Method and system for determining distributed photovoltaic consumption capability of power distribution network

Abstract

The invention relates to the technical field of power system management, and discloses a method and a system for determining distributed photovoltaic consumption capability of a power distribution network. The method comprises the steps of collecting multi-dimensional power grid data, extracting characteristics of photovoltaic output fluctuation and the like, generating a fused power grid characteristic graph, constructing a consumption capacity knowledge base, and determining the maximum photovoltaic installation capacity and the minimum photovoltaic installation capacity through a dynamic adjustment algorithm based on the two. Carrying out multi-scale decomposition on load time sequence data, and processing the data by adopting an edge weight distribution algorithm and the like; optimizing by using an improved particle swarm algorithm; detecting meteorological abnormity and dynamically re-evaluating the photovoltaic capacity; and constructing the compliance of the knowledge graph verification scheme. The system comprises a data acquisition module, a feature extraction module and the like. According to the method, various factors are comprehensively considered, the absorption capacity can be accurately determined, the photovoltaic installation capacity and the power grid stability can be balanced, the operation safety and efficiency of the power grid are improved, and effective support is provided for planning and operation of distributed photovoltaic in the power distribution network.

Description

Technical Field

[0001] The present invention relates to the technical field of power system management, and in particular to a method and system for determining the distributed photovoltaic absorption capacity of a distribution network. Background Art

[0002] With the growing global demand for clean energy, distributed photovoltaic power generation is increasingly being used in distribution networks due to its clean, flexible, and decentralized advantages. The integration of distributed photovoltaic power into distribution networks can effectively reduce traditional fossil fuel consumption, lower carbon emissions, and provide strong support for sustainable energy development. However, the large-scale integration of distributed photovoltaic power also poses numerous challenges to distribution networks, with absorption capacity being a particularly prominent issue.

[0003] The structure and operational characteristics of distribution networks are complex, with numerous nodes and diverse line layouts, resulting in significant variations in load characteristics across regions. Distributed photovoltaic power generation is intermittent and fluctuating, significantly affected by meteorological conditions such as sunlight intensity and temperature. During periods of abundant sunlight, photovoltaic output can surge; however, during periods of cloudy weather and at night, output can plummet or even reach zero. This volatile output characteristic makes it difficult to precisely match the relatively stable load demands of the distribution network, easily leading to power imbalances in certain areas. When photovoltaic output exceeds the distribution network's capacity, voltage overshoot can occur, causing voltages at certain nodes to become excessively high or low, impacting the normal operation of various equipment in the grid, shortening their service life, and even potentially damaging them. It can also cause line overloads, increase line losses, and reduce grid efficiency, posing a serious threat to the grid's safe and stable operation.

[0004] Existing methods for determining the distributed photovoltaic (PV) absorption capacity of distribution networks suffer from numerous shortcomings. Some traditional methods simply consider a single factor, such as estimating absorption capacity based solely on historical load data or PV output data. These methods ignore the grid topology, changing meteorological conditions, and the complex coupling relationship between the two, resulting in poorly accurate calculation results and an inability to provide a reliable basis for practical engineering projects. Some methods suffer from low computational efficiency when faced with large-scale data and complex grid structures, making it difficult to meet real-time requirements and respond promptly to dynamic changes in grid operating conditions. Furthermore, most existing methods lack a comprehensive consideration of grid operating constraints, such as inter-node power transmission constraints and equipment capacity limitations, making the determined absorption capacity unfeasible in practical applications.

[0005] With the continuous growth of distributed photovoltaic installed capacity and the demand for intelligent development of distribution networks, there is an urgent need for a method and system that can comprehensively consider multiple factors and accurately and efficiently determine the distributed photovoltaic absorption capacity of the distribution network, so as to solve the current technical difficulties and achieve the coordinated development of distributed photovoltaic and distribution networks. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for determining the distributed photovoltaic absorption capacity of a distribution network to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for determining the distributed photovoltaic absorption capacity of a distribution network, the method comprising:

[0008] Collect multi-dimensional grid data on the operating status of the distribution network, including distributed photovoltaic equipment parameters, load time series data, grid topology and real-time meteorological monitoring data;

[0009] Extracting features from the multidimensional power grid data to obtain a power grid feature vector corresponding to each dimension, the power grid feature vector including: photovoltaic output fluctuation characteristics, load demand spatiotemporal distribution characteristics, topological connection characteristics, and meteorological correlation characteristics;

[0010] Mapping the power grid feature vector to a unified graph structure space based on a distributed graph computing framework to generate a fused power grid feature graph;

[0011] Building an absorption capacity knowledge base based on historical absorption records, wherein the knowledge base stores absorption benchmark data, including: the maximum allowable photovoltaic capacity of each node, power transmission constraints between nodes, and historical absorption efficiency records;

[0012] Based on the integrated grid characteristic diagram and the absorption benchmark data in the absorption capacity knowledge base, the maximum installed capacity and the minimum installed capacity of the distributed photovoltaic power generation system of the distribution network are determined through a dynamic adjustment algorithm.

