Power distribution network scheduling method based on heterogeneous feature graph model aggregation analysis

Abstract

The invention discloses a power distribution network scheduling method based on heterogeneous feature graph model aggregation analysis, and belongs to the technical field of intelligent power distribution networks, and the method comprises the steps: S1, obtaining the operation data of a power distribution network, and carrying out the normalization processing, and obtaining an initial data set; s2, establishing a heterogeneous feature graph model based on the initial data set and a corresponding power distribution network topological structure, and determining an initial feature set; s3, performing feature aggregation on the initial feature set based on a heterogeneous feature graph neural network in combination with a feature aggregation formula to determine aggregation features; s4, calculating embedded features according to the aggregation features and different nodes in the heterogeneous feature graph model; s5, calculating adjusting power and line loss of the power distribution network according to the embedded characteristics, and determining a scheduling strategy of the power distribution network according to the adjusting power and the line loss of the power distribution network; the method overcomes the problems that the line loss calculation result error is large and the operation stability of the corresponding power distribution network is poor due to the fact that the operation data change characteristics of the power distribution network are not considered in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of smart distribution networks, and in particular to a distribution network scheduling method based on heterogeneous feature graph model aggregation analysis. Background Art

[0002] With the development of new energy technologies, a large number of photovoltaic power generation systems have been integrated into the distribution network. Since the access points of photovoltaic power generation systems are widely distributed, with significant intermittency and volatility, their power generation capacity changes over time, resulting in increasing uncertainty and volatility in the power flow of the distribution network. At the same time, the output power demand of the distribution network is growing, and the corresponding load equipment and flexible intelligent switches are constantly increasing, which further complicates the power transmission and distribution of the distribution network. However, the traditional power flow calculation method cannot accurately capture the dynamic changes of power in such a complex working environment, which increases the error of the calculated line loss and makes it difficult to design a corresponding solution based on the calculated line loss to maintain the stable operation of the distribution network.

[0003] Chinese patent, publication number: CN117239754A, publication date: December 15, 2023, discloses a wind-solar-thermal multi-power coordinated operation method and system considering carbon emissions, including: aggregating photovoltaic, wind and gas turbines inside the multi-power body, constructing an upper-level multi-power load autonomous output consumption model to solve the output of each power source and the power consumption of each load in the distribution network; according to the power generation of the multi-power body and the power consumption of each load participating in the autonomous output consumption, establish a lower-level distribution network optimization model to solve the distribution network flow results; introduce a demurrage fee agreement and set a double-layer model convergence condition, construct a carbon emission flow model to calculate the carbon emission flow results when the double-layer model converges; collect multi-power parameter data and distribution network parameter data, input the constructed model and solve it to obtain the best operation scheduling result; however, the invention does not take into account the changing characteristics of the distribution network operation data, resulting in a large line loss error obtained by its flow calculation. Summary of the invention

[0004] The purpose of the present invention is to address the problem that the existing technology does not take into account the changing characteristics of distribution network operation data, resulting in large errors in line loss calculation results and poor operation stability of the corresponding distribution network; a distribution network scheduling method based on heterogeneous feature graph model aggregation analysis is proposed, a heterogeneous feature graph model is constructed through the distribution network topology and operation data, and the initial features are determined at the same time, and the initial features are aggregated and analyzed to obtain aggregated features, and then embedded features are calculated based on the aggregated features and the heterogeneous feature graph model, and then the regulated power and distribution network line loss are calculated based on the embedded features, and the regulated power and distribution network line loss are correlated and analyzed to obtain the basis for designing power compensation actions - power compensation factors, and finally, the distribution network executes scheduling actions in response to the scheduling strategy determined according to the power compensation action, thereby significantly improving the stability of the distribution network.

[0005] In a first aspect, a technical solution provided in an embodiment of the present invention is a distribution network dispatching method based on heterogeneous feature graph model aggregation analysis, comprising the following steps: S1, obtain the distribution network operation data and perform normalization processing to obtain the initial data set; S2. Establishing a heterogeneous feature graph model based on the initial data set and the corresponding distribution network topology and determining an initial feature set; S3, based on the heterogeneous feature graph neural network combined with the feature aggregation formula, the initial feature set is aggregated to determine the aggregated features; S4, the embedded features are calculated according to the aggregated features and different nodes in the heterogeneous feature graph model; S5. Calculate the regulated power and the line loss of the distribution network according to the embedded features, and determine the dispatching strategy of the distribution network according to the regulated power and the line loss of the distribution network.

