is solved to obtain the power flow analysis result of the power distribution network. The problem of insufficient feature utilization, poor accuracy and reliability of power flow analysis is solved.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, and in particular to a power flow analysis method, apparatus, equipment, medium, and product for power distribution networks. Background Technology

[0002] In recent years, with the rapid development of distributed photovoltaic, wind power and other renewable energy sources in rural power distribution networks, the operation mode of rural power distribution networks has gradually evolved from one-way power supply to two-way power flow.

[0003] Currently, rural power distribution networks mainly adopt a centralized power flow analysis method based on the unidirectional power flow assumption. This method involves constructing a node-branch model and performing power flow calculations based on the operating data of each node to assess the operating status of the power distribution network.

[0004] However, in bidirectional power flow operation scenarios, traditional unidirectional or static power flow models are unable to effectively capture the coupling relationship between electrical spatial characteristics and temporal dynamic characteristics, resulting in insufficient utilization of features, poor accuracy and reliability of power flow analysis, and difficulty in adapting to the needs of bidirectional power flow operation in rural areas. Summary of the Invention

[0005] This invention provides a power flow analysis method, apparatus, equipment, medium, and product for power distribution networks, in order to solve the problems of insufficient utilization of features and poor accuracy and reliability of power flow analysis in existing power flow analysis methods.

[0006] According to one aspect of the present invention, a power flow analysis method for a distribution network is provided, comprising: The historical operation data and topology data of each node in the distribution network to be analyzed are obtained, wherein the nodes include new energy access nodes and new energy non-access nodes; Based on the historical operating data of each new energy access node, the power output of new energy is predicted, and the power output prediction results of each new energy access node at the target time are obtained. For each node, based on the historical operation data and topology data of the current node, the temporal dynamic characteristics and electrical spatial coupling characteristics of the current node are determined; the temporal dynamic characteristics and electrical spatial coupling characteristics are dynamically weighted and fused through an attention gating mechanism to obtain the fused characteristics of the current node; Based on the fusion characteristics of each node, the network-wide fusion characteristic matrix is ​​determined; Based on the power output prediction results, the network-wide integrated feature matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network.

[0007] According to another aspect of the present invention, a power flow analysis device for a distribution network is provided, comprising: The data acquisition module is used to acquire historical operating data and topology data of each node in the distribution network to be analyzed, wherein the nodes include new energy access nodes and new energy non-access nodes; The power output prediction module is used to predict the power output of new energy sources based on the historical operating data of each new energy access node, and to obtain the power output prediction results of each new energy access node at the target time. The feature fusion module is used to determine the temporal dynamic features and electrical-spatial coupling features of each node based on its historical operation data and topology data; and to dynamically weight and fuse the temporal dynamic features and electrical-spatial coupling features through an attention gating mechanism to obtain the fused features of the current node. The matrix generation module is used to determine the network-wide fusion feature matrix based on the fusion characteristics of each node. The power flow solution module is used to perform power flow analysis and solution based on the power output prediction results, the network-wide integrated feature matrix, and the topology data, to obtain the power flow analysis results of the distribution network.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the power flow analysis method for the distribution network according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the power flow analysis method for a distribution network according to any embodiment of the present invention.

[0010] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the power flow analysis method for a distribution network according to any embodiment of the present invention.

[0011] The technical solution of this invention involves acquiring historical operating data and topology data of each node in the distribution network to be analyzed, wherein the nodes include renewable energy access nodes and renewable energy non-access nodes; based on the historical operating data of each renewable energy access node, renewable energy output prediction is performed to obtain the output prediction results of each renewable energy access node at the target time; for each node, based on the historical operating data and topology data of the current node, the temporal dynamic characteristics and electrical spatial coupling characteristics of the current node are determined; the temporal dynamic characteristics and electrical spatial coupling characteristics are dynamically weighted and fused through an attention gating mechanism to obtain the fused characteristics of the current node; based on the fused characteristics of each node, a network-wide fused characteristic matrix is ​​determined; based on the output prediction results, the network-wide fused characteristic matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network. By simultaneously extracting the temporal dynamic features and electrical spatial coupling features of each node, and using an attention gating mechanism for dynamic weighted fusion to obtain the whole-network fusion feature matrix, this approach overcomes the limitations of traditional unidirectional or static models that rely solely on static electrical parameters and cannot integrate temporal variation characteristics. It achieves deep coupling and adaptive utilization of electrical spatial features and temporal dynamic features, thereby solving the problem of insufficient feature utilization. On the other hand, by combining the new energy output prediction results, the whole-network fusion feature matrix, and topology data to jointly participate in power flow analysis, and relying on multi-source information and whole-domain fusion features to collaboratively support computational analysis, the input information dimensions of power flow solution are enriched, making the power flow analysis more comprehensive and its representation more closely aligned with the actual power grid operating state, thus solving the problem of poor accuracy and reliability in power flow analysis.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart of a power flow analysis method for a distribution network provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the structure of a power flow analysis device for a power distribution network provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] Example 1 Figure 1 This is a flowchart of a power flow analysis method for a distribution network provided in Embodiment 1 of the present invention. This embodiment is applicable to power flow analysis of rural distribution networks. The distribution network can include both bidirectional power flow operation (where power flows between sources and loads are bidirectional) and unidirectional power flow operation (where the output of new energy sources is zero or less than the load). This method can be executed by a power flow analysis device for the distribution network, which can be implemented in hardware and / or software and can be configured in electronic equipment. Figure 1 As shown, the method includes: S110. Obtain historical operating data and topology data of each node in the distribution network to be analyzed, wherein the nodes include new energy access nodes and new energy non-access nodes.

[0018] In this embodiment, the distribution network to be analyzed is a distribution network requiring power flow analysis, consisting of several electrical nodes and connecting branches. The distribution network to be analyzed in this embodiment is a rural distribution network that includes distributed photovoltaic, wind power, and other renewable energy sources. This distribution network can dynamically switch its operating state according to the output of renewable energy and the load level, supporting both bidirectional power flow conditions with both source and load bidirectional power flow and unidirectional power receiving conditions when renewable energy output is insufficient. Nodes include renewable energy access nodes and renewable energy non-access nodes. Renewable energy access nodes are nodes connected to renewable energy sources. Historical operating data can be understood as the operating status data collected by each node within a historical period, which may include electrical measurement data, load data, and environmental data. Topology data is used to describe the connection relationships between nodes and branches in the distribution network to be analyzed, reflecting the spatial structure and electrical topology of the power grid.