[0013] Preferably, the determining of the maximum and minimum installed capacity of distributed photovoltaics in the distribution network by a dynamic adjustment algorithm includes:

[0014] generating an initial set of candidate nodes for consumption according to the association between the fused power grid characteristic graph and the nodes in the knowledge base;

[0015] Constructing an absorption capacity constraint graph based on the topological connection characteristics, wherein nodes in the constraint graph represent grid nodes and edges represent power transmission paths and capacity constraints;

[0016] According to the absorption capacity constraint graph and real-time meteorological monitoring data, the initial absorption candidate node set is iteratively adjusted through a graph optimization algorithm to output the final maximum photovoltaic installation capacity and minimum photovoltaic installation capacity.

[0017] Preferably, the load time series data is subjected to feature extraction using a multi-scale decomposition network, including:

[0018] Use wavelet transform to separate the fundamental frequency component and high frequency harmonic component of load data;

[0019] The daily and weekly cycle features are extracted from the fundamental frequency component through the time convolution network;

[0020] A gated recurrent unit is used to capture transient impact characteristics of high-frequency harmonic components.

[0021] Preferably, mapping the power grid feature vector to a unified graph structure space based on a distributed graph computing framework includes:

[0022] Quantify the feasibility and efficiency of power transfer between nodes through edge weight allocation algorithm;

[0023] The weighted features are input into the graph attention network to aggregate the absorption capacity correlation features of adjacent nodes.

[0024] Preferably, the dynamic adjustment algorithm further includes:

[0025] A capacity optimization agent strategy is constructed based on a multi-objective optimization model. The agent strategy dynamically balances the photovoltaic installation capacity and grid stability through constraint functions. The constraint functions include voltage over-limit penalty, line overload penalty and absorption efficiency reward.

[0026] Preferably, the graph optimization algorithm is an improved particle swarm optimization algorithm, comprising:

[0027] Encoding the absorption capacity constraint graph into a particle swarm position vector, wherein the vector dimension represents the photovoltaic capacity of the node;

[0028] The particle velocity and position are updated through a global optimal guidance strategy, and a local perturbation operator is introduced to avoid premature convergence;

[0029] The optimal solution set is screened according to a fitness function, where the fitness function is calculated based on capacity utilization and constraint violation.

[0030] Preferably, the method further comprises:

[0031] Applying an anomaly detection model to the real-time meteorological monitoring data to identify sudden changes in light intensity or extreme weather events;

[0032] When abnormal weather is detected, dynamic reassessment of photovoltaic capacity is triggered, and the installed capacity range is updated based on the reassessment results.

[0033] Preferably, the dynamic reassessment adopts a fuzzy clustering rule engine, and the specific method includes:

[0034] Define the fuzzy membership of meteorological anomaly level, node capacity elasticity and grid robustness;

[0035] Candidate capacity adjustment plans are generated through a fuzzy decision tree, and the optimal plan is selected based on the maximum membership principle.

[0036] Preferably, the method further comprises:

[0037] Constructing a knowledge graph of consumption rules, in which nodes represent power grid equipment entities and edges represent consumption rules or operation constraints;

[0038] During the capacity determination process, the compliance of the installation capacity plan with the knowledge graph is verified through a subgraph isomorphism algorithm.

[0039] Preferably, the present invention further includes a system for determining the distributed photovoltaic absorption capacity of a distribution network, the system comprising:

[0040] Data acquisition module, used to obtain multi-dimensional grid data, including distributed photovoltaic equipment parameters, load time series data, grid topology and real-time meteorological monitoring data;

[0041] a feature extraction module for extracting photovoltaic output fluctuation characteristics, load demand spatiotemporal distribution characteristics, topological connection characteristics, and meteorological correlation characteristics from the multidimensional power grid data;

[0042] A graph structure fusion module maps the power grid feature vector to a unified graph structure space based on a distributed graph computing framework to generate a fused power grid feature graph;

[0043] The knowledge base construction module is used to store the consumption benchmark data in the historical consumption records, including the maximum allowable photovoltaic capacity of each node, the power transmission constraints between nodes, and the historical consumption efficiency records;

