CN120834603A - A control method for grid-connected energy storage inverter based on spatiotemporal optimization of complex networks - Google Patents
A control method for grid-connected energy storage inverter based on spatiotemporal optimization of complex networksInfo
- Publication number
- CN120834603A CN120834603A CN202511335563.0A CN202511335563A CN120834603A CN 120834603 A CN120834603 A CN 120834603A CN 202511335563 A CN202511335563 A CN 202511335563A CN 120834603 A CN120834603 A CN 120834603A
- Authority
- CN
- China
- Prior art keywords
- node
- time
- dependent
- power grid
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—ELECTRIC POWER NETWORKS; CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for AC mains or AC distribution networks
- H02J3/38—Arrangements for feeding a single network from two or more generators or sources in parallel; Arrangements for feeding already energised networks from additional generators or sources in parallel
- H02J3/46—Controlling the sharing of generated power between the generators, sources or networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
本发明公开了一种基于复杂网络时空优化的构网型储能逆变器控制方法，包括：在时间轴上定位一历史时间戳，并构建所述历史时间戳的功能依赖图，其中，功能依赖图包含K个表征电网功能单元的节点，以预设时间步长，滑动构建L个历史时间戳的功能依赖图，并基于历史时间戳的顺序排列生成依赖图序列，将所述依赖图序列切割为M个时空依赖样本，基于M个时空依赖样本，对图神经网络模型进行迭代监督训练，获得时空状态预测模型，基于时空状态预测模型，确定构网型储能逆变器的控制指令，本发明将时空依赖度确定为边权重，增强了模型对电网未来状态的预测能力，进而支持构网型储能逆变器生成基于预测的控制指令。
The present invention discloses a control method for a grid-type energy storage inverter based on spatiotemporal optimization of a complex network, comprising: locating a historical timestamp on a time axis, and constructing a functional dependency graph of the historical timestamp, wherein the functional dependency graph includes K nodes representing functional units of a power grid, slidingly constructing functional 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, and iteratively supervising training a graph neural network model based on the M spatiotemporal dependency samples to obtain a spatiotemporal state prediction model, and determining control instructions for the grid-type energy storage inverter based on the spatiotemporal state prediction model. The present invention determines the spatiotemporal dependency as an edge weight, thereby enhancing the model's ability to predict the future state of the power grid, thereby supporting the grid-type energy storage inverter to generate control instructions based on prediction.
Description
Technical Field
The invention relates to the field of energy storage inverter control, in particular to a network construction type energy storage inverter control method based on space-time optimization of a complex network.
Background
With the continuous improvement of the access proportion of new energy sources such as photovoltaic power generation, wind power generation and the like, the fluctuation of the voltage and the frequency of a power grid is aggravated, and the difficulty of adjusting the power balance is increased. The prior art with the document number of CN119298603A discloses an energy storage inverter and a control method thereof, which can reduce the load current and the risk of overcurrent damage of devices by adjusting the output voltage. However, the method relies on local or local running state information when generating the control instruction, and lacks perception of the overall space-time evolution trend of the system, so that the control instruction has a prospective deficiency.
Furthermore, the prior art generally does not take into account their distribution in geospatial relation when integrally modeling the grid function. For example, there may be stronger voltage effects between the photovoltaic units and the load units that are closer, while there is often a significant delay in the power response between the remote units, which may not be indicative of spatial position differences, making it difficult to accurately predict local voltage fluctuations.
Meanwhile, although the state prediction is carried out by adopting a graph neural network in part of schemes, the constructed graph structure depends on fixed electrical topology, the edge weights among nodes are set to be constant or manually, and the dynamic dependency relationship among power grid functional units along with the change of the operation working conditions cannot be reflected. The capturing capability of the system state change trend is limited, and the supporting of the grid-type energy storage inverter is difficult to generate a prospective control instruction with high reliability under a complex space-time condition.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a control method of a network-structured energy storage inverter based on space-time optimization of a complex network, which solves the technical problems in the background art by introducing edge weights constructed based on time dependence and space dependence and constructing a functional dependency graph according to the edge weights.