[0019] S120. Based on the historical operating data of each new energy access node, perform new energy output prediction to obtain the output prediction results of each new energy access node at the target time.

[0020] In this embodiment, renewable energy output prediction is a process of estimating the renewable energy generation power at future times based on historical operating data of renewable energy access nodes. The target time is the future time at which power flow analysis needs to be performed.

[0021] Specifically, based on the historical operating data of each renewable energy access node, the renewable energy power generation of each renewable energy access node at a future target time is estimated using a renewable energy output prediction model to obtain the corresponding output prediction results. The renewable energy output prediction model can be a power prediction model based on machine learning or deep learning, or a data-driven prediction model that integrates physical constraints and historical data; this embodiment does not impose specific limitations on it.

[0022] S130. For each node, based on the historical operation data and topology data of the current node, determine the temporal dynamic characteristics and electrical spatial coupling characteristics of the current node; use an attention gating mechanism to dynamically weight and fuse the temporal dynamic characteristics and the electrical spatial coupling characteristics to obtain the fused characteristics of the current node.

[0023] In this embodiment, temporal dynamic features are used to characterize the regularity and trend of node operating status changes over time, reflecting the temporal fluctuation characteristics of parameters such as renewable energy output and load. Electrical spatial coupling features are used to characterize the electrical connections between nodes and adjacent nodes and branches, reflecting the spatial electrical coupling characteristics under the topological constraints of the distribution network. The attention gating mechanism is a mechanism that can automatically allocate weights according to data importance and achieve adaptive fusion of multiple features, dynamically adjusting the fusion ratio of temporal dynamic features and electrical spatial coupling features.

[0024] Specifically, in order to simultaneously utilize the temporal dynamic information and electrical-spatial information of node operation, for each node, its temporal dynamic characteristics are mined based on the historical operation data of the current node, and its electrical-spatial coupling characteristics are extracted based on the topology data. Furthermore, an attention gating mechanism is used to dynamically weight and fuse the two types of features. The fusion process is as follows: ; Where, in the formula, Let be the fused feature vector of node i. Let i be the temporal dynamic feature of node i. Let i be the electrical spatial coupling characteristic of node i. This is used for attention weighting. Through this mechanism, the contribution ratio of temporal features and electrical features is automatically adjusted under different operating scenarios, so that the fused features can take into account both the dynamic laws of time and the spatial electrical coupling relationship, ensuring that the features remain reliable under measurement fluctuations or power output changes, and providing a reliable node-level comprehensive characterization for subsequent power flow analysis.

[0025] S140. Based on the fusion characteristics of each node, determine the network-wide fusion characteristic matrix.

[0026] In this embodiment, the whole-network integrated feature matrix is ​​used to characterize the operating status and electrical coupling relationship of the entire distribution network to be analyzed.

[0027] Specifically, based on the fusion characteristics of each node, they can be directly combined and spliced ​​according to the node dimension to form a network-wide fusion feature matrix covering the entire network; or, based on the credibility of the data of each node (such as measurement integrity, data stability, etc.), the fusion characteristics of each node can be adaptively weighted and adjusted before participating in the matrix construction to improve the reliability of the network-wide fusion feature matrix.

[0028] S150. Based on the output prediction results, the network-wide integrated feature matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network.

[0029] In this embodiment, the power flow analysis results include the operating status results of the distribution network to be analyzed at the target time. The operating status results may include the voltage amplitude and phase angle of each node, the active power and reactive power of each branch, etc.

[0030] Specifically, the output prediction results are used as the power input conditions for the new energy side, and the whole network integrated feature matrix is ​​used as a high-dimensional feature input reflecting the time-series dynamics and electrical spatial coupling characteristics of the power grid. Combined with the network connection constraints provided by the topology data, power flow calculation is performed to obtain the operating status results of the distribution network at the target time, which are used as the power flow analysis results.

[0031] It is worth noting that the power flow analysis results can be used to assess the renewable energy absorption capacity of rural distribution networks, verify safe operation, and make optimized dispatch decisions, providing technical support for the safe operation and optimized control of distribution networks under conditions of high renewable energy penetration. For example, based on the power flow analysis results, the operating status of the distribution network to be analyzed can be evaluated, and corresponding dispatch control commands can be generated.

[0032] The technical solution provided in Embodiment 1 of this invention involves acquiring historical operating data and topology data of each node in the distribution network to be analyzed, wherein the nodes include new energy access nodes and new energy non-access nodes; based on the historical operating data of each new energy access node, new energy output prediction is performed to obtain the output prediction results of each new energy access node at the target time; for each node, based on the historical operating data and topology data of the current node, the temporal dynamic characteristics and electrical spatial coupling characteristics of the current node are determined; the temporal dynamic characteristics and electrical spatial coupling characteristics are dynamically weighted and fused through an attention gating mechanism to obtain the fused characteristics of the current node; based on the fused characteristics of each node, a network-wide fused characteristic matrix is ​​determined; based on the output prediction results, the network-wide fused characteristic matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network. By simultaneously extracting the temporal dynamic features and electrical spatial coupling features of each node, and using an attention gating mechanism for dynamic weighted fusion to obtain the whole-network fusion feature matrix, this approach overcomes the limitations of traditional unidirectional or static models that rely solely on static electrical parameters and cannot integrate temporal variation characteristics. It achieves deep coupling and adaptive utilization of electrical spatial features and temporal dynamic features, thereby solving the problem of insufficient feature utilization. On the other hand, by combining the new energy output prediction results, the whole-network fusion feature matrix, and topology data to jointly participate in power flow analysis, and relying on multi-source information and whole-domain fusion features to collaboratively support computational analysis, the input information dimensions of power flow solution are enriched, making the power flow analysis more comprehensive and its representation more closely aligned with the actual power grid operating state, thus solving the problem of poor accuracy and reliability in power flow analysis.

[0033] In some embodiments, the historical operating data includes a sequence of historical operating features, each of which includes historical renewable energy output power, active power of node loads, reactive power of node loads, node voltage data, environmental meteorological characteristics, and measurement availability indicators for the corresponding node.