[0044] The dynamic adjustment module determines the maximum and minimum installed capacity of distributed photovoltaics by combining the integrated grid characteristic diagram and the absorption benchmark data through a dynamic adjustment algorithm.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The method and system for determining the distributed photovoltaic absorption capacity of the distribution network provided by the present invention have shown significant beneficial effects in many aspects. In the data processing and feature extraction link, by collecting multi-dimensional grid data such as distributed photovoltaic equipment parameters, load time series data, grid topology structure and real-time meteorological monitoring data, and performing targeted feature extraction, the photovoltaic output fluctuation characteristics, load demand spatiotemporal distribution characteristics, topological connection characteristics and meteorological correlation characteristics are obtained. This multi-dimensional data fusion and feature extraction method can comprehensively and accurately characterize the operating status of the distribution network and the complex relationship between distributed photovoltaic and distribution networks. Compared with traditional methods that only consider a single or a few factors, the present invention can more deeply explore the information behind the data, laying a solid foundation for the subsequent accurate determination of absorption capacity. For example, a multi-scale decomposition network is used to extract features from load time series data. By using wavelet transform, time convolution network and gated recurrent unit, load characteristics can be analyzed from different time scales and frequency components, and daily cycle characteristics, weekly cycle characteristics and transient impact characteristics can be captured. These characteristics are crucial for accurately predicting load change trends and rationally arranging photovoltaic power absorption.

[0047] To generate a converged grid characteristic diagram and build a knowledge base for absorptive capacity, the distributed graph computing framework is used to map grid characteristic vectors into a unified graph structure space. A knowledge base for absorptive capacity is constructed based on historical absorptive records, storing absorptive benchmark data such as the maximum allowable photovoltaic capacity of each node, inter-node power transmission constraints, and historical absorptive efficiency records. The converged grid characteristic diagram uses an intuitive graph structure to display the relationships and characteristic distributions between various grid components, facilitating subsequent analysis and processing. The absorptive capacity knowledge base provides rich historical empirical data support for determining absorptive capacity. This combination allows the absorptive capacity to be determined by fully considering both the grid's historical operation and its current structure and characteristics, improving the accuracy and reliability of the results.

[0048] In determining the PV installation capacity, a dynamic adjustment algorithm, combined with a fusion of grid characteristic graphs and absorption benchmark data, is used to determine the maximum and minimum installed capacity of distributed PV. This dynamic adjustment algorithm constructs a capacity optimization agent strategy based on a multi-objective optimization model. It dynamically balances PV installation capacity with grid stability through constraint functions such as voltage over-limit penalties, line overload penalties, and absorption efficiency rewards. Simultaneously, iterative adjustments are made using graph optimization algorithms, such as an improved particle swarm optimization algorithm, to avoid premature convergence and ensure that an optimal or near-optimal PV installation capacity solution is found. This approach maximizes the absorption capacity of distributed PV while ensuring the safe and stable operation of the grid. It avoids problems such as voltage over-limit and line overload caused by inappropriate PV access capacity, reduces grid operation risks, and improves grid efficiency.

[0049] In addition, the present invention also performs anomaly detection on real-time meteorological monitoring data. When a sudden change in light intensity or an extreme weather event is detected, it triggers a dynamic reassessment of photovoltaic capacity and uses a fuzzy clustering rule engine to determine the optimal capacity adjustment plan based on the meteorological anomaly level, node capacity elasticity, and grid robustness. This enables the system to respond promptly to the impact of changing meteorological conditions on photovoltaic output, dynamically adjust the photovoltaic installation capacity range, and further ensure the safe and stable operation of the power grid under different meteorological conditions. Moreover, by constructing a knowledge graph of absorption rules and verifying the compliance of the installation capacity plan with the knowledge graph through a subgraph isomorphism algorithm, it ensures that the determined installation capacity plan complies with the absorption rules and operating constraints of the power grid, thereby improving the feasibility and practicality of the plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a working principle diagram of the method for determining the distributed photovoltaic absorption capacity of the distribution network according to the present invention;

[0051] Figure 2 Flowchart for load time series data feature extraction;

[0052] Figure 3 This is a flow chart of dynamic reassessment of photovoltaic capacity under abnormal weather conditions;

[0053] Figure 4 This is a flowchart for verifying the compliance of the installation capacity plan based on the consumption rule knowledge graph. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] See also Figures 1-4 The present invention provides a technical solution: The present invention provides a method for determining the distributed photovoltaic absorption capacity of a distribution network, and the specific steps are as follows:

[0056] Utilizing various sensors, monitoring equipment, and data acquisition systems, we collect multidimensional grid data on the operating status of the distribution network. Distributed photovoltaic equipment parameters include the model, specifications, rated power, and conversion efficiency of the photovoltaic panels, which directly affect the power generation capacity of the photovoltaic equipment. Load time series data records the magnitude of power load at different times, reflecting how power demand changes over time. The grid topology details the connections between various nodes in the grid (such as substations, transformers, and user terminals), as well as information such as the direction, length, and impedance of the lines, and is the fundamental architecture for power transmission. Real-time meteorological monitoring data includes information such as light intensity, temperature, humidity, and wind speed. Meteorological conditions have a significant impact on the output of distributed photovoltaic systems.

