A network construction type energy storage inverter control method based on complex network space-time optimization

Abstract

The application discloses a network construction type energy storage inverter control method based on complex network space-time optimization, comprising the following steps: positioning a historical timestamp on a time axis, and constructing a function dependency graph of the historical timestamp, wherein the function dependency graph contains K nodes representing power grid functional units; slidingly constructing function dependency graphs of L historical timestamps at a preset time step, and generating a dependency graph sequence based on the sequence arrangement of the historical timestamps; cutting the dependency graph sequence into M space-time dependency samples; iteratively supervising and training a graph neural network model based on the M space-time dependency samples, obtaining a space-time state prediction model, and determining a control instruction of the network construction type energy storage inverter based on the space-time state prediction model. The application determines the space-time dependency as an edge weight, enhances the prediction ability of the model on the future state of the power grid, and further supports the network construction type energy storage inverter to generate a prediction-based control instruction.

Description

Technical Field

[0001] This invention relates to the field of energy storage inverter control, specifically to a control method for a grid-type energy storage inverter based on spatiotemporal optimization of complex networks. Background Technology

[0002] With the increasing proportion of new energy sources such as photovoltaic and wind power being integrated into the grid, grid voltage and frequency fluctuations are intensifying, making power balance regulation more difficult. Existing control methods for grid-connected energy storage inverters can generate control commands based on historical or real-time operating data; for example, the prior art (CN119298603A) discloses an energy storage inverter and its control method, which can reduce load current and mitigate the risk of overcurrent damage to devices by adjusting the output voltage. However, such methods rely heavily on local or partial operating status information when generating control commands, lacking awareness of the overall spatiotemporal evolution trend of the system, resulting in insufficient foresight in the control commands.

[0003] Furthermore, existing technologies typically do not consider the geographical distribution of grid functional units when modeling them as a whole. For example, there may be a stronger voltage impact between photovoltaic units and load units that are close to each other, while the power response between units that are far apart often has a significant delay, failing to reflect spatial differences and making it difficult to accurately predict local voltage fluctuations.

[0004] Meanwhile, although some solutions employ graph neural networks for state prediction, the constructed graph structure relies on a fixed electrical topology, and the edge weights between nodes are often set to constants or manually configured, failing to reflect the dynamic dependencies between grid functional units as operating conditions change. This limits the ability to capture trends in system state changes, making it difficult to support grid-based energy storage inverters in generating highly reliable forward-looking control commands under complex spatiotemporal conditions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a control method for a grid-type energy storage inverter based on spatiotemporal optimization of complex networks. This method solves the technical problems mentioned in the background by introducing edge weights based on time and space dependencies and constructing a functional dependency graph accordingly.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A control method for a grid-type energy storage inverter based on spatiotemporal optimization of complex networks includes the following steps:

[0008] S1. Locate a historical timestamp on the timeline and construct a functional dependency graph of the historical timestamp; wherein, the functional dependency graph contains K nodes representing power grid functional units;

[0009] S2. With a preset time step, construct a functional dependency graph with L historical timestamps by sliding, and generate a dependency graph sequence based on the order of the historical timestamps.

[0010] S3. Divide the dependency graph sequence into M spatiotemporal dependency samples;

[0011] S4. Based on M spatiotemporally dependent samples, perform iterative supervised training on the graph neural network model to obtain a spatiotemporal state prediction model.

[0012] S5. Based on the spatiotemporal state prediction model, determine the control commands for the grid-type energy storage inverter.

[0013] In some specific embodiments, constructing the functional dependency graph of the historical timestamps includes:

[0014] S1-1. Obtain the set of functional nodes containing K nodes under the historical timestamp;

[0015] S1-2. Establish directed edges between the K nodes of the functional node set.

[0016] S1-3. Determine edge weights based on the upstream and downstream power grid functional units of directed edges;

[0017] S1-4. Traverse the set of functional nodes, repeatedly establish directed edges and their weights, until the functional dependency graph corresponding to K power grid functional units is constructed.

[0018] In some specific embodiments, obtaining the set of functional nodes containing K nodes under the historical timestamp includes:

[0019] S1-1-1. Delineate the target data collection area on the electronic map and anchor K power grid functional units within the target data collection area;

[0020] S1-1-2. Obtain several operating status parameters of K power grid functional units at the same historical timestamp;

[0021] S1-1-3. Standardize the characteristics of several operating status parameters to construct the operating status characteristics corresponding to the power grid functional units;

[0022] S1-1-4. Define K power grid functional units as K nodes in a graph structure, and their corresponding operating state characteristics as node vectors, forming a functional node set.

[0023] Each of the K nodes in the functional node set has a node index, and each node index corresponds to its power grid functional unit.