[0034] The historical operational feature sequence is a multi-dimensional feature set formed by arranging each node in chronological order within a historical time period. The historical time period covers continuous sampling moments from the past to the present. Historical operational features are used to characterize the specific operational state of the corresponding node at a given sampling moment. For example, the historical operational feature sequence can be represented as: ; Where n is the length of the historical input window, For each element in the sequence at the current time. These are the aforementioned historical operational characteristics.

[0035] Historical renewable energy output power is the total renewable energy output of the renewable energy access node at the sampling time, obtained by aggregating the output power of all renewable energy power plants connected to that node. For example, the historical renewable energy output power of the current time tk can be expressed as: ; In the formula, For the set of new energy power stations connected to node i, For station s at the sampling time The output power, The active power output of renewable energy after aggregation at node i; for nodes not connected to renewable energy power plants, their historical renewable energy output is zero.

[0036] The active power of a node load refers to the active power consumed by the load at the distribution network node. The reactive power of a node load refers to the reactive power consumed by the load at the distribution network node. Node voltage data refers to electrical measurement data such as voltage amplitude and phase angle at the distribution network node. Environmental meteorological characteristics refer to meteorological parameters of the environment where the new energy power station is located, including light intensity, temperature, wind speed, and wind direction. The measurement availability indicator is a status indicator used to identify whether the measurement data of the corresponding node at the sampling time is complete, valid, without missing data, and without abnormalities.

[0037] For example, at the current moment The historical operational characteristics are composed of the above components in the following vector form: ; In the formula, For the active power of the node load, For the reactive power of the node load, For node voltage data, As environmental and meteorological characteristics, This is a metric for availability.

[0038] In some embodiments, determining the temporal dynamic characteristics and electrical spatial coupling characteristics of the current node based on its historical operating data and topology data includes: processing the historical operating characteristic sequence of the current node using a temporal feature extraction network to obtain the temporal dynamic characteristics of the current node; and processing the node voltage data of the current node and the adjacency relationships in the topology data using an electrical topology feature encoding network, aggregating the node voltage data of neighboring nodes to obtain the electrical spatial coupling characteristics of the current node.

[0039] In this embodiment, the temporal feature extraction network is a computational network used to process the historical operational feature sequences of nodes and extract their temporal dynamic features, enabling it to learn the dynamic patterns and trends in time series data. The electrical topology feature encoding network is a computational network used to encode spatial features based on adjacency relationships and node electrical states in the topology data, capable of aggregating the electrical information of neighboring nodes and extracting the electrical spatial coupling features between nodes. Adjacency relationships are data in the topology data that characterize the connection relationships between nodes and branches, reflecting whether there is a direct electrical connection between nodes.

[0040] Specifically, for each node, a temporal feature extraction network is used to process the historical execution feature sequence of the current node to obtain the temporal dynamic features of the current node. In this embodiment, the temporal feature extraction network adopts a sequence modeling structure based on a self-attention mechanism to learn the temporal dependencies in the node's historical execution feature sequence. The processing procedure is as follows: ; in, The temporal dynamic characteristics of node i; For node i up to the current time The network represents the historical operational characteristic sequence, where n is the length of the historical input window. Through this network, the patterns and trends of parameters such as renewable energy output and load over time are extracted.

[0041] Simultaneously, an electrical topology feature coding network is used to process the adjacency relationships in the node voltage data and topology data of the current node, aggregating the node voltage data of neighboring nodes to obtain the electrical spatial coupling characteristics of the current node. In this embodiment, the electrical topology feature coding network performs spatial feature coding on the node voltage topology adjacency relationships based on a graph structure message passing mechanism. The processing procedure is as follows: ; in, The electrical spatial coupling characteristics of node i; Let i be the set of neighboring nodes of node i; These are trainable weights; For activation functions; For neighbor node j at the current time Node voltage data, , This provides auxiliary electrical parameters such as active power and reactive power. The network extracts spatial coupling information between nodes and branches by aggregating node voltage data from neighboring nodes, providing electrical constraint information for power flow analysis.

[0042] This embodiment realizes the parallel extraction of temporal dynamic features and electrical-spatial coupling features, overcoming the shortcomings of traditional static models that are difficult to integrate temporal change characteristics and ignore network topology constraints, and providing high-quality feature inputs that take into account both temporal evolution and spatial coupling for subsequent power flow analysis.

[0043] In some embodiments, determining the network-wide fusion feature matrix based on the fusion characteristics of each node includes: determining the node weight of each node based on the measurement availability flag of each node; determining the modified fusion characteristics of each node based on the fusion characteristics of each node and the corresponding node weight; and determining the network-wide fusion feature matrix based on the modified fusion characteristics of each node.

[0044] In this embodiment, node weights are used to characterize the importance of each node's fusion feature in constructing the overall network fusion feature matrix, and can be adaptively determined based on the reliability of the node data. The corrected fusion feature is a node-level feature vector obtained by weighting the fusion feature based on the node weights, used to reduce the impact of low-reliability nodes on the overall network fusion feature matrix.

[0045] Specifically, considering the possibility of incomplete measurement data in rural power distribution networks, with missing or abnormal measurement data at each node leading to differences in the reliability of fused features from different nodes, in order to reduce the impact of low-quality nodes and improve matrix reliability, the node weights of each node are determined based on the measurement availability flag of each node at the current moment. In one implementation, a soft-threshold attention mechanism is introduced to adaptively weight key nodes. The calculation process for node weights is as follows: ; in, Let node i be at the current time. The node weights (normalized weights). Let node i be at the current time. Measuring availability indicator To adjust the parameters.

[0046] Furthermore, based on the fusion characteristics of each node and the corresponding node weights, the modified fusion characteristics of each node are determined: ; in, The modified fusion feature for node i. Let i be the fusion feature of node i.

[0047] Finally, based on the corrected fusion features of each node, the features are concatenated according to the node dimension to determine the network-wide fusion feature matrix: ; Wherein, H is the network-wide fusion feature matrix, which is used for initial value generation, sensitivity correction, and constraint update of the subsequent power flow model.

[0048] This mechanism enhances the contribution of valid measurement nodes to feature modeling and reduces the impact of missing or abnormal measurement nodes on power flow analysis.