[0057] For the collected multi-dimensional power grid data, appropriate technical means are used to extract features to obtain the power grid feature vectors corresponding to each dimension. For the parameters of distributed photovoltaic equipment, their output characteristics under different operating conditions are analyzed, and the fluctuation characteristics of photovoltaic output are extracted. For example, the characteristics are characterized by statistically analyzing the amplitude and frequency of changes in photovoltaic power within a certain time interval. For load time series data, the patterns in its time series are mined to obtain the spatiotemporal distribution characteristics of load demand, such as analyzing the time and intensity of load peaks and troughs in different regions and time periods. Topological connection features are extracted from the power grid topology, including the degree of the node (the number of connected edges), the shortest path between nodes, and the connectivity of the network. Combining the correlation between real-time meteorological monitoring data and photovoltaic output, meteorological correlation features are extracted, such as the linear relationship between light intensity and photovoltaic power, and the influence of temperature on photovoltaic conversion efficiency.

[0058] Leveraging a distributed graph computing framework, the grid feature vectors obtained above are mapped to a unified graph space. During this process, the different feature vectors are integrated and transformed to conform to the graph representation, generating a fused grid feature graph. This graph visually displays the relationships between grid components and the distribution of various features using nodes and edges, providing a unified data structure for subsequent analysis.

[0059] Based on historical absorption records, a knowledge base of absorption capacity is constructed. From this historical data, the maximum allowable photovoltaic capacity of each node is collected and organized. This is the maximum power of photovoltaic equipment that can be connected to each node, determined based on factors such as the rated capacity of grid equipment and safe operating restrictions. Inter-node power transmission constraints involve transmission capacity limits and power flow restrictions to ensure the safe and stable transmission of power within the grid. Historical absorption efficiency records reflect the absorption of photovoltaic power under different historical circumstances, including absorption ratios under different time periods and different photovoltaic connection capacities. This data is stored in the knowledge base and serves as an important reference for subsequent determination of photovoltaic absorption capacity.

[0060] By integrating grid characteristic maps with absorption benchmark data from a knowledge base of absorption capacity, a dynamic adjustment algorithm is used to determine the maximum and minimum installed capacity of distributed photovoltaic power generation within the distribution network. This algorithm comprehensively considers various grid characteristics, constraints, and historical absorption experience. Through continuous adjustment and optimization, it determines the range of photovoltaic installation capacity that can meet power demand and ensure safe and stable operation under the current grid conditions.

[0061] The present invention will be further described below in conjunction with Examples 1 to 6:

[0062] Example 1:

[0063] In the process of determining the maximum and minimum installed capacity of distributed photovoltaic power generation in the distribution network, the specific implementation is as follows:

[0064] An initial set of candidate nodes for distributed photovoltaic installation is generated based on the correlation between the integrated grid characteristic graph and the nodes in the knowledge base. The integrated grid characteristic graph displays the various relationships between grid nodes. By analyzing these relationships and combining the performance of each node in the historical absorption process and the degree of correlation between them recorded in the knowledge base, nodes that may be suitable for distributed photovoltaic installation are screened to form the initial set of candidate nodes for absorption. For example, nodes that are closely connected to high-efficiency nodes in historical data and have a certain power carrying capacity are prioritized for inclusion in the candidate set.

[0065] A capacity constraint graph is constructed based on topological connectivity features. In this constraint graph, grid nodes serve as the graph's nodes, while the power transmission paths and capacity constraints between nodes are represented by edges. Each edge is assigned corresponding attributes, including parameters such as the line's maximum transmission power, resistance, and reactance. These parameters limit the power transmission capacity between nodes, reflecting the physical characteristics and operational constraints of the power grid. For example, due to the limitations of the conductor gauge and length, the maximum transmission power of a line is Pmax. In the constraint graph, this is reflected as the power transmission limit of the edge connecting the two nodes being Pmax.