[0024] In some specific embodiments, directed edges are established between K nodes of the functional node set, including:

[0025] S1-2-1. In the set of functional nodes, anchor a target node and select J candidate dependent nodes that have a candidate dependency relationship with the target node to obtain J candidate node pairs, where J=K-1;

[0026] S1-2-2. For each candidate node pair, determine whether the candidate dependent nodes are valid dependent nodes based on the pre-constructed knowledge fact graph.

[0027] S1-2-3. If it is a valid dependent node, then extract the predefined entity relationship in the knowledge fact graph as the dependency relationship between the target node and the valid dependent node;

[0028] S1-2-4. Based on the aforementioned dependency relationship, establish a directed edge between the target node and the effective dependent nodes.

[0029] In some specific embodiments, determining whether candidate dependency nodes are valid dependency nodes based on a pre-built knowledge fact graph includes:

[0030] S1-2-2-1, Anchoring the two power grid functional units corresponding to the candidate node pair;

[0031] S1-2-2-2: Label the two power grid functional units with knowledge facts and pair them to generate entity pairs to be judged;

[0032] S1-2-2-3. In the pre-constructed knowledge fact graph, traverse each knowledge fact triplet, extract its head entity and tail entity, and pair them to generate several fact entity pairs.

[0033] S1-2-2-4. Calculate the matching degree between the entity pair to be judged and several fact entity pairs;

[0034] S1-2-2-5. Select the maximum matching degree from several matching degrees and compare it with a set threshold.

[0035] S1-2-2-6. If the maximum matching degree is greater than the set threshold, then the candidate dependent node in the candidate node pair is determined to be a valid dependent node; otherwise, it is an invalid dependent node.

[0036] In some specific embodiments, the edge weights are determined based on the grid functional units upstream and downstream of the directed edge, including:

[0037] S1-3-1. Calculate the spatial dependency of directed edges;

[0038] S1-3-2, Calculate the time dependency of directed edges;

[0039] S1-3-3. The spatial dependency and temporal dependency are weighted and fused to construct the spatiotemporal dependency, and the spatiotemporal dependency is determined as the edge weight.

[0040] In some specific embodiments, calculating the spatial dependency of directed edges includes:

[0041] S1-3-1-1, Anchor the target nodes and effective dependent nodes upstream and downstream of the directed edge.

[0042] S1-3-1-2. Based on the node index of the target node and the effective dependent node, extract the two corresponding power grid functional units.

[0043] S1-3-1-3, Obtain the spatial coordinates of the two power grid functional units on the electronic map;

[0044] S1-3-1-4. Calculate the spatial distance between the two power grid functional units based on the spatial coordinates.

[0045] S1-3-1-5. The spatial distance is inversely proportional to the spatial dependency between the target node and the effective dependent node;

[0046] In some specific embodiments, calculating the time dependency of directed edges includes:

[0047] S1-3-2-1. Anchor the historical timestamps on the timeline;

[0048] S1-3-2-2. Based on historical timestamps, determine a fixed-length time window on the timeline;

[0049] S1-3-2-3. Within the time window, obtain the running status parameters of the target node and the effective dependent nodes at several consecutive timestamps, and arrange the running status parameters into a target parameter sequence and a dependent parameter sequence based on the order of the several consecutive timestamps.

[0050] S1-3-2-4. Calculate the mean of the objective parameter sequence and the dependent parameter sequence;

[0051] S1-3-2-5. Based on the mean of the target parameter sequence and the dependent parameter sequence, calculate the unnormalized covariance and unnormalized standard deviation of the target parameter sequence and the dependent parameter sequence.

[0052] S1-3-2-6. Calculate the time dependence between the target node and the effective dependent node based on the unnormalized covariance and unnormalized standard deviation of the target parameter sequence and the dependent parameter sequence.

[0053] The formula for calculating the time dependence is:

[0054] ;

[0055] in, Indicates time dependence. This represents the i-th target node. This represents the j-th valid dependency node. This represents the Pearson coefficient. This represents the absolute value of the Pearson coefficient; This represents the running status parameters of the i-th target node at the k-th consecutive timestamp. This represents the mean of the target parameter sequence. This represents the running status parameters of the j-th valid dependent node at the k-th consecutive timestamp. This represents the mean of the parameter sequence. This indicates that the variance is not uniformly covariant. This represents the product of unnormalized standard deviations.

[0056] S1-3-2-7. Obtain the absolute value of the Pearson coefficient, which is defined as the time dependence.

[0057] In some specific embodiments, the dependency graph sequence is segmented into M spatiotemporal dependency samples, including:

[0058] S3-1, Set the sample length for preprocessing;

[0059] S3-2. Based on the length of the preprocessed samples, extract the dependency graph subsequences sequentially from the dependency graph sequence;

[0060] The dependency graph subsequence contains a functional dependency graph with consecutive historical timestamps;

[0061] S3-3. For each dependency graph subsequence, take the running status parameters of the K nodes in the functional dependency graph corresponding to its end timestamp as the target label value.