[0049] In some embodiments, the power output prediction result includes the power output prediction interval of the corresponding renewable energy access node at the target time; the step of predicting renewable energy power output based on the historical operating data of each renewable energy access node to obtain the power output prediction result of each renewable energy access node at the target time includes: for each renewable energy access node, inputting the historical operating feature sequence of the current renewable energy access node into the trained renewable energy power output prediction model, and outputting the active power output of the current renewable energy access node at the target time; wherein, the renewable energy power output prediction model uses historical operating feature samples of multiple benchmark renewable energy access nodes as a training set, and is trained using a model-independent meta-learning framework, and the loss function used in the training process includes a first constraint loss term constructed based on the physical laws of photovoltaic power output and / or a second constraint loss term constructed based on the physical laws of wind power output; based on the active power output, using a generative diffusion model, generating the active power output of the current renewable energy access node in each scenario at the target time, and determining the power output prediction interval of the current renewable energy access node at the target time based on the active power output in each scenario.

[0050] In this embodiment, the power output prediction interval is the predicted range of active power output of the renewable energy access node at the target time, determined by the lower and upper limits of power in the multi-scenario prediction results, used to characterize the uncertainty boundary of renewable energy output. The renewable energy output prediction model is a power prediction model trained based on historical operating feature samples, used to output the active power output of the renewable energy access node at the target time. The Model Independent Meta-Learning Framework (MAML) is a meta-learning training framework that uses historical operating feature samples of multiple benchmark renewable energy access nodes as a training set for training, learning transferable model initialization parameters, enabling the model to quickly adapt to the characteristics of new power stations. The physical information constraint loss term is a loss function term that incorporates physical laws as hard constraints into the model training process; wherein, the first constraint loss term refers to the physical information constraint loss term constructed based on the physical laws of photovoltaic power output (such as the solar radiation transfer equation), and the second constraint loss term refers to the physical information constraint loss term constructed based on the physical laws of wind power output (such as the wind turbine dynamics model). The generative diffusion model is used to generate the active power output of each scenario at the target time based on the active power output, so as to determine the power output prediction interval.

[0051] Specifically, considering the dispersed nature of new energy power stations in rural power distribution networks, the low quality and high volatility of data, and the scarcity of historical data for new power stations, which makes it difficult to directly apply traditional prediction models and result in insufficient generalization ability, this embodiment uses the Model Independent Meta-Learning (MAML) framework to train the new energy output prediction model. It uses historical operating feature samples from multiple benchmark new energy access nodes as the training set to learn transferable new energy output change patterns, thereby improving the model's ability to quickly adapt to power stations with scarce data.

[0052] During training, to ensure the prediction results conform to physical laws, the loss function used includes a first constraint loss term constructed based on the physical laws of photovoltaic power output and a second constraint loss term constructed based on the physical laws of wind power output. In one implementation, the above constraint loss terms can be implemented by incorporating the corresponding physical equations as hard constraints into the loss function through a Physical Information Neural Network (PINN). The photovoltaic power output physical constraint corresponding to the first constraint loss term is: ; In the formula, For photovoltaic module conversion efficiency, For component area, Let be the solar irradiance at the target time d. This is the error correction term.

[0053] The physical constraint on wind power output corresponding to the second constraint loss term is: ; In the formula, ρ is the air density. The swept area of ​​the fan impeller. It is a function of the wind turbine power coefficient as a function of wind speed. Let be the wind speed at the target time d. This is a correction term. By incorporating the aforementioned physical constraints into the loss function, the prediction results are made to consider both historical data and the physical mechanism of new energy power generation.

[0054] Through the above training, a new energy output prediction model is obtained.

[0055] Furthermore, for each renewable energy access node, the historical operating characteristic sequence of the current renewable energy access node is input into the trained renewable energy output prediction model, and the active power output of the node at the target time d is output: ; In the formula, Let be the active power output of node i at the target time d. A new energy power output prediction model with parameter θ.

[0056] Furthermore, based on the aforementioned active power output, a generative diffusion model is used to generate the active power output of the current renewable energy access node under various scenarios at the target time d: ; In the formula, r is the scene number, and R is the total number of scenes. Let be the active power output of node i in the r-th scenario at target time d.

[0057] Finally, based on the active power output under each scenario, the predicted output range of the current renewable energy access node at the target time d is determined: ; In the formula, and These represent the lower and upper limits of the predicted power, respectively, used to denote the uncertainty boundary. This power output prediction range can be directly used as the injection power constraint for renewable energy access nodes in subsequent power flow calculations, ensuring the robustness of power flow analysis under different power output conditions.

[0058] By integrating meta-learning with physical constraint training, this embodiment improves the power output prediction accuracy and station adaptability under data-scarce conditions; by constructing a power output prediction interval through a diffusion model to represent uncertainty, it provides stable power supply constraints for power flow analysis and improves the robustness of the solution.

[0059] In some embodiments, the step of performing power flow analysis based on the output prediction results, the network-wide integrated feature matrix, and the topology data to obtain the power flow analysis results of the distribution network includes: constructing node power balance equations based on the topology data; determining the net injected power of each node at the target time and the constraint boundary of the net injected power according to the node load power of each node at the target time and the output prediction interval, and embedding the constraint boundary of the net injected power into the node power balance equations; generating initial values ​​for power flow solution using the network-wide integrated feature matrix, and solving the node power balance equations embedded with the constraint boundary in combination with node voltage safety constraints and branch capacity constraints to obtain the power flow analysis results.

[0060] In this embodiment, the node power balance equation describes the balance between active and reactive power at each node of the distribution network and serves as the fundamental constraint equation for power flow analysis. Node load power is the power consumed by the load at each node in the distribution network, including both active and reactive power. Net injected power is the difference between the power output of renewable energy sources at a node and the load power at that node, reflecting the net power injected into or absorbed from the distribution network by that node. For nodes without renewable energy connections, the renewable energy output is zero, and the net injected power is determined by the load power. The constraint boundary of net injected power is the allowable variation range of net injected power determined based on the output prediction interval. This range characterizes the constraint and limitation of renewable energy output uncertainty on node power injection, ensuring the feasibility of the power flow model under conditions of renewable energy output fluctuations. The initial values ​​for power flow solution are the initial estimates of node voltage amplitude and phase angle generated using the network-wide fusion feature matrix to initiate the power flow calculation iteration process. Node voltage safety constraints are the allowable range of node voltage amplitude set to ensure the safe operation of the distribution network. Branch capacity constraints are the upper limit of branch transmission power set to ensure the safe operation of the distribution network.