[0066] Based on the absorption capacity constraint graph and real-time meteorological monitoring data, a graph optimization algorithm iteratively adjusts the initial set of candidate absorption nodes. Real-time meteorological monitoring data can affect the output of distributed photovoltaic systems. For example, when sunlight intensity increases, photovoltaic output increases. The graph optimization algorithm takes these factors into account during the iterative adjustment process. For example, at a certain moment, if real-time meteorological data indicates sufficient sunlight and photovoltaic output is expected to increase, the algorithm checks the power acceptance capacity of each node in the candidate node set based on the absorption capacity constraint graph. If a node's power acceptance capacity approaches or reaches the upper limit, the algorithm adjusts its priority in the candidate set or removes it from the set, while considering other nodes with acceptable capacity. After multiple iterations, the algorithm ultimately outputs the final maximum and minimum installed photovoltaic capacity that meets the requirements. This capacity determination takes into account both the grid topology and power transmission constraints, as well as the impact of real-time meteorological conditions on photovoltaic output, ensuring the accuracy and practicality of the results.

[0067] Example 2:

[0068] A multi-scale decomposition network is used to extract features from load time series data. The specific operations are as follows:

[0069] The first step is to use wavelet transform to separate the fundamental frequency component and high-frequency harmonic component of the load data. Wavelet transform is a time-frequency analysis method that can decompose complex time series signals into components of different frequencies and time scales. For load time series data, by selecting appropriate wavelet functions and decomposition layers, it is decomposed into fundamental frequency components and high-frequency harmonic components. The fundamental frequency component reflects the basic trend of load changes over time, which is usually related to changes in longer time scales such as daily cycles and weekly cycles; the high-frequency harmonic component contains the rapid changes and transient impact information of the load in a short period of time. For example, in the load data of a certain city, after decomposition through wavelet transform, it was found that the fundamental frequency component showed obvious daily peak and trough changes, while the high-frequency harmonic component would fluctuate greatly at certain special moments (such as when large industrial equipment starts or stops).

[0070] The second step is to extract diurnal and weekly cycle features from the fundamental frequency component using a temporal convolutional network. A temporal convolutional network (TCN) is a neural network structure specifically designed for processing time series data. It extracts data features along the time dimension through convolution operations. The separated fundamental frequency component is input into the TCN, where the convolution kernel slides along the time axis, capturing the periodic features within the fundamental frequency component. For example, the convolution kernel can learn similar load variation patterns within the same time period each day, thereby extracting diurnal cycle features. Similarly, by learning from data within a week, weekly cycle features can be extracted. These periodic features are important for predicting load trends and rationally arranging power supply.

[0071] The third step is to use a gated recurrent unit (GRU) to capture transient impact characteristics of high-frequency harmonic components. The gated recurrent unit (GRU) is a special recurrent neural network structure that can effectively handle long-term dependencies and short-term transient changes in time series. Transient impact characteristics in high-frequency harmonic components are often brief and intense. Through its gating mechanism, the GRU can selectively memorize and forget information, accurately capturing these transient impact characteristics. For example, when a large commercial building suddenly turns on a large number of lighting equipment, a brief spike will appear in the load data. The GRU can promptly perceive this change and learn and record it as a transient impact characteristic, providing a basis for subsequent analysis and prediction of abnormal load changes. Through these three steps, the multi-scale decomposition network can comprehensively and deeply extract the characteristics of load time series data, providing strong support for determining the distributed photovoltaic absorption capacity of the distribution network.

[0072] Example 3:

[0073] Based on the distributed graph computing framework, the grid feature vector is mapped to a unified graph structure space. The specific implementation is as follows:

[0074] The feasibility and efficiency of power transmission between nodes are quantified using an edge weight assignment algorithm. In power grids, power transmission between nodes is affected by a variety of factors, such as line impedance, transmission distance, and device capacity. The edge weight assignment algorithm comprehensively considers these factors and assigns a weight value to each edge in the graph structure. For example, for a short, low-impedance line with a large connected device capacity, the corresponding edge weight can be set high, indicating high feasibility and efficiency of power transmission on this line; conversely, for a long, high-impedance line with limited connected device capacity, the edge weight is lower. In this way, the complex characteristics of power transmission between nodes are converted into concrete numerical representations, which are intuitively reflected in the graph structure.