[0006] In this scheme, the acquired distribution network operation data is normalized to obtain an initial data set with a uniform data magnitude, and a heterogeneous feature graph model including grid connection points of various types of power equipment in the distribution network is established based on the initial data set and the corresponding distribution network topology. At the same time, the data in the initial data set is used as the initial feature to clarify the change characteristics of electrical data such as power, voltage, and current in the distribution network. Then, the initial features are aggregated and analyzed to obtain the aggregated features of the comprehensive change characteristics of the corresponding node data, and the embedded features used to calculate the regulated power and the distribution network line loss are calculated based on the aggregated features and the heterogeneous feature graph model. Then, the regulated power and the distribution network line loss are calculated based on the embedded features, and the regulated power and the distribution network line loss are correlated and analyzed to obtain the basis for designing the power compensation action - the power compensation factor. Finally, the distribution network executes the scheduling action in response to the scheduling strategy determined according to the power compensation action, thereby significantly improving the stability of the distribution network.

[0007] Preferably, in S1, the normalization formula corresponding to the normalization process is specifically: Where P norm is the normalized power value, P min is the minimum value of the power data, P max is the maximum value of the power data, P is the power value that needs to be normalized, and U norm is the normalized voltage value, U min is the minimum value of the voltage data, U max is the maximum value of the voltage data, and U is the voltage value that needs to be normalized.

[0008] In this solution, after the corresponding operating data is normalized into data of the same magnitude through a normalization formula, it is convenient for comparative analysis between different power equipment, different time periods, and different types of data, making the changing characteristics of the data easier to capture.

[0009] Preferably, in S2, the specific process of establishing the heterogeneous feature graph model based on the initial data set and the corresponding distribution network topology structure is: S21, determining and marking the nodes corresponding to the power equipment in the distribution network according to the topological structure of the distribution network, and establishing a corresponding relationship between the initial data set and the marked nodes based on the power equipment; Synchronously, extracting data features from the initial data set and marking the data features as initial features; S22, constructing a node topology map according to the topology information of the power equipment in the distribution network, and determining corresponding initial features of the nodes based on the corresponding relationship; S23, integrating the node topology graph and the corresponding initial features of the nodes to obtain a heterogeneous feature graph model; The data features in the initial data set include at least the power generation and voltage of photovoltaic power generation equipment in the distribution network, the power consumption and current consumption of load equipment, and the input power, output power and voltage of the intelligent soft switch.

[0010] In this solution, the corresponding relationship between the initial features and the nodes in the distribution network topology is established, and the trend of the data corresponding to the same node feature is clarified to support the subsequent aggregation analysis.

[0011] Preferably, in S3, the feature aggregation formula is specifically: In the formula, is the PV type aggregate feature of node x after the i-th iteration, a represents the PV type feature, b represents the load type feature, c represents the smart soft switch type feature, N i (x) is the set of neighbor nodes of node x in round i, GRU(g b (x′)) represents the initial characteristic g of the load type of neighbor node x′ b (x′) is aggregated and calculated by GRU, GRU(g c (x′)) represents the initial feature g of the intelligent soft switch type for the neighbor node x′ c (x′) is aggregated and calculated through GRU, represents the feature concatenation operation after bidirectional GRU aggregation, |N i (x)| is the number of neighbor nodes of node x; the GRU is a gated recurrent unit used to process the initial features of neighbor nodes in the heterogeneous feature graph neural network.

[0012] In this solution, after obtaining the aggregate characteristics of the nodes in the distribution network, the node power change, voltage change, and current change corresponding to different data changes of the nodes can be obtained according to the aggregate characteristics.

[0013] Preferably, the aggregated features output by the feature aggregation formula after completing the node feature aggregation in the i-th iteration include at least the photovoltaic type feature corresponding to the node, the load type feature corresponding to the node and the intelligent soft switch type feature corresponding to the node.

[0014] In this solution, the aggregation features include at least the photovoltaic type features corresponding to the node, the load type features corresponding to the node, and the intelligent soft switch type features corresponding to the node, which facilitates the analysis of the impact on the node status when the photovoltaic power generation system, load equipment, and intelligent soft switch corresponding to the node change, and at the same time clarifies the flow of data corresponding to the node status change.