[0062] S3-4. Combine each dependency graph subsequence and its corresponding target label value to form the spatiotemporal dependency sample.

[0063] In some specific embodiments, control commands for the grid-type energy storage inverter are determined based on a spatiotemporal state prediction model, including:

[0064] S5-1. Obtain the operating status parameters of K power grid functional units in the target acquisition area at the current timestamp, and construct a functional dependency graph for the current timestamp based on the operating status parameters.

[0065] S5-2. Input the functional dependency graph of the current timestamp into the spatiotemporal state prediction model, and output the predicted state parameters of the target timestamp in K power grid functional units.

[0066] S5-3. Based on the predicted state parameters of the K grid functional units at the target timestamp, generate control commands for the grid-type energy storage inverter.

[0067] This invention provides a control method for a grid-type energy storage inverter based on spatiotemporal optimization of complex networks, which has the following advantages:

[0068] This invention calculates the spatial distance between upstream and downstream target nodes and effective dependent nodes of directed edges, and inversely proportionals it to the spatial dependence between target nodes and effective dependent nodes. This reflects the proximity relationship of power grid functional units in geographic space. It introduces spatial distribution into the construction process of the functional dependency graph, enabling the model to have the ability to perceive the geographic layout when identifying functional dependencies. This helps to distinguish the differences in the impact of power grid functional units in different spatial locations on operating state parameters.

[0069] Furthermore, this invention uses the absolute value of the Pearson coefficient to calculate the time dependence, which can capture the dynamic response consistency of the operating state parameters of power grid functional units at consecutive time stamps, identify power grid functional unit pairs that have strong correlations in the time evolution process, and improve the characterization of dynamic behavior.

[0070] Furthermore, this invention weights and fuses spatial dependence and temporal dependence into spatiotemporal dependence as edge weights. These edge weights consider both the geographical proximity of power grid functional units and their dynamic response consistency over time, enabling the functional dependence graph to comprehensively reflect the coupling within the power grid.

[0071] Furthermore, this invention defines spatiotemporal dependence as edge weights, so that the directed edges in the functional dependency graph not only express the connection direction but also quantify the relative strength of the dependency relationship. This strength is determined by both spatial proximity and temporal synchronization, enabling the graph structure to dynamically reflect the comprehensive influence between power grid functional units over historical periods. This provides higher-quality sequence samples for the graph neural network model, enhances the model's ability to predict the future state of the power grid, and thus supports the generation of prediction-based control commands by grid-connected energy storage inverters. Attached Figure Description

[0072] Figure 1 This is a flowchart illustrating a control method for a grid-type energy storage inverter based on spatiotemporal optimization of complex networks according to the present invention.

[0073] Figure 2 This is a schematic diagram illustrating the construction process of the functional dependency graph described in this invention;

[0074] Figure 3 This is a schematic diagram of the process for determining valid dependent nodes according to the present invention;

[0075] Figure 4 This is a schematic diagram of the calculation process for the spatial dependence described in this invention;

[0076] Figure 5This is a schematic diagram of the calculation process for the time dependence described in this invention. Detailed Implementation

[0077] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] Example 1: Please refer to Figure 1 This invention provides a control method for a grid-type energy storage inverter based on spatiotemporal optimization of complex networks, comprising the following application steps:

[0079] S1. Locate a historical timestamp on the timeline and construct a functional dependency graph of the historical timestamp; wherein, the functional dependency graph contains K nodes representing power grid functional units;

[0080] In this embodiment, the power grid functional unit represents the unit corresponding to wind, solar, diesel and energy storage in the power grid system; for example, the K power grid functional units include: photovoltaic power generation unit, wind power generation unit, diesel power generation unit, energy storage unit, load cluster unit and power grid connection unit.

[0081] S2. With a preset time step, construct a functional dependency graph with L historical timestamps by sliding, and generate a dependency graph sequence based on the order of the historical timestamps.

[0082] S3. Divide the dependency graph sequence into M spatiotemporal dependency samples;

[0083] S4. Based on M spatiotemporally dependent samples, perform iterative supervised training on the graph neural network model to obtain a spatiotemporal state prediction model.

[0084] Specifically, the iterative training steps of the spatiotemporal state prediction model include:

[0085] Input the M spatiotemporally dependent samples into the graph neural network model in sequence;

[0086] Based on the dependency graph sequence in the input samples, the model outputs preliminary predictions of the future operating status of power grid functional units;

[0087] The preliminary prediction results are compared with the actual collected operating status parameters to calculate the prediction error;

[0088] Adjusting the internal parameters of the graph neural network model based on the prediction error;

[0089] Repeat the above process until the model's prediction results converge or reach the preset training accuracy, thus completing the model training.