[0061] Specifically, based on topology data, node power balance equations are constructed, including active power balance equations and reactive power balance equations: ; In the formula, These are the net active power injection and net reactive power injection of node i at the target time d, respectively. Let be the voltage amplitude of node i and node j at the target time d. Let be the phase angle difference between node i and node j; These are the conductance and susceptance components of the nodal admittance matrix, respectively.

[0062] Furthermore, based on the node load power of each node at the target time and the power output prediction interval, the net injected power of each node at the target time is determined. The net active power injection power and the net reactive power injection power are respectively: ; In the formula, These are the active power and reactive power output from new energy sources, respectively. These represent the active and reactive power of the node load, respectively. For nodes not connected to renewable energy sources, the renewable energy output is zero; for nodes with no load, the node load power is zero.

[0063] The constraint boundary for net active power injection is determined by the output prediction interval. The output prediction interval is embedded within the node active power injection condition: ; Furthermore, the net active power injection constraint is obtained: ; in, ; This enables the power flow model to maintain the feasibility of its solution under conditions of fluctuating renewable energy output. Furthermore, the constraint boundary of the net active power injection is embedded into the active power balance equation.

[0064] Then, the initial values ​​for power flow solution are generated using the network-wide fused feature matrix: ; In the formula, Let be the initial voltage value of the node at the target time d. The initial value of the node phase angle. Here, H is the feature mapping function, and H is the network-wide fusion feature matrix.

[0065] Finally, combining the node voltage safety constraints and branch capacity constraints, the node power balance equations embedded in the constraint boundaries are solved to obtain the power flow analysis results: ; In the formula, These are the upper and lower limits of the node voltage. Let (i,j) be the complex power of branch (i,j) at the target time d. This represents the upper limit of branch capacity. The final output is the feasible power flow solution at the target time, which serves as the power flow analysis result.

[0066] By embedding the power output prediction interval into the nodal power balance equation, the power flow model is made adaptable to the power output fluctuations of new energy sources; the initial value is generated by using the fused feature matrix to reduce the sensitivity of the initial value and improve the convergence robustness; and the power flow solution is ensured to meet the safe operation requirements by combining voltage safety and branch capacity constraints, thereby improving the accuracy and reliability of the analysis.

[0067] In some embodiments, power flow analysis is performed based on the output prediction results, the network-wide integrated feature matrix, and the topology data to obtain the power flow analysis results of the distribution network. This includes: adaptively dividing the distribution network to be analyzed into multiple autonomous regions based on the output prediction results and the topology data; determining the local topology data, local output prediction results, and local integrated feature matrix of each autonomous region based on the topology data, the output prediction results, and the network-wide integrated feature matrix; establishing and solving local power flow equations for each autonomous region based on its corresponding local topology data, local output prediction results, and local integrated feature matrix to obtain local power flow solutions; using the voltage and phase angle of the corresponding boundary nodes in the local power flow solutions as boundary variables, and coordinating the consistency of the boundary variables with adjacent autonomous regions using the alternating direction multiplier method, while combining a longitudinal federated learning mechanism for encrypted gradient exchange; dynamically adjusting the penalty parameters of the alternating direction multiplier method based on the consistency residuals of the boundary variables between adjacent autonomous regions during the iterative coordination process; and outputting the network-wide node voltage, phase angle, and branch power results after global coordination as the power flow analysis results.

[0068] In this embodiment, an autonomous region is a set of local computational regions adaptively divided from the distribution network to be analyzed. Each region can independently solve for power flow. Local topology data refers to the topology data within the autonomous region, describing the connection relationships between nodes and branches within that region. Local output prediction results are the output prediction results of renewable energy access nodes within the autonomous region. The local fusion feature matrix is ​​a matrix formed by combining the fusion features of each node within the autonomous region according to the node dimension. The local power flow equations are power flow equations established based on local topology data, local output prediction results, and the local fusion feature matrix, used for independent solution by each region. Boundary nodes are nodes located at the boundary of the autonomous region that are electrically connected to adjacent regions. Boundary variables are the voltage and phase angle of the boundary nodes in the local power flow solution, serving as common variables for cross-regional consistency coordination. The alternating direction multiplier method is a distributed optimization algorithm used to achieve consistency of boundary variables between adjacent autonomous regions through iterative coordination. The longitudinal federated learning mechanism is a privacy-preserving mechanism where each region achieves collaborative optimization through encrypted gradients or parameter exchange, without exposing the original operating data. The consistency residual is the difference in calculation of the same boundary variable between adjacent autonomous regions, used to measure the degree of boundary consistency. The penalty parameter is a coordination parameter used in the alternating direction multiplier method to control the strength of boundary constraints.

[0069] Specifically, considering the continuously expanding scale of rural power distribution networks, the dispersed data, and the low efficiency of centralized power flow solutions due to privacy restrictions, the power distribution network to be analyzed is adaptively divided into multiple autonomous partitions based on output prediction results and topology data. Using the DBSCAN algorithm, nodes are divided into several autonomous units based on electrical distance and output correlation between nodes: ; In the formula, N is the entire set of network nodes. Let Z be the set of nodes in the z-th autonomous partition, where Z is the total number of partitions. This partitioning method groups nodes with high output relevance into the same partition, thereby reducing the complexity of cross-regional coordination.

[0070] Furthermore, based on topology data, power output prediction results, and the network-wide fusion feature matrix, the local topology data, local power output prediction results, and local fusion feature matrix of each autonomous region are determined.

[0071] By establishing and solving local power flow equations for each autonomous region based on its corresponding local topology data, local power output prediction results, and local fused feature matrix, the local power flow solutions are obtained. The locally established local power flow equations for each region are as follows: ; In the formula, Let be the power flow equations for the z-th partition. This represents the node voltage and phase angle vector for this partition.