[0075] The weighted features are input into the graph attention network to aggregate the absorption capacity correlation features of adjacent nodes. The graph attention network (GAT) is a graph neural network model based on the attention mechanism, which can automatically learn the importance weights between different nodes. In this embodiment, the power grid feature vector processed by the edge weight distribution algorithm is input into the GAT. The GAT calculates the importance weight of each adjacent node to the current node based on the characteristics of each node and its relationship with its adjacent nodes. For example, for a specific node, those adjacent nodes that are closely connected to it and have strong absorption capacity will be given higher weights, while those adjacent nodes with weak absorption capacity or weak connection relationship will be given lower weights. Then, the GAT performs weighted aggregation on the absorption capacity correlation features of the adjacent nodes according to these weights to obtain the comprehensive absorption capacity features of the current node. In this way, the association information between the nodes in the graph structure is fully utilized, so that the generated fusion power grid feature map can more accurately reflect the absorption capacity relationship between the various parts of the power grid, providing a more reliable data basis for the subsequent determination of photovoltaic absorption capacity.

[0076] Example 4:

[0077] The process of determining the distributed photovoltaic absorption capacity of the distribution network involves the specific implementation of the dynamic adjustment algorithm and the graph optimization algorithm:

[0078] A capacity optimization agent strategy is constructed based on a multi-objective optimization model. The multi-objective optimization model aims to balance multiple conflicting objectives. In the present invention, a capacity optimization agent strategy is constructed to balance the photovoltaic installation capacity and the grid stability. Among them, the constraint function plays a key role, including voltage over-limit penalty, line overload penalty and absorption efficiency reward. The voltage over-limit penalty is to ensure that the grid voltage operates within a safe and stable range. When the increase in photovoltaic installation capacity causes the voltage of some nodes in the grid to exceed the allowable range, the objective function will be penalized according to the degree of voltage over-limit. Assume that the actual voltage of node i is V i , the allowable voltage range is [V min ,V max ], the voltage limit penalty function can be expressed as:

[0079]

[0080] where k V1 and k V2 is the penalty coefficient, which is set according to the actual situation of the power grid. The line overload penalty is used to prevent the line transmission power from exceeding its rated capacity, avoiding problems such as line overheating and damage. If the actual transmission power of line j is P j , the rated transmission power is P jmax , the line overload penalty function is:

[0081]

[0082] where k L is the penalty coefficient. The absorption efficiency reward encourages the improvement of the absorption ratio of photovoltaic power. If the photovoltaic absorption efficiency of a certain area is η, the absorption efficiency reward function can be R η =k η η,k η is the reward coefficient. Through the combined effect of these constraint functions, the photovoltaic installation capacity and grid stability are dynamically balanced. Secondly, the graph optimization algorithm adopts the improved particle swarm algorithm. The absorption capacity constraint graph is encoded as a particle swarm position vector, and the vector dimension represents the photovoltaic capacity of the node. Each particle represents a possible photovoltaic installation capacity allocation scheme, and each element in its position vector corresponds to the photovoltaic capacity of a grid node. For example, for a grid containing n nodes, the particle position vector can be expressed as X = [x1, x2,…, x n ], where x i represents the photovoltaic installed capacity of node i. The particle speed and position are updated through the global optimal guidance strategy. In each iteration, the particle adjusts its speed and position according to its own historical optimal position pBest and the global optimal position gBest. The particle speed update formula is:

[0083]

[0084] in is the velocity of particle i at the tth iteration, ω is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1]; the particle position update formula is At the same time, a local perturbation operator is introduced to avoid premature convergence. During the iteration process, the particle position is locally perturbed with a certain probability, such as randomly adjusting some dimensions of a particle, to prevent the algorithm from falling into a local optimal solution. Finally, the optimal solution set is selected according to the fitness function, which is calculated based on capacity utilization and constraint violation. Capacity utilization reflects the effective utilization of photovoltaic installation capacity, and constraint violation measures the violation of the solution to constraints such as voltage limit and line overload. The fitness function can be expressed as:

[0085]

[0086] where α is a weight coefficient used to balance the importance of capacity utilization and constraint violation, x imax is the maximum allowable PV capacity of node i, m and l are the number of lines and nodes, respectively. Through continuous iteration and screening, a PV installation capacity plan that meets the requirements is finally obtained.

[0087] Example 5:

[0088] In the process of determining the accommodation capacity of distributed photovoltaic in the distribution network, the processing of real-time meteorological monitoring data and the dynamic re-evaluation of photovoltaic capacity are implemented as follows:

[0089] An anomaly detection model is used for real-time meteorological monitoring data to identify sudden changes in light intensity or extreme weather events. The anomaly detection model can be constructed using methods such as statistical analysis and machine learning. For example, the method based on statistical analysis can calculate the mean and standard deviation of historical meteorological data and set the range of normal meteorological data. When the real-time monitored light intensity exceeds this range by a certain degree, it is determined as a sudden change in light intensity; for extreme weather events such as heavy rain and strong wind, a corresponding meteorological feature model can be established and judged by combining multiple meteorological parameters. Assuming that the mean of light intensity is μ and the standard deviation is σ obtained through historical data statistics, when the real-time light intensity I satisfies |I - μ| > kσ (k is a threshold set according to the actual situation), an anomaly detection alarm is triggered.