[0090] S5. Based on the spatiotemporal state prediction model, determine the control commands for the grid-type energy storage inverter.

[0091] In this embodiment, a functional dependency graph representing the grid functional units and their dependencies under historical timestamps is constructed, generating a time-ordered dependency graph sequence. This sequence is then segmented into spatiotemporal dependency samples for iterative supervised training of the graph neural network model to generate a spatiotemporal state prediction model. Based on this model, control commands for the grid-connected energy storage inverter are determined, realizing spatiotemporal modeling and prediction of the historical operating state of the grid functional units to future control behavior. This improves the accuracy of predicting the trends of grid voltage, frequency, and power changes, supports the grid-connected energy storage inverter in generating control commands with grid support functions in advance under complex operating conditions, and enhances the grid's proactive adjustment capability.

[0092] Example 2: See Figures 2 to 5 The technical solution of this embodiment 2 differs from that of embodiment 1 in that it discloses the specific execution sub-steps of each application step in embodiment 1.

[0093] The specific execution sub-steps of step S1 in the control method are as follows:

[0094] S1-1. Obtain the set of functional nodes containing K nodes under the historical timestamp;

[0095] S1-2. Establish directed edges between the K nodes of the functional node set.

[0096] S1-3. Determine edge weights based on the upstream and downstream power grid functional units of directed edges;

[0097] S1-4. Traverse the set of functional nodes, repeatedly establish directed edges and their weights, until the functional dependency graph corresponding to K power grid functional units is constructed.

[0098] In this embodiment, the constructed functional dependency graph, through a set of functional nodes with K nodes, directed edges between nodes, and edge weights determined based on upstream and downstream power grid functional units, realizes a structured expression of the connection direction and interaction strength of each power grid functional unit under historical timestamps.

[0099] Furthermore, sub-step S1-1 specifically includes:

[0100] S1-1-1. Delineate the target data collection area on the electronic map and anchor K power grid functional units within the target data collection area;

[0101] S1-1-2. Obtain several operating status parameters of K power grid functional units at the same historical timestamp;

[0102] In this embodiment, the operating status parameters represent the core operating status parameters of the power grid functional unit at the timestamp, including dimensions such as environmental input, operating status, and power response.

[0103] Specifically, the operating parameters of a photovoltaic power generation unit are irradiance, module temperature, and output power;

[0104] The operating parameters of a wind power generation unit are wind speed, component rotation speed, and output power;

[0105] The operating parameters of the diesel generator unit are fuel reserves, unit fuel consumption, and output power;

[0106] The operating status parameters of the energy storage unit are state of charge, terminal voltage, and charging / discharging power;

[0107] The operating status parameters of the load cluster unit are load type (industrial / commercial / residential), electricity consumption period, and load power;

[0108] The operating status parameters of the power grid connection unit are the grid connection point voltage, grid connection point frequency, and electricity price period (peak / flat / valley).

[0109] S1-1-3. Standardize the characteristics of several operating status parameters to construct the operating status characteristics corresponding to the power grid functional units;

[0110] Specifically, feature standardization means normalizing the dimensions of each operating state parameter to eliminate numerical differences between different physical quantities; the feature standardization adopts Z-score standardization or Min-Max normalization method.

[0111] The operating status features are dimensionless feature vectors obtained after standardization, used to characterize the operating status of power grid functional units.

[0112] S1-1-4. Define K power grid functional units as K nodes in a graph structure, and their corresponding operating state characteristics as node vectors, forming a functional node set.

[0113] Each of the K nodes in the functional node set has a node index, and each node index corresponds to its power grid functional unit.

[0114] In this embodiment, the constructed functional node set maps K power grid functional units in the target acquisition area to K nodes in a graph structure, and uses standardized operating status features as node vectors for each node. This allows the operating status parameters of each power grid functional unit under the same historical timestamp to be organized into nodes with node indices, providing a unified data foundation for establishing directed edges and determining edge weights between nodes in the functional node set.

[0115] Furthermore, sub-steps S1-2 specifically include:

[0116] S1-2-1. In the set of functional nodes, anchor a target node and select J candidate dependent nodes that have a candidate dependency relationship with the target node to obtain J candidate node pairs, where J=K-1;

[0117] Specifically, in the functional node set, candidate dependent nodes refer to the other nodes corresponding to the remaining K-1 power grid functional units besides the target node. These nodes may have potential interactions with the target node in terms of electrical topology, geographical proximity, or historical operating data.

[0118] S1-2-2. For each candidate node pair, determine whether the candidate dependent nodes are valid dependent nodes based on the pre-constructed knowledge fact graph.