[0072] The voltage and phase angle of the corresponding boundary nodes in the local power flow solution are used as boundary variables. Consistency coordination of boundary variables with adjacent autonomous regions is achieved through the alternating direction multiplier method. The alternating direction multiplier method is introduced, and Lagrange multipliers are defined to constrain the boundary variables of the region and its adjacent regions. ; In the formula, Let ρ be the Lagrange multiplier for partition z, and ρ be the penalty parameter. The variables are used for coordination at the partition boundaries. Through iterative methods using the alternating direction multiplier method, the local power flow solutions of each partition are gradually coordinated, ultimately yielding a consistent power flow solution for the entire network.

[0073] It also incorporates a vertical federated learning mechanism for encrypted gradient exchange. Each partition achieves collaborative optimization through encrypted gradient exchange without uploading raw power, load, or voltage data. ; In the formula, Let z be the model parameters for the z-th partition in the r-th iteration. Let ξ be the local power flow loss function, and ξ be the learning rate. These are global aggregation parameters. This approach achieves both distributed collaborative solving and data privacy protection.

[0074] During the iterative coordination process of the alternating direction multiplier method, the penalty parameter of the alternating direction multiplier method is dynamically adjusted based on the consistency residuals of boundary variables between adjacent autonomous partitions. A dynamic step size adjustment strategy is adopted to automatically adjust the penalty parameter according to the changes in local residuals of the partitions. ; in, For partition residuals, , These represent increasing or decreasing the step size coefficient, respectively. This is the residual threshold. Dynamic adjustment can accelerate convergence while maintaining accuracy. Finally, the system outputs the network node voltage, phase angle, and branch power results after global coordination as the power flow analysis results; at the same time, it outputs the node voltage and branch power range results formed under multiple scenarios, which are used for the analysis of the new energy absorption capacity of rural distribution networks, the verification of safe operation, and the optimization scheduling decision.

[0075] By implementing local parallel solution through adaptive partitioning, and combining the alternating direction multiplier method with the vertical federated learning mechanism, the efficiency and scalability of power flow solution for large-scale distribution networks are improved while protecting data privacy, and convergence is accelerated by dynamically adjusting the penalty parameters.

[0076] In some embodiments, acquiring historical operating data and topology data of each node in the distribution network to be analyzed includes: acquiring topology information of the distribution network to be analyzed, and multi-source operating data of each node at multiple sampling times; constructing the topology connection relationship of the distribution network to be analyzed based on the topology information, and performing consistency verification on the topology connection relationship; preprocessing the multi-source operating data of each node at the multiple sampling times, and determining the measurement availability flag of each node to obtain the preprocessed operating data of each node; using the verified topology connection relationship as the topology data, and standardizing the preprocessed operating data of each node to form the historical operating data of each node.

[0077] In this embodiment, topology information refers to the raw data describing the spatial distribution and connection attributes of nodes and branches in the distribution network to be analyzed. Multi-source operational data refers to the operational status data collected from various data sources at multiple sampling times for each node, which may include electrical measurement data, load data, etc. Topology connection relationships are data constructed based on topology information, characterizing the connection relationships between nodes and branches in the distribution network, reflecting the spatial structure of the power grid. Consistency verification involves verifying the topology connection relationships to detect and eliminate topology inconsistencies, forming a computable topology basis. Preprocessing involves performing missing data identification, noise suppression, and RMS value reconstruction on the multi-source operational data to improve data quality. Measurement availability flags are status indicators used to identify whether the measurement data of the corresponding node at the sampling time is complete, valid, without missing data, and without anomalies. Standardization processing involves standardizing the preprocessed operational data according to a unified dimension and format, forming historical operational data that can be directly used for subsequent analysis.

[0078] Specifically, the topology information of the distribution network to be analyzed is obtained through geographic information systems and ledger data, and multi-source operational data is collected through a data acquisition and monitoring control system. To ensure data availability, the multi-source operational data of each node at multiple sampling times are time-aligned, and the units of various physical quantities are standardized (e.g., voltage is standardized to kV, current to A, and power to kW / kVar).

[0079] Based on topology information, the topological connection relationships of the distribution network to be analyzed are constructed, such as: ; In the formula, This indicates whether there is a branch connection between node i and node j. Furthermore, it detects isolated nodes and closed-loop anomalies: if the total number of connections for a node is zero, it is determined to be an isolated node, and an attempt is made to establish a connection with its nearest neighbor node; for detected closed-loop anomalies, a simulated switch operation is used to disconnect the connection, ensuring that the network meets the requirements for distribution network operation and forming a computable, conflict-free topology.

[0080] The multi-source operational data of each node at multiple sampling times are preprocessed, and the measurement availability flag of each node is determined to obtain the preprocessed operational data of each node. In one implementation, a node measurement availability flag is defined to address the case of sparse or missing measurements: ; When the measurement availability flag is 1, the original valid measurement value is retained; when the measurement availability flag is 0, the neighbor weighted average is used for estimation.

[0081] For example, missing data filling for voltage measurements: ; In the formula, For node i at time... The estimated voltage, Let i be the set of neighboring nodes. The weights between nodes can be set based on admittance or geographical distance. To further suppress noise, a sliding window low-pass filter is applied to the estimation results: ; In the formula, p represents the window radius, and s is the relative time offset within the sliding window (from -p to p). For other electrical measurement data such as node load power and branch power, similar preprocessing is performed based on the measurement availability flag of the corresponding node, in accordance with the above method, to ensure the overall quality of multi-source operational data. The above preprocessing operation is performed at each sampling time, thereby obtaining the preprocessed operational data of each node at multiple consecutive sampling times.

[0082] The verified topological connection relationships are used as topological structure data, and the preprocessed running data of each node is standardized to form the historical running data of each node.

[0083] Example 2 Figure 2 This is a schematic diagram of the power flow analysis device for a distribution network provided in Embodiment 2 of the present invention. Figure 2 As shown, the device includes: The data acquisition module 21 is used to acquire historical operating data and topology data of each node in the distribution network to be analyzed, wherein the nodes include new energy access nodes and new energy non-access nodes; The power output prediction module 22 is used to predict the power output of new energy sources based on the historical operating data of each new energy access node, and to obtain the power output prediction results of each new energy access node at the target time. The feature fusion module 23 is used to determine the temporal dynamic features and electrical spatial coupling features of each node based on the historical operation data and topology data of the current node; and to dynamically weight and fuse the temporal dynamic features and the electrical spatial coupling features through an attention gating mechanism to obtain the fused features of the current node. The matrix generation module 24 is used to determine the network-wide fusion feature matrix based on the fusion features of each node. The power flow solution module 25 is used to perform power flow analysis and solution based on the power output prediction results, the network-wide integrated feature matrix and the topology data, and to obtain the power flow analysis results of the distribution network.