[0015] Preferably, in S4, the calculation formula of the embedded feature is specifically: In the formula, E x represents the embedding feature of node x, g(x) represents the feature of node x, O x represents the set of neighbor nodes related to node x, β x,x Represents the weight coefficient of the feature corresponding to node x, g y (x) represents the characteristics of neighbor node y, β x,y Represents the weight coefficient of the corresponding feature of the neighbor node y.

[0016] Preferably, in the calculation formula of the embedded feature, the feature g(x) of node x at least includes the photovoltaic power feature and photovoltaic voltage feature in the aggregated feature of node x, and the feature g(x) of neighbor node y. y (x) at least includes the load type feature and the intelligent soft switch type feature in the node y aggregation feature.

[0017] In this scheme, the embedded features of node x are calculated through the photovoltaic power features, photovoltaic voltage features of node x, load type features of neighboring nodes of node x, and intelligent soft switch type features. At the same time, the corresponding weight coefficient is set according to the aggregated features, that is, β x,x and β x,y , the priority of the node state change corresponding to the photovoltaic characteristics of the node and the corresponding characteristics of the neighboring nodes is obtained, which supports the design of the distribution network scheduling action.

[0018] Preferably, in S5, the calculation formula for adjusting the power is specifically: P out =f adjust (P in ,Z line); Where P out is the output power of the transmission line, that is, the regulated power; P in is the power feature embedded in the feature of the node on the transmission line; Z line is the impedance of the transmission line; f adjust is the adjustment function; The calculation formula of distribution network line loss is as follows: In the formula, Z j is the impedance of the jth transmission line, I j is the current flowing through the jth transmission line, n is the total number of transmission lines, P loss is the total loss of the transmission line.

[0019] In this solution, the current data is the current calculated based on the embedded features of different nodes on the transmission line, which reflects the current and voltage changes of the corresponding transmission line.

[0020] Preferably, in S5, after calculating the regulating power and the distribution network line loss, it is also necessary to calculate the error value according to the regulating power and error calculation formula. When the error value is greater than or equal to a preset threshold, the initial feature set is updated based on the regulating power and the distribution network line loss and step S3 is executed.

[0021] In this scheme, in order to ensure that the calculated distribution network line loss and regulated power are consistent with the current operating status of the distribution network, an error value is used for judgment, and recalculation is performed when the error value is greater than or equal to a preset threshold, thereby ensuring the accuracy and effectiveness of the distribution network line loss and regulated power.

[0022] Preferably, in S5, the specific process of determining the dispatching strategy of the distribution network according to the regulated power and the line loss of the distribution network is: S51, performing correlation analysis on the regulated power and the line loss of the distribution network to determine a power compensation factor; S52, designing a power compensation action of an intelligent soft switch in a distribution network according to the power compensation factor; S53, organizing the power compensation actions to determine a dispatching strategy for the distribution network.

[0023] In this scheme, the correlation analysis between the regulated power and the line loss of the distribution network can determine the power compensation factors corresponding to different areas and time periods of the distribution network. The power compensation actions of different intelligent soft switches in different time periods can also be designed according to the power compensation factors to maximize the utilization of the overall power resources of the distribution network.

[0024] The present invention has at least the following substantial effects: The present application normalizes the acquired distribution network operation data to obtain an initial data set with a uniform data magnitude, thereby eliminating the difficulty of data comparison and analysis caused by different data magnitudes, and establishes a heterogeneous feature graph model including grid connection points of various types of power equipment in the distribution network based on the initial data set and the corresponding distribution network topology structure, and clarifies the position of the node in the distribution network and the power equipment corresponding to the node, and at the same time uses the data in the initial data set as the initial feature to clarify the change characteristics of electrical data such as power, voltage, and current in the distribution network, and then aggregates and analyzes the initial features to obtain aggregate features of the comprehensive change characteristics of the corresponding node data, and calculates embedded features for calculating the regulated power and the distribution network line loss based on the aggregate features and the heterogeneous feature graph model, clarifies the influence of the corresponding features of different power equipment and the neighboring node features on the state change of the corresponding node, and then calculates the regulated power and the distribution network line loss based on the embedded features, and performs correlation analysis on the regulated power and the distribution network line loss to obtain the basis for designing the power compensation action - the power compensation factor, and finally, the distribution network executes the scheduling action in response to the scheduling strategy determined according to the power compensation action, thereby significantly improving the stability of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0026] Figure 1 It is a flow chart of the distribution network dispatching method based on heterogeneous feature graph model aggregation analysis of the present invention; Figure 2 Schematic diagram of the distribution network topology. DETAILED DESCRIPTION

[0027] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0029] Embodiment 1: like Figure 1 As shown, this embodiment provides a distribution network scheduling method based on heterogeneous feature graph model aggregation analysis, including the following steps: S1. Obtain distribution network operation data and perform normalization processing to obtain an initial data set.