[0119] The process for determining the effective dependent nodes is as follows:

[0120] S1-2-2-1, Anchoring the two power grid functional units corresponding to the candidate node pair;

[0121] S1-2-2-2: Label the two power grid functional units with knowledge facts and pair them to generate entity pairs to be judged;

[0122] S1-2-2-3. In the pre-constructed knowledge fact graph, traverse each knowledge fact triplet, extract its head entity and tail entity, and pair them to generate several fact entity pairs.

[0123] S1-2-2-4. Calculate the matching degree between the entity pair to be judged and several fact entity pairs;

[0124] For example, the matching degree is determined by mapping the names of power grid functional units in the entity pair to be judged and the fact entity pair to pre-trained power domain word vectors, and calculating the cosine similarity between the corresponding word vectors;

[0125] The matching degree between the entity pair to be judged and a certain fact entity pair is the average of the similarity between the word vectors of the head entity and the word vectors of the tail entity.

[0126] S1-2-2-5. Select the maximum matching degree from several matching degrees and compare it with a set threshold.

[0127] S1-2-2-6. If the maximum matching degree is greater than the set threshold, then the candidate dependent node in the candidate node pair is determined to be a valid dependent node; otherwise, it is an invalid dependent node.

[0128] Specifically, the knowledge fact graph is a predefined rule-based knowledge base that contains the physical connection rules and operational coupling relationships between various power grid functional units, for example:

[0129] (Photovoltaic array, power supply, load cluster)

[0130] (Energy storage system, support, grid frequency)

[0131] (Diesel generator, backup power, critical load)

[0132] (Wind turbine, controlled by wind speed)

[0133] Each knowledge fact is in the form of a triple: (head entity, relation, tail entity), which constitutes the knowledge fact triple.

[0134] In this embodiment, by anchoring candidate nodes to the corresponding two power grid functional units, knowledge fact annotation is performed to generate entity pairs to be judged. Knowledge fact triples in the knowledge fact graph are traversed to generate fact entity pairs. The matching degree between the entity pairs to be judged and the fact entity pairs is calculated. The maximum matching degree is selected and compared with a set threshold. Based on the comparison result, it is determined whether the candidate dependent node is a valid dependent node, thus completing the validity confirmation of the dependency relationship between power grid functional units.

[0135] Furthermore, sub-steps S1-2 specifically include...

[0136] S1-2-3. If it is a valid dependent node, then extract the predefined entity relationship in the knowledge fact graph as the dependency relationship between the target node and the valid dependent node;

[0137] S1-2-4. Based on the aforementioned dependency relationship, establish a directed edge between the target node and the effective dependent nodes.

[0138] In this embodiment, by anchoring the target node among K nodes in the functional node set and selecting candidate dependent nodes, the validity of the candidate dependent nodes is determined by combining the pre-constructed knowledge fact graph, and the dependency relationship is determined based on the entity relationship in the knowledge fact graph. Based on the dependency relationship, a directed edge is established to realize the establishment of the connection direction between the power grid functional units.

[0139] Furthermore, sub-steps S1-3 specifically include:

[0140] S1-3-1. Calculate the spatial dependency of directed edges;

[0141] S1-3-2, Calculate the time dependency of directed edges;

[0142] S1-3-3. The spatial dependency and temporal dependency are weighted and fused to construct the spatiotemporal dependency, and the spatiotemporal dependency is determined as the edge weight.

[0143] In this embodiment, the spatial and temporal dependencies of directed edges are calculated separately, and the two are weighted and merged into an edge weight. This allows the edge weight to integrate the calculation results of spatial and temporal dependencies, thereby completing the quantitative establishment of the connection strength between power grid functional units.

[0144] For example, the calculation steps of the spatial dependency include:

[0145] S1-3-1-1, Anchor the target nodes and effective dependent nodes upstream and downstream of the directed edge.

[0146] S1-3-1-2. Based on the node index of the target node and the effective dependent node, extract the two corresponding power grid functional units.

[0147] S1-3-1-3, Obtain the spatial coordinates of the two power grid functional units on the electronic map;

[0148] S1-3-1-4. Calculate the spatial distance between the two power grid functional units based on the spatial coordinates.

[0149] For example, the spatial distance can be calculated based on the Euclidean distance between two spatial coordinates.

[0150] S1-3-1-5. The spatial distance is inversely proportional to the spatial dependency between the target node and the effective dependent node;

[0151] For example, the inverse formula for calculating the spatial dependence is:

[0152] ;

[0153] in, Indicates spatial dependence. This represents the spatial distance between the i-th target node and the j-th effective dependent node. This represents the attenuation coefficient of spatial distance; in this embodiment, the attenuation coefficient is used to adjust the intensity of the nonlinear influence of spatial distance on spatial dependence. Specifically:

[0154] When the attenuation coefficient As the distance increases, the impact of spatial distance on spatial dependence becomes more significant. That is, the weight decreases faster as the spatial distance increases.

[0155] When the attenuation coefficient As the distance decreases, the impact of spatial distance on spatial dependence weakens. That is, as the spatial distance increases, the weight decreases more slowly.