[0084] The technical solution provided in Embodiment 2 of this invention extracts the temporal dynamic features and electrical spatial coupling features of each node simultaneously, and uses an attention gating mechanism to perform dynamic weighted fusion to obtain a network-wide fusion feature matrix. This overcomes the limitations of traditional unidirectional or static models that rely solely on static electrical parameters and cannot integrate temporal variation characteristics. It achieves deep coupling and adaptive utilization of electrical spatial features and temporal dynamic features, thereby solving the problem of insufficient feature utilization. On the other hand, it combines the new energy output prediction results, the network-wide fusion feature matrix, and topology data to jointly participate in power flow analysis. Relying on multi-source information and the collaborative support of the network-wide fusion features for computational analysis, it enriches the input information dimensions of power flow solution, making the power flow analysis more comprehensive and the representation more closely aligned with the actual power grid operating state, thus solving the problem of poor accuracy and reliability of power flow analysis.

[0085] Optionally, the historical operating data includes a historical operating feature sequence, and each historical operating feature in the historical operating feature sequence includes historical renewable energy output power, node load active power, node load reactive power, node voltage data, environmental meteorological characteristics, and the measurement availability indicator of the corresponding node.

[0086] Optionally, the matrix generation module 24 includes: The weight determination unit is used to determine the node weight of each node based on the measurement availability flag of each node. The feature correction unit is used to determine the corrected fusion features of each node based on the fusion features of each node and the corresponding node weights. The matrix determination unit is used to determine the network-wide fusion feature matrix based on the modified fusion features of each node.

[0087] Optionally, the power output prediction result includes the power output prediction range of the corresponding new energy access node at the target time; Optionally, the output prediction module 22 includes: The power prediction unit is used to input the historical operating feature sequence of the current renewable energy access node into the trained renewable energy output prediction model for each renewable energy access node, and output the active power output of the current renewable energy access node at the target time. The renewable energy output prediction model uses historical operating feature samples of multiple benchmark renewable energy access nodes as training sets and is trained using a model-independent meta-learning framework. The loss function used in the training process includes a first constraint loss term based on the physical laws of photovoltaic power output and / or a second constraint loss term based on the physical laws of wind power output. The interval generation unit is used to generate the output active power of the current renewable energy access node under each scenario at the target time based on the output active power and using a generative diffusion model, and to determine the output prediction interval of the current renewable energy access node at the target time based on the output active power under each scenario.

[0088] Optionally, the power flow solving module 25 includes: The equation construction unit is used to construct node power balance equations based on the topology data. The constraint embedding unit is used to determine the net injected power of each node at the target time and the constraint boundary of the net injected power based on the node load power of each node at the target time and the output prediction interval, and to embed the constraint boundary of the net injected power into the node power balance equation. The power flow calculation unit is used to generate initial values ​​for power flow solution using the network-wide fusion feature matrix, and to solve the node power balance equations embedded in the constraint boundaries by combining node voltage security constraints and branch capacity constraints, thereby obtaining the power flow analysis results.

[0089] Optionally, the feature fusion module 23 includes: The temporal feature determination unit is used to process the historical running feature sequence of the current node using a temporal feature extraction network to obtain the temporal dynamic features of the current node. The electrical feature determination unit is used to process the node voltage data of the current node and the adjacency relationship in the topology data using an electrical topology feature encoding network, and to aggregate the node voltage data of neighboring nodes to obtain the electrical spatial coupling features of the current node.

[0090] Optionally, the power flow solving module 25 includes: A partitioning unit is used to adaptively divide the power distribution network to be analyzed into multiple autonomous partitions based on the output prediction results and the topology data. The local data determination unit is used to determine the local topology data, local power output prediction results, and local fusion feature matrix of each autonomous region based on the topology data, the power output prediction results, and the network-wide fusion feature matrix. The local power flow solution unit is used to establish and solve local power flow equations based on the corresponding local topology data, local output prediction results and local fusion feature matrix of each autonomous region, so as to obtain local power flow solutions. The boundary coordination unit is used to take the voltage and phase angle of the corresponding boundary node in the local power flow solution as boundary variables, coordinate the consistency of the boundary variables with the adjacent autonomous regions through the alternating direction multiplier method, and perform encrypted gradient exchange in combination with the vertical federated learning mechanism. The parameter adjustment unit is used to dynamically adjust the penalty parameter of the alternating direction multiplier method according to the consistency residual of the boundary variables of the adjacent autonomous partitions during the iterative coordination process of the alternating direction multiplier method. The result output unit is used to output the network node voltage, phase angle and branch power results after global coordination, as the power flow analysis results.

[0091] Optionally, the data acquisition module 21 includes: The data acquisition unit is used to acquire the topology information of the power distribution network to be analyzed, as well as the multi-source operating data of each node at multiple sampling times; The topology verification unit is used to construct the topology connection relationship of the distribution network to be analyzed based on the topology information, and to perform consistency verification on the topology connection relationship; The data preprocessing unit is used to preprocess the multi-source operational data of each node at the multiple sampling times, determine the measurement availability flag of each node, and obtain the preprocessed operational data of each node. The standardization processing unit is used to take the verified topological connection relationship as the topological structure data, and to standardize the preprocessed running data of each node to form the historical running data of each node.

[0092] The power flow analysis device for distribution networks provided in the embodiments of the present invention can execute the power flow analysis method for distribution networks provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0093] Example 3 Figure 3 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0094] like Figure 3As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0095] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0096] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as power flow analysis methods for distribution networks.

[0097] In some embodiments, the power flow analysis method for a distribution network can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the power flow analysis method for a distribution network described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the power flow analysis method for a distribution network by any other suitable means (e.g., by means of firmware).

[0098] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0099] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0100] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0101] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0102] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0103] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0104] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0105] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0106] This invention also provides a computer program product, including a computer program and / or instructions, which, when executed by a processor, implements the power flow analysis method for distribution networks as provided in any embodiment of this invention.