[0090] When abnormal meteorology is detected, the dynamic re-evaluation of photovoltaic capacity is triggered. The dynamic re-evaluation uses a fuzzy clustering rule engine, and the specific method is as follows: Define the fuzzy membership degrees of meteorological anomaly level, node capacity elasticity, and grid robustness. The meteorological anomaly level can be divided into different levels such as mild anomaly, moderate anomaly, and severe anomaly, and the fuzzy membership degree function of the anomaly level is determined by setting the thresholds of indicators such as the change rate of light intensity and the temperature change range. For example, for the change rate of light intensity r, if r1 < r ≤ r2, it belongs to a mild anomaly, and its fuzzy membership degree is μ light (r), which is calculated through the corresponding function. The node capacity elasticity represents the ability of the node to withstand changes in photovoltaic capacity under different working conditions, and its fuzzy membership degree function is determined according to the equipment parameters and historical operation data of the node. The grid robustness reflects the ability of the grid to maintain stable operation under abnormal conditions, and its fuzzy membership degree function is also constructed by analyzing factors such as the topological structure and power transmission capacity of the grid.

[0091] Then, candidate capacity adjustment plans are generated through a fuzzy decision tree. The fuzzy decision tree generates multiple candidate capacity adjustment plans according to the fuzzy membership degrees of the input meteorological anomaly level, node capacity elasticity, and grid robustness, following the pre-set decision rules. For example, when the meteorological anomaly level is moderate anomaly, and a certain node has high capacity elasticity and good grid robustness, the decision tree may generate a plan to increase the photovoltaic capacity of this node.

[0092] Finally, the optimal plan is selected based on the principle of maximum membership degree. Calculate the membership degree of each candidate plan after comprehensively considering various factors, select the plan with the largest membership degree as the final capacity adjustment plan, and update the installation capacity range according to this plan. In this way, when abnormal meteorology is encountered, the photovoltaic capacity can be adjusted in time to ensure the safe and stable operation of the power grid.

[0093] Example 6:

[0094] In the process of determining the distributed photovoltaic absorption capacity of the distribution network, the knowledge graph of absorption rules is constructed and the compliance of the installation capacity plan is verified as follows:

[0095] Construct a knowledge graph of consumption rules. In this knowledge graph, nodes represent grid equipment entities, such as transformers, lines, distributed photovoltaic equipment, load nodes, etc. Edges represent consumption rules or operating constraints, such as transformer capacity limitations, line power transmission limitations, photovoltaic equipment access rules, etc. Organize this information in the form of a graph to form a structured knowledge network. For example, for a rated capacity of S T The edges between the transformer and the connected lines and load nodes contain the constraints on power transmission and distribution, which stipulate the maximum power transmission from the transformer to each node.

[0096] During the capacity determination process, the compliance of the installation capacity plan with the knowledge graph is verified using a subgraph isomorphism algorithm. This algorithm determines whether a subgraph is identical in structure and attributes to another graph. The determined PV installation capacity plan is converted into a subgraph structure, where nodes represent the location and capacity of PV installations and edges represent power transmission relationships. This subgraph is then matched with the consumption rule knowledge graph. If all nodes and edges in the subgraph can be matched to nodes and edges in the knowledge graph that satisfy the constraints, the installation capacity plan complies with the consumption rules and operational constraints. Conversely, if there is a mismatch, such as if the PV installation capacity of a node exceeds the maximum allowable capacity specified in the knowledge graph or if the power transmission path violates the transmission limits of the line, the plan is deemed non-compliant. In this way, when determining the PV installation capacity plan, the consumption rule knowledge graph can be used for rapid and accurate compliance verification, ensuring the feasibility and safety of the plan and avoiding grid failures or safety hazards caused by rule violations.

[0097] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0098] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for determining the distributed photovoltaic absorption capacity of a distribution network, characterized in that: include: Collect multi-dimensional grid data on the operating status of the distribution network, including distributed photovoltaic equipment parameters, load time series data, grid topology and real-time meteorological monitoring data; Extracting features from the multidimensional power grid data to obtain a power grid feature vector corresponding to each dimension, the power grid feature vector including: photovoltaic output fluctuation characteristics, load demand spatiotemporal distribution characteristics, topological connection characteristics, and meteorological correlation characteristics; Mapping the power grid feature vector to a unified graph structure space based on a distributed graph computing framework to generate a fused power grid feature graph; Building an absorption capacity knowledge base based on historical absorption records, wherein the knowledge base stores absorption benchmark data, including: the maximum allowable photovoltaic capacity of each node, power transmission constraints between nodes, and historical absorption efficiency records; Based on the integrated grid characteristic diagram and the absorption benchmark data in the absorption capacity knowledge base, the maximum installed capacity and the minimum installed capacity of the distributed photovoltaic power generation system of the distribution network are determined through a dynamic adjustment algorithm.

2. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 1, characterized in that: The method of determining the maximum and minimum installed capacity of distributed photovoltaic power generation in the distribution network by a dynamic adjustment algorithm includes: generating an initial set of candidate nodes for consumption according to the association between the fused power grid characteristic graph and the nodes in the knowledge base; Constructing an absorption capacity constraint graph based on the topological connection characteristics, wherein nodes in the constraint graph represent grid nodes and edges represent power transmission paths and capacity constraints; According to the absorption capacity constraint graph and real-time meteorological monitoring data, the initial absorption candidate node set is iteratively adjusted through a graph optimization algorithm to output the final maximum photovoltaic installation capacity and minimum photovoltaic installation capacity.

3. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 2, characterized in that: A multi-scale decomposition network is used to extract features from the load time series data, including: Use wavelet transform to separate the fundamental frequency component and high frequency harmonic component of load data; The daily and weekly cycle features are extracted from the fundamental frequency component through the time convolution network; A gated recurrent unit is used to capture transient impact characteristics of high-frequency harmonic components.

4. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 1, characterized in that: Mapping the power grid feature vector to a unified graph structure space based on a distributed graph computing framework includes: Quantify the feasibility and efficiency of power transfer between nodes through edge weight allocation algorithm; The weighted features are input into the graph attention network to aggregate the absorption capacity correlation features of adjacent nodes.

5. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 2, characterized in that: The dynamic adjustment algorithm also includes: A capacity optimization agent strategy is constructed based on a multi-objective optimization model. The agent strategy dynamically balances the photovoltaic installation capacity and grid stability through constraint functions. The constraint functions include voltage over-limit penalty, line overload penalty and absorption efficiency reward.

6. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 5, characterized in that: The graph optimization algorithm is an improved particle swarm optimization algorithm, which includes: Encoding the absorption capacity constraint graph into a particle swarm position vector, wherein the vector dimension represents the photovoltaic capacity of the node; The particle velocity and position are updated through a global optimal guidance strategy, and a local perturbation operator is introduced to avoid premature convergence; The optimal solution set is screened according to a fitness function, where the fitness function is calculated based on capacity utilization and constraint violation.

7. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 1, characterized in that: The method further comprises: Applying an anomaly detection model to the real-time meteorological monitoring data to identify sudden changes in light intensity or extreme weather events; When abnormal weather is detected, dynamic reassessment of photovoltaic capacity is triggered, and the installed capacity range is updated based on the reassessment results.

8. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 7, characterized in that: The dynamic reassessment adopts a fuzzy clustering rule engine, and the specific method includes: Define the fuzzy membership of meteorological anomaly level, node capacity elasticity and grid robustness; Candidate capacity adjustment plans are generated through a fuzzy decision tree, and the optimal plan is selected based on the maximum membership principle.

9. The method for determining the distributed photovoltaic absorption capacity of a distribution network according to claim 1, characterized in that: The method further comprises: Constructing a knowledge graph of consumption rules, in which nodes represent power grid equipment entities and edges represent consumption rules or operation constraints; During the capacity determination process, the compliance of the installation capacity plan with the knowledge graph is verified through a subgraph isomorphism algorithm.

10. A system for determining the distributed photovoltaic absorption capacity of a distribution network, characterized in that: include: Data acquisition module, used to obtain multi-dimensional grid data, including distributed photovoltaic equipment parameters, load time series data, grid topology and real-time meteorological monitoring data; a feature extraction module for extracting photovoltaic output fluctuation characteristics, load demand spatiotemporal distribution characteristics, topological connection characteristics, and meteorological correlation characteristics from the multidimensional power grid data; A graph structure fusion module maps the power grid feature vector to a unified graph structure space based on a distributed graph computing framework to generate a fused power grid feature graph; The knowledge base construction module is used to store the consumption benchmark data in the historical consumption records, including the maximum allowable photovoltaic capacity of each node, the power transmission constraints between nodes, and the historical consumption efficiency records; The dynamic adjustment module determines the maximum and minimum installed capacity of distributed photovoltaics by combining the integrated grid characteristic diagram and the absorption benchmark data through a dynamic adjustment algorithm.

Images