[0156] In this embodiment, by anchoring the target nodes and effective dependent nodes upstream and downstream of the directed edge, the corresponding power grid functional units are extracted according to the node index, their spatial coordinates on the electronic map are obtained, the spatial distance is calculated, and the spatial distance is inversely proportional to the spatial dependency, thus completing the expression of the spatial relationship between power grid functional units.

[0157] For example, the calculation process of the time dependency includes:

[0158] S1-3-2-1. Anchor the historical timestamps on the timeline;

[0159] S1-3-2-2. Based on historical timestamps, determine a fixed-length time window on the timeline;

[0160] For example, in this embodiment, the time window is This means that a time sub-axis of length T is extracted as a time window by going back T lengths from the historical timestamp t; where the length T is preferably 1 hour.

[0161] S1-3-2-3. Within the time window, obtain the running status parameters of the target node and the effective dependent nodes at several consecutive timestamps, and arrange the running status parameters into a target parameter sequence and a dependent parameter sequence based on the order of the several consecutive timestamps.

[0162] For example, the sequence of running state parameters represents the core dynamic variables of a node, such as:

[0163] Photovoltaic power generation unit → Output power;

[0164] Energy storage unit → charging and discharging power;

[0165] Load cluster unit → Load power;

[0166] Power grid connection unit → Grid connection evaluation rate;

[0167] Of course, the core dynamic variables only need to reflect the energy interaction behavior of the power grid functional units. The appropriate observable dynamic variables can be selected based on the actual system requirements.

[0168] S1-3-2-4. Calculate the mean of the objective parameter sequence and the dependent parameter sequence;

[0169] S1-3-2-5. Based on the mean of the target parameter sequence and the dependent parameter sequence, calculate the unnormalized covariance and unnormalized standard deviation of the target parameter sequence and the dependent parameter sequence.

[0170] S1-3-2-6. Calculate the time dependence between the target node and the effective dependent node based on the unnormalized covariance and unnormalized standard deviation of the target parameter sequence and the dependent parameter sequence.

[0171] The formula for calculating the time dependence is:

[0172] ;

[0173] in, Indicates time dependence. This represents the i-th target node. This represents the j-th valid dependency node. This represents the Pearson coefficient. This represents the absolute value of the Pearson coefficient; This represents the running status parameters of the i-th target node at the k-th consecutive timestamp. This represents the mean of the target parameter sequence. This represents the running status parameters of the j-th valid dependent node at the k-th consecutive timestamp. This represents the mean of the parameter sequence. This indicates that the variance is not uniformly covariant. This represents the product of unnormalized standard deviations.

[0174] S1-3-2-7. Obtain the absolute value of the Pearson coefficient, which is defined as the time dependence.

[0175] In this embodiment, a fixed-length time window is determined based on historical timestamps. The operating status parameters of the target node and the effective dependent node at several consecutive timestamps within the time window are obtained and constructed into a target parameter sequence and a dependent parameter sequence. The mean, unnormalized covariance, and unnormalized standard deviation of the two sequences are calculated. The correlation is calculated based on the Pearson coefficient formula, and the absolute value is taken as the time dependence. This completes the quantitative representation of the relationship between the operating status parameters of the power grid functional units at consecutive timestamps.

[0176] In this embodiment, the specific execution sub-steps of step S3 of the control method are as follows:

[0177] S3-1, Set the sample length for preprocessing;

[0178] S3-2. Based on the length of the preprocessed samples, extract the dependency graph subsequences sequentially from the dependency graph sequence;

[0179] The dependency graph subsequence contains a functional dependency graph with consecutive historical timestamps;

[0180] S3-3. For each dependency graph subsequence, take the running status parameters of the K nodes in the functional dependency graph corresponding to its end timestamp as the target label value.

[0181] S3-4. Combine each dependency graph subsequence and its corresponding target label value to form the spatiotemporal dependency sample.

[0182] In this embodiment, by setting the sample length of the preprocessing, the dependency graph subsequence containing consecutive historical timestamps of the functional dependency graph is extracted sequentially from the dependency graph sequence. The running state parameters of K nodes in the functional dependency graph corresponding to the end timestamp of each dependency graph subsequence are used as target label values. The dependency graph subsequence is combined with its corresponding target label values ​​to obtain spatiotemporal dependency samples that can be used for model input.

[0183] In this embodiment, the specific execution sub-steps of step S5 of the control method are as follows:

[0184] S5-1. Obtain the operating status parameters of K power grid functional units in the target acquisition area at the current timestamp, and construct a functional dependency graph for the current timestamp based on the operating status parameters.

[0185] In this embodiment, the functional dependency graph of the current timestamp is constructed in the same way as the historical timestamp, including: obtaining the current running state parameters, constructing a set of functional nodes, and establishing directed edges and edge weights based on the knowledge fact graph.