[0107] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0108] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A power flow analysis method for a distribution network, characterized in that, include: The historical operation data and topology data of each node in the distribution network to be analyzed are obtained, wherein the nodes include new energy access nodes and new energy non-access nodes; Based on the historical operating data of each new energy access node, the power output of new energy is predicted, and the power output prediction results of each new energy access node at the target time are obtained. For each node, based on the historical operation data and topology data of the current node, the temporal dynamic characteristics and electrical spatial coupling characteristics of the current node are determined; the temporal dynamic characteristics and electrical spatial coupling characteristics are dynamically weighted and fused through an attention gating mechanism to obtain the fused characteristics of the current node; Based on the fusion characteristics of each node, the network-wide fusion characteristic matrix is ​​determined; Based on the power output prediction results, the network-wide integrated feature matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network.

2. The method according to claim 1, characterized in that, The historical operating data includes a historical operating feature sequence. Each historical operating feature in the historical operating feature sequence includes historical renewable energy output power, node load active power, node load reactive power, node voltage data, environmental meteorological characteristics, and the measurement availability indicator of the corresponding node.

3. The method according to claim 2, characterized in that, The determination of the network-wide fusion feature matrix based on the fusion characteristics of each node includes: The node weight of each node is determined based on the measured availability indicators of each node. Based on the fusion characteristics of each node and the corresponding node weights, the modified fusion characteristics of each node are determined. Based on the modified fusion characteristics of each node, the network-wide fusion characteristic matrix is ​​determined.

4. The method according to claim 2, characterized in that, The power output prediction result includes the power output prediction range of the corresponding new energy access node at the target time; The process of predicting the output of new energy sources based on historical operating data of each new energy access node, and obtaining the predicted output of each new energy access node at the target time, includes: For each renewable energy access node, the historical operating feature sequence of the current renewable energy access node is input into the trained renewable energy output prediction model, and the active power output of the current renewable energy access node at the target time is output. The renewable energy output prediction model uses historical operating feature samples of multiple benchmark renewable energy access nodes as training sets and is trained using a model-independent meta-learning framework. The loss function used in the training process includes a first constraint loss term based on the physical laws of photovoltaic power output and / or a second constraint loss term based on the physical laws of wind power output. Based on the output active power, a generative diffusion model is used to generate the output active power of the current renewable energy access node under each scenario at the target time, and the output prediction range of the current renewable energy access node at the target time is determined based on the output active power under each scenario.

5. The method according to claim 4, characterized in that, Based on the output prediction results, the network-wide integrated feature matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network, including: Based on the aforementioned topology data, node power balance equations are constructed. Based on the node load power of each node at the target time and the output prediction interval, determine the net injected power of each node at the target time and the constraint boundary of the net injected power, and embed the constraint boundary of the net injected power into the node power balance equation. The initial values ​​for power flow analysis are generated using the network-wide integrated feature matrix. Combined with node voltage safety constraints and branch capacity constraints, the node power balance equations embedded in the constraint boundaries are solved to obtain the power flow analysis results.

6. The method according to claim 2, characterized in that, The determination of the temporal dynamic characteristics and electrical-spatial coupling characteristics of the current node based on its historical operational data and topology data includes: The historical operation feature sequence of the current node is processed using a temporal feature extraction network to obtain the temporal dynamic features of the current node; The electrical topology feature coding network is used to process the node voltage data of the current node and the adjacency relationship in the topology data, and the node voltage data of neighboring nodes are aggregated to obtain the electrical spatial coupling characteristics of the current node.

7. The method according to claim 1, characterized in that, Based on the output prediction results, the network-wide integrated feature matrix, and the topology data, power flow analysis is performed to obtain the power flow analysis results of the distribution network, including: Based on the power output prediction results and the topology data, the distribution network to be analyzed is adaptively divided into multiple autonomous partitions; Based on the topology data, the power output prediction results, and the network-wide fusion feature matrix, the local topology data, local power output prediction results, and local fusion feature matrix of each autonomous region are determined. By establishing and solving local power flow equations for each autonomous region based on corresponding local topology data, local power output prediction results, and local fusion feature matrix, local power flow solutions are obtained. The voltage and phase angle of the corresponding boundary node in the local power flow solution are used as boundary variables. The consistency of the boundary variables is coordinated with the adjacent autonomous regions through the alternating direction multiplier method, and the encrypted gradient exchange is carried out in combination with the vertical federated learning mechanism. During the iterative coordination process of the alternating direction multiplier method, the penalty parameter of the alternating direction multiplier method is dynamically adjusted based on the consistency residuals of the boundary variables of adjacent autonomous partitions. The output, after global coordination, shows the total network node voltage, phase angle, and branch power as the power flow analysis results.

8. The method according to claim 1, characterized in that, The acquisition of historical operating data and topology data of each node in the distribution network to be analyzed includes: Obtain the topology information of the distribution network to be analyzed, as well as the multi-source operating data of each node at multiple sampling times; Based on the topology information, the topology connection relationship of the distribution network to be analyzed is constructed, and the consistency of the topology connection relationship is checked. The multi-source operational data of each node at the multiple sampling times are preprocessed, and the measurement availability flag of each node is determined to obtain the preprocessed operational data of each node. The verified topological connection relationship is used as the topological structure data, and the preprocessed running data of each node is standardized to form the historical running data of each node.

9. A power flow analysis device for a power distribution network, characterized in that, include: The data acquisition module is used to acquire historical operating data and topology data of each node in the distribution network to be analyzed, wherein the nodes include new energy access nodes and new energy non-access nodes; The power output prediction module is used to predict the power output of new energy sources based on the historical operating data of each new energy access node, and to obtain the power output prediction results of each new energy access node at the target time. The feature fusion module is used to determine the temporal dynamic features and electrical-spatial coupling features of each node based on its historical operation data and topology data; and to dynamically weight and fuse the temporal dynamic features and electrical-spatial coupling features through an attention gating mechanism to obtain the fused features of the current node. The matrix generation module is used to determine the network-wide fusion feature matrix based on the fusion characteristics of each node. The power flow solution module is used to perform power flow analysis and solution based on the power output prediction results, the network-wide integrated feature matrix, and the topology data, to obtain the power flow analysis results of the distribution network.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the power flow analysis method for the distribution network according to any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the power flow analysis method for the distribution network as described in any one of claims 1-8.

12. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the power flow analysis method for the distribution network as described in any one of claims 1-8.

Images