[0030] In one embodiment, the normalization formula corresponding to the normalization process is specifically: Where P norm is the normalized power value, P min is the minimum value of the power data, P max is the maximum value of the power data, P is the power value that needs to be normalized, and U norm is the normalized voltage value, U min is the minimum value of the voltage data, U max is the maximum value of the voltage data, and U is the voltage value that needs to be normalized.

[0031] In this embodiment, firstly, the measurement data of the photovoltaic equipment, the load equipment, and the intelligent soft switch equipment in the distribution network are obtained. These data can be measured manually or automatically according to the facilities of the corresponding distribution network. The measurement data include the power generation power and voltage of the photovoltaic power generation equipment, the power consumption and current of the load equipment, and the input power, output power, and voltage of the intelligent soft switch equipment; secondly, the normalization formula is used to normalize the obtained measurement data to eliminate the influence of different data magnitudes on the line loss calculation. Based on the fact that the normalization formula only involves power and voltage, and the current is fixed under the premise that the voltage and power remain unchanged, the current, voltage, and power can be linked together through the power calculation formula, and then the normalized current is obtained by normalizing the power and voltage. The power calculation formula is as follows: P=U×I; In the formula, P represents power, U represents voltage, and I represents current.

[0032] S2. Establish a heterogeneous feature graph model based on the initial data set and the corresponding distribution network topology and determine an initial feature set.

[0033] S21, determining and marking the nodes corresponding to the power equipment in the distribution network according to the topological structure of the distribution network, and establishing a corresponding relationship between the initial data set and the marked nodes based on the power equipment; Synchronously, extracting data features from the initial data set and marking the data features as initial features; S22, constructing a node topology map according to the topology information of the power equipment in the distribution network, and determining corresponding initial features of the nodes based on the corresponding relationship; S23, integrating the node topology graph and the corresponding initial features of the nodes to obtain a heterogeneous feature graph model; The data features in the initial data set include at least the power generation and voltage of photovoltaic power generation equipment in the distribution network, the power consumption and current consumption of load equipment, and the input power, output power and voltage of the intelligent soft switch.

[0034] In this embodiment, the measurement data obtained in step S1 and the changing characteristics of the measurement data are used as initial features, and a corresponding relationship is established according to the nodes incorporated into the distribution network. Therefore, in addition to being able to represent nodes of various types of equipment in the distribution network, such as photovoltaic equipment, load equipment, and intelligent soft switching equipment, the heterogeneous feature graph model can also represent the corresponding characteristics of different nodes based on the corresponding relationship.

[0035] S3. Based on the heterogeneous feature graph neural network and the feature aggregation formula, the initial feature set is aggregated to determine the aggregated features.

[0036] In one embodiment, the feature aggregation formula is specifically: In the formula, is the PV type aggregate feature of node x after the i-th iteration, a represents the PV type feature, b represents the load type feature, c represents the smart soft switch type feature, N i (x) is the set of neighbor nodes of node x in round i, GRU(g b (x′)) represents the initial characteristic g of the load type of neighbor node x′ b (x′) is aggregated and calculated by GRU, GRU(g c (x′)) represents the initial feature g of the intelligent soft switch type for the neighbor node x′ c (x′) is aggregated and calculated through GRU, represents the feature concatenation operation after bidirectional GRU aggregation, |N i(x)| is the number of neighbor nodes of node x; the GRU is a gated recurrent unit used to process the initial features of neighbor nodes in the heterogeneous feature graph neural network.

[0037] Among them, the aggregation features output by the feature aggregation formula after completing the node feature aggregation in the i-th iteration at least include the photovoltaic type feature corresponding to the node, the load type feature corresponding to the node and the intelligent soft switch type feature corresponding to the node.

[0038] In this embodiment, the calculation process of the aggregated features is essentially an intermediate step of the embedded feature calculation, and is also a manifestation of the message transmission mechanism in this embodiment, that is, data is transmitted through this calculation process. The i in the figure corresponds to the accuracy of the last calculated error value in judging the line loss, which reflects that the present embodiment updates the calculation result according to the real-time state of the distribution network in a cyclic detection manner, and timely and accurately captures the dynamic changes of the power of the distribution network.