[0186] S5-2. Input the functional dependency graph of the current timestamp into the spatiotemporal state prediction model, and output the predicted state parameters of the target timestamp in K power grid functional units.

[0187] The predicted state parameters described in this invention refer to the operating state parameters characterizing the functional units of the power grid at a future target time stamp, including but not limited to: output power, terminal voltage, state of charge, grid connection point frequency, environmental input, etc. Therefore, the operating state parameters in this embodiment can be divided into "historical operating state parameters," "current operating state parameters," and "predicted state parameters" according to their time stage. Their data content is consistent, only their time attributes differ.

[0188] S5-3. Generate control commands for the grid-type energy storage inverter based on the predicted state parameters of the K grid functional units at the target timestamp.

[0189] Specifically, the generation of the control commands includes:

[0190] If the comparison of the predicted state parameters reveals that the grid connection point frequency is lower than the set threshold, then a control command is generated to actively support the grid connection point frequency, causing the energy storage unit to enter the discharge state and enhance the grid inertial response.

[0191] If comparing the predicted state parameters reveals that the load power will increase significantly and the output of photovoltaic / wind power is insufficient, then a control command is generated to maintain the stability of the bus voltage, causing the energy storage unit to release active power in advance to provide voltage support.

[0192] If comparing the predicted state parameters reveals that the electricity price period is about to enter the "valley period" and the state of charge is lower than the set threshold, then a control command for low-cost energy storage is generated, causing the energy storage unit to switch to charging mode to absorb low-priced electricity.

[0193] If comparing the predicted state parameters reveals that the wind and solar power output is significantly higher than the load demand and the energy storage is not fully utilized, then control commands for peak shaving and valley filling are generated, so that the energy storage unit absorbs excess power to balance the system power.

[0194] The aforementioned control commands include active / reactive power setpoints, virtual inertia, droop coefficient, and operating mode signals. These are dynamically configured according to the control architecture of the grid-type inverter and sent to the grid-type energy storage inverter to drive it to complete the operating mode switch before the target timestamp.

[0195] In summary, this embodiment constructs a functional dependency graph to express the connection direction and dependency strength between power grid functional units, uses spatiotemporal dependency samples to reflect the evolution of the system state over continuous timestamps, generates operating state parameters for the target timestamp based on a prediction model, and generates corresponding control commands. This expands the basis for generating control commands from the operating state at a single point in time to a composite representation encompassing the spatial connection relationship and temporal evolution relationship between power grid functional units. The control behavior changes from responsive regulation to prediction-based early intervention, achieving forward-looking regulation of power grid voltage, frequency, and power balance, and supporting the proactive adjustment capability of grid-type energy storage inverters in complex network environments.

[0196] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means.

[0197] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A network-constructed energy storage inverter control method based on complex network space-time optimization, characterized in that, The method comprises the following steps: Positioning a historical timestamp on a time axis and constructing a function dependency graph of the historical timestamp; wherein the function dependency graph contains K nodes representing power grid functional units; Slidingly constructing function dependency graphs of L historical timestamps with a preset time step, and generating a dependency graph sequence based on the sequential arrangement of the historical timestamps; Cutting the dependency graph sequence into M spatiotemporal dependency samples; Iteratively supervising and training a graph neural network model based on the M spatiotemporal dependency samples to obtain a spatiotemporal state prediction model; Determining a control instruction of a grid-forming energy storage inverter based on the spatiotemporal state prediction model; The step of constructing the function dependency graph of the historical timestamp comprises: Obtaining a function node set containing K nodes under the historical timestamp; Establishing a directed edge between the K nodes of the function node set, Determining an edge weight based on the upstream and downstream power grid functional units of the directed edge; Iterating through the function node set to repeatedly establish the directed edge and its edge weight until the function dependency graph corresponding to the K power grid functional units is constructed; The step of determining the control instruction of the grid-forming energy storage inverter based on the spatiotemporal state prediction model comprises: Obtaining the operating state parameters of the K power grid functional units in the target collection area at the current timestamp, and constructing a function dependency graph of the current timestamp based on the operating state parameters; Inputting the function dependency graph of the current timestamp into the spatiotemporal state prediction model to output the predicted state parameters of the K power grid functional units at the target timestamp; Generating the control instruction of the grid-forming energy storage inverter according to the predicted state parameters of the K power grid functional units at the target timestamp.

2. The complex network space-time optimization-based network-structured energy storage inverter control method according to claim 1, characterized in that, The step of obtaining the function node set containing K nodes under the historical timestamp comprises: Delimiting the target collection area on an electronic map and anchoring the K power grid functional units within the target collection area; Obtaining a plurality of operating state parameters of the K power grid functional units at the same historical timestamp; Performing feature standardization on the plurality of operating state parameters to construct the operating state features corresponding to the power grid functional units; Defining the K power grid functional units as K nodes of a graph structure, and the corresponding operating state features as node vectors to construct a function node set; wherein the K nodes of the function node set are all established with node indexes, and each node index corresponds to its power grid functional unit.