[0039] S4. Calculate embedded features according to the aggregated features and different nodes in the heterogeneous feature graph model.

[0040] In one embodiment, the calculation formula of the embedding feature is specifically: In the formula, E x represents the embedding feature of node x, g(x) represents the feature of node x, O x represents the set of neighbor nodes related to node x, β x,x Represents the weight coefficient of the feature corresponding to node x, g y (x) represents the characteristics of neighbor node y, β x,y Represents the weight coefficient of the corresponding feature of the neighbor node y.

[0041] The feature g(x) of node x at least includes the photovoltaic power feature and photovoltaic voltage feature in the aggregated feature of node x, and the feature g(x) of neighbor node y. y (x) at least includes the load type feature and the intelligent soft switch type feature in the node y aggregation feature.

[0042] In this embodiment, combined with step S3, the heterogeneous feature graph neural network first performs aggregation analysis based on the nodes and corresponding features of the nodes in the heterogeneous feature graph model to obtain aggregate features of different nodes, and then performs weighted calculation on the aggregate features of different nodes according to the calculation formula of the embedded features to obtain embedded features, which not only takes into account the impact of corresponding data of different devices, such as photovoltaic equipment, load equipment, and intelligent soft switch equipment on the nodes, but also takes into account the mutual influence between different nodes.

[0043] S5. Calculate the regulated power and the line loss of the distribution network according to the embedded features, and determine the dispatching strategy of the distribution network according to the regulated power and the line loss of the distribution network.

[0044] In one embodiment, the calculation formula for adjusting the power is specifically: P out =f adjust (P in ,Z line ); Where P out is the output power of the transmission line, that is, the regulated power; P in is the power feature embedded in the feature of the node on the transmission line; Z line is the impedance of the transmission line; f adjust is the adjustment function.

[0045] In this embodiment, the intelligent soft switch device can monitor the power and voltage status of neighboring nodes in real time, and the intelligent soft switch device can adjust the input and output power according to the power flow demand of the distribution network to reduce the line loss caused by power fluctuations. Therefore, the intelligent soft switch device is used to calculate and adjust the power. Secondly, the adjustment function f adjust It specifically describes how the intelligent soft switch dynamically adjusts the output power under different power input and line conditions, thereby balancing the power flow in the power grid and reducing line losses.

[0046] In one embodiment, the calculation formula of the power distribution network line loss is specifically: In the formula, Z j is the impedance of the jth transmission line, I j is the current flowing through the jth transmission line, n is the total number of transmission lines, P loss is the total loss of the transmission line.

[0047] In addition, in step S5, after calculating the regulated power and the distribution network line loss, it is also necessary to calculate the error value according to the regulated power and error calculation formula. When the error value is greater than or equal to the preset threshold, the initial feature set is updated based on the regulated power and the distribution network line loss and step S3 is executed.

[0048] In this embodiment, the preset threshold value can be set manually or automatically according to the configuration of the distribution network related equipment. The error value is obtained by comparing the calculated regulated power with the real-time operation data of the distribution network. The specific comparison formula is as follows: Where P measured is the power data in the actual operation data of the distribution network, P calculatedis the calculated regulated power, and Error is the error value; if the error value is greater than or equal to the preset threshold, the power, load, and voltage data in the initial feature set are updated according to the calculated regulated power and the distribution network line loss, and step S3 is executed to re-perform the aggregation analysis; if the error value is less than the preset threshold, the dispatching action of the distribution network is designed according to the regulated power and the distribution network line loss, which is essentially the regulation action of the output power and input power of the intelligent soft switches corresponding to different nodes in the distribution network.

[0049] In one embodiment, the specific process of determining the dispatching strategy of the distribution network according to the regulated power and the line loss of the distribution network is as follows: S51, performing correlation analysis on the regulated power and the line loss of the distribution network to determine a power compensation factor; S52, designing a power compensation action of an intelligent soft switch in a distribution network according to the power compensation factor; S53, organizing the power compensation actions to determine a dispatching strategy for the distribution network.

[0050] In this embodiment, the statistical analysis module in Matlab can be used to perform correlation analysis between the regulated power and the line loss of the distribution network and determine the power compensation factor. After obtaining the power compensation factor, the input power and output power of the intelligent soft switch can be designed according to the power compensation factor to determine the power compensation action, and finally the dispatching strategy of the distribution network can be determined according to the power compensation action.

[0051] To verify the specific effect of this embodiment, a set of historical data of actual application is obtained for verification: Get Figure 2 The historical operation data of the three transmission lines A, B and C of the distribution network shown in the figure are aggregated and analyzed when a large number of photovoltaic devices are connected. Among them, line A is taken as a typical analysis object. The line contains 13 photovoltaic power generation equipment, 20 load equipment (including residential users and commercial users, etc.), and 1 intelligent soft switch.

[0052] Specifically, a time period (9:00 to 16:00) with high photovoltaic power generation output in a certain day is taken for analysis. The output power and line loss of line A are compared and analyzed using the traditional algorithm and the heterogeneous feature graph neural network algorithm proposed in the present invention. The comparison results are shown in Table 1: Table 1. Comparative analysis results of line loss calculated by traditional algorithm and this embodiment Time 9 o'clock 10 o'clock 11 o'clock 12 o'clock 13:00 14:00 15:00 16:00 Traditional algorithm line loss 2.32％ 3.15％ 2.58％ 2.59％ 3.24％ 2.85％ 3.98％ 2.12％ The algorithm line loss in this embodiment 1.98％ 3.02％ 2.33％ 2.21％ 2.88％ 2.52％ 3.64％ 1.78％ According to the comparative analysis results in Table 1, it can be inferred that when photovoltaic power generation equipment outputs a large amount, the algorithm of this embodiment is significantly better than the traditional algorithm and can significantly reduce line losses, especially during peak hours. The algorithm of this embodiment can optimize the power flow between devices through the bidirectional adjustment function of the intelligent soft switch, reduce the loss caused by power fluctuations and multi-directional power flows, and thus improve the operation efficiency of the distribution network. Therefore, this embodiment has a significant line loss optimization effect when multiple photovoltaic devices are connected and the load is complex. In particular, under the premise of not relying on a clear line topology structure, it can achieve more efficient power grid power flow calculation and line loss control through feature aggregation and the adjustment mechanism of the intelligent soft switch.

[0053] This embodiment has at least the following substantial effects: This embodiment normalizes the acquired distribution network operation data to obtain an initial data set with a uniform data magnitude, thereby eliminating the difficulty in data comparison and analysis caused by different data magnitudes, and establishes a heterogeneous feature graph model including grid connection points of various types of power equipment in the distribution network based on the initial data set and the corresponding distribution network topology structure, and clarifies the position of the node in the distribution network and the power equipment corresponding to the node. At the same time, the data in the initial data set is used as the initial feature to clarify the change characteristics of electrical data such as power, voltage, and current in the distribution network, and then the initial features are aggregated and analyzed to obtain the aggregated features of the comprehensive change characteristics of the corresponding node data, and the use is calculated based on the aggregated features and the heterogeneous feature graph model. The embedded features of the regulated power and the line loss of the distribution network are calculated, the influence of the corresponding features of different power equipment and the characteristics of neighboring nodes on the state change of the corresponding nodes is clarified, and then the regulated power and the line loss of the distribution network are calculated according to the embedded features, and the correlation analysis between the regulated power and the line loss of the distribution network is performed to obtain the basis for designing the power compensation action - the power compensation factor. Finally, the distribution network executes the scheduling action in response to the scheduling strategy determined according to the power compensation action, which significantly improves the stability of the distribution network. In addition, this embodiment can also optimize the power flow between the devices through the bidirectional adjustment function of the intelligent soft switch, reduce the loss caused by power fluctuations and multi-directional power flows, and effectively improve the operation efficiency of the distribution network.

[0054] The above specific embodiments are preferred embodiments of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the present specific embodiments. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A distribution network dispatching method based on heterogeneous feature graph model aggregation analysis, characterized in that: The method comprises: S1, obtain the distribution network operation data and perform normalization processing to obtain the initial data set; S2. Establishing a heterogeneous feature graph model based on the initial data set and the corresponding distribution network topology and determining an initial feature set; S3, based on the heterogeneous feature graph neural network combined with the feature aggregation formula, the initial feature set is aggregated to determine the aggregated features; S4, calculating embedded features according to the aggregated features and different nodes in the heterogeneous feature graph model; S5. Calculate the regulated power and the line loss of the distribution network according to the embedded features, and determine the dispatching strategy of the distribution network according to the regulated power and the line loss of the distribution network.

2. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1 is characterized in that: In S1, the normalization formula corresponding to the normalization process is specifically: Where P norm is the normalized power value, P min is the minimum value of the power data, P max is the maximum value of the power data, P is the power value that needs to be normalized, and U norm is the normalized voltage value, U min is the minimum value of the voltage data, U max is the maximum value of the voltage data, and U is the voltage value that needs to be normalized.

3. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1 is characterized in that: In S2, the specific process of establishing a heterogeneous feature graph model based on the initial data set and the corresponding distribution network topology is: S21, determining and marking the nodes corresponding to the power equipment in the distribution network according to the topological structure of the distribution network, and establishing a corresponding relationship between the initial data set and the marked nodes based on the power equipment; Synchronously, extracting data features from the initial data set and marking the data features as initial features; S22, constructing a node topology map according to the topology information of the power equipment in the distribution network, and determining corresponding initial features of the nodes based on the corresponding relationship; S23, integrating the node topology graph and the corresponding initial features of the nodes to obtain a heterogeneous feature graph model; The data features in the initial data set include at least the power generation and voltage of photovoltaic power generation equipment in the distribution network, the power consumption and current consumption of load equipment, and the input power, output power and voltage of the intelligent soft switch.

4. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1 is characterized in that: In S3, the feature aggregation formula is specifically: In the formula, is the PV type aggregate feature of node x after the i-th iteration, a represents the PV type feature, b represents the load type feature, c represents the smart soft switch type feature, N i (x) is the set of neighbor nodes of node x in round i, GRU(g b (x′)) represents the initial characteristic g of the load type of neighbor node x′ b (x′) is aggregated and calculated by GRU, GRU(g c (x′)) represents the initial feature g of the intelligent soft switch type for the neighbor node x′ c (x′) is aggregated and calculated through GRU, represents the feature concatenation operation after bidirectional GRU aggregation, |N i (x)| is the number of neighbor nodes of node x; the GRU is a gated recurrent unit used to process the initial features of neighbor nodes in the heterogeneous feature graph neural network.

5. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 4 is characterized in that: The aggregated features output by the feature aggregation formula after completing the node feature aggregation in the i-th iteration at least include the photovoltaic type feature corresponding to the node, the load type feature corresponding to the node, and the intelligent soft switch type feature corresponding to the node.

6. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1 is characterized in that: In S4, the calculation formula of the embedded feature is specifically: In the formula, E x represents the embedding feature of node x, g(x) represents the feature of node x, O x represents the set of neighbor nodes related to node x, β x,x Represents the weight coefficient of the feature corresponding to node x, g y (x) represents the characteristics of neighbor node y, β x,y Represents the weight coefficient of the corresponding feature of the neighbor node y.

7. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 6 is characterized in that: In the calculation formula of the embedded feature, the feature g(x) of node x at least includes the photovoltaic power feature and photovoltaic voltage feature in the aggregated feature of node x, and the feature g(x) of neighbor node y. y (x) at least includes the load type feature and the intelligent soft switch type feature in the node y aggregation feature.

8. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1 is characterized in that: In S5, the calculation formula for adjusting the power is specifically: P out =f adjust (P in ,Z line ); Where P out is the output power of the transmission line, that is, the regulated power; P in The power characteristics of the nodes on the transmission line are embedded in the characteristics; Z line is the impedance of the transmission line; f adjust is the adjustment function; The calculation formula of distribution network line loss is as follows: In the formula, Z j is the impedance of the jth transmission line, I j is the current flowing through the jth transmission line, n is the total number of transmission lines, P loss is the total loss of the transmission line.

9. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1 is characterized in that: In S5, after calculating the regulated power and the distribution network line loss, it is necessary to calculate the error value according to the regulated power and error calculation formula. When the error value is greater than or equal to the preset threshold, the initial feature set is updated based on the regulated power and the distribution network line loss and step S3 is executed.

10. The distribution network dispatching method based on heterogeneous feature graph model aggregation analysis according to claim 1, characterized in that: In S5, the specific process of determining the dispatching strategy of the distribution network according to the regulated power and the line loss of the distribution network is as follows: S51, performing correlation analysis on the regulated power and the line loss of the distribution network to determine a power compensation factor; S52, designing a power compensation action of an intelligent soft switch in a distribution network according to the power compensation factor; S53, organizing the power compensation actions to determine a dispatching strategy for the distribution network.

Images