3. The complex network space-time optimization-based network-structured energy storage inverter control method according to claim 2, characterized in that, The step of establishing a directed edge between the K nodes of the function node set comprises: In the function node set, anchoring a target node and selecting J candidate dependency nodes having candidate dependency relationships with the target node to obtain J candidate node pairs, wherein J=K-1; For each candidate node pair, determining whether the candidate dependency nodes thereof belong to valid dependency nodes based on a pre-constructed knowledge fact graph; If they belong to valid dependency nodes, extracting a predefined entity relationship in the knowledge fact graph as the dependency relationship between the target node and the valid dependency nodes; Based on the dependency relationship, establishing a directed edge between the target node and the valid dependency nodes.

4. The complex network space-time optimization-based network-structured energy storage inverter control method according to claim 3, characterized in that, The step of determining whether the candidate dependency nodes belong to valid dependency nodes based on the pre-constructed knowledge fact graph comprises: Anchoring the two power grid functional units corresponding to the candidate node pair; Performing knowledge fact annotation on the two power grid functional units to pair generate a to-be-determined entity pair; In the pre-built knowledge fact graph, traverse each knowledge fact triple, extract its head entity and tail entity, and pair to generate a plurality of fact entity pairs; Calculate the matching degrees of the to-be-determined entity pair and the plurality of fact entity pairs; Select the maximum matching degree from the plurality of matching degrees and compare it with the set threshold value; If the maximum matching degree is greater than the set threshold value, the candidate dependent node in the candidate node pair is determined to be a valid dependent node; otherwise, it is an invalid dependent node.

5. The complex network space-time optimization-based network-constructed energy storage inverter control method according to claim 4, characterized in that, Based on the upstream and downstream power grid functional units of the directed edge, the edge weight is determined, including: Calculate the spatial dependency of the directed edge; Calculate the time dependency of the directed edge; Weighted fusion of spatial dependency and time dependency to construct spatio-temporal dependency, and determine the spatio-temporal dependency as the edge weight.

6. The complex network space-time optimization-based network-structured energy storage inverter control method according to claim 5, characterized in that, The spatial dependency of the directed edge includes: Anchoring the target node and the valid dependent node upstream and downstream of the directed edge, According to the node index of the target node and the valid dependent node, extract the corresponding two power grid functional units; Obtain the spatial coordinates of the two power grid functional units in the electronic map; According to the spatial coordinates, calculate the spatial distance of the two power grid functional units; The spatial distance is inversely proportional to the spatial dependency between the target node and the valid dependent node.

7. The complex network space-time optimization-based network-constructed energy storage inverter control method according to claim 6, characterized in that, The time dependency of the directed edge includes: Anchoring the historical timestamp on the time axis; Determine a fixed length time window on the time axis based on the historical timestamp; In the time window, obtain the running state parameters of the target node and the valid dependent node at a plurality of continuous time stamps, and arrange the running state parameters based on the order of the plurality of continuous time stamps into target parameter sequence and dependent parameter sequence respectively; Calculate the mean of the target parameter sequence and the dependent parameter sequence; According to the mean of the target parameter sequence and the dependent parameter sequence, calculate the non-normalized covariance and non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence; According to the non-normalized covariance and non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence, calculate the time dependency between the target node and the valid dependent node; The calculation formula of the time dependency is: ； wherein, denotes a time dependency, denotes the i-th target node, denotes the j-th valid dependency node, denotes a Pearson coefficient, denotes an absolute value of a Pearson coefficient; denotes a running state parameter of the i-th target node at the k-th consecutive time stamp, denotes a mean of a target parameter sequence, denotes a running state parameter of the j-th valid dependency node at the k-th consecutive time stamp, denotes a mean of a dependency parameter sequence, denotes an unnormalized covariance, denotes an unnormalized standard deviation product; The absolute value of the Pearson coefficient is obtained, which is defined as the time dependency.

8. The complex network space-time optimization-based network-constructed energy storage inverter control method according to claim 7, characterized in that, Cutting the dependency graph sequence into M spatio-temporal dependency samples includes: Set the preprocessed sample length; Based on the preprocessed sample length, sequentially cut the dependency graph subsequence from the dependency graph sequence; Wherein, the dependency graph subsequence contains the functional dependency graph of the continuous historical time stamp; For each dependency graph subsequence, the running state parameters of the K nodes in the functional dependency graph corresponding to the end time stamp are taken as the target label value; Combine each dependency graph subsequence and its corresponding target label value to form the spatio-temporal dependency sample.

Images