In order to achieve the above purpose, the invention is realized by the following technical scheme:
a method for controlling a network-structured energy storage inverter based on space-time optimization of a complex network comprises the following steps:
s1, positioning a historical time stamp on a time axis, and constructing a functional dependency graph of the historical time stamp, wherein the functional dependency graph comprises K nodes representing power grid functional units;
s2, sliding to construct a functional dependency graph of L historical time stamps according to a preset time step, and generating a dependency graph sequence based on the sequential arrangement of the historical time stamps;
S3, cutting the dependency graph sequence into M space-time dependency samples;
s4, performing iterative supervision training on the graph neural network model based on M space-time dependent samples to obtain a space-time state prediction model;
S5, determining a control instruction of the grid-built energy storage inverter based on the space-time state prediction model.
In some specific embodiments, constructing the functional dependency graph of the historical timestamp includes:
S1-1, acquiring a functional node set containing K nodes under the historical timestamp;
S1-2, establishing a directed edge among K nodes of the functional node set,
S1-3, determining edge weights based on power grid functional units on the upstream and downstream sides of the directed edges;
s1-4, traversing the functional node set, and repeatedly establishing a directed edge and edge weights thereof until the functional dependency graphs corresponding to the K power grid functional units are established.
In some specific embodiments, obtaining the set of functional nodes including K nodes under the historical timestamp includes:
s1-1-1, defining a target acquisition area on an electronic map, and anchoring K power grid functional units in the target acquisition area;
s1-1-2, acquiring a plurality of running state parameters of K power grid functional units in the same historical time stamp;
s1-1-3, carrying out feature standardization on a plurality of operation state parameters, and constructing operation state features corresponding to the power grid functional units;
S1-1-4, defining K power grid functional units as K nodes of a graph structure, wherein corresponding operation state features are node vectors, and forming a functional node set;
The K nodes of the functional node set are respectively provided with node indexes, and each node index corresponds to a power grid functional unit.
In some specific embodiments, establishing a directed edge between K nodes of the set of functional nodes includes:
S1-2-1, anchoring a target node in a functional node set, and selecting J candidate dependent nodes with candidate dependent relations with the target node to obtain J candidate node pairs, wherein J=K-1;
s1-2-2, judging whether candidate dependent nodes in each candidate node pair belong to effective dependent nodes or not based on a pre-constructed knowledge fact graph;
s1-2-3, if the knowledge fact map belongs to the effective dependent node, extracting a predefined entity relationship from the knowledge fact map as a dependent relationship between the target node and the effective dependent node;
s1-2-4, based on the dependency relationship, establishing a directed edge between the target node and the effective dependency node.
In some specific embodiments, determining whether the candidate dependent nodes belong to valid dependent nodes based on the pre-constructed knowledge-fact graph includes:
S1-2-2-1, anchoring two power grid functional units corresponding to the candidate node pairs;
S1-2-2-2, labeling knowledge facts of two power grid functional units, and pairing to generate an entity pair to be judged;
S1-2-2-3, traversing each knowledge fact triplet in a pre-constructed knowledge fact map, extracting a head entity and a tail entity of the knowledge fact triplet, and pairing to generate a plurality of fact entity pairs;
S1-2-2-4, calculating a plurality of matching degrees of an entity pair to be judged and a plurality of fact entity pairs;
s1-2-2-5, selecting the maximum matching degree from a plurality of matching degrees, and comparing the maximum matching degree with a set threshold value;
S1-2-2-6, if the maximum matching degree is larger than a set threshold value, judging that the candidate dependent node in the candidate node pair belongs to an effective dependent node, and if not, judging that the candidate dependent node belongs to an ineffective dependent node.
In some specific embodiments, determining the edge weight based on the grid functional units downstream on the directed edge includes:
s1-3-1, calculating the spatial dependence of the directed edge;
s1-3-2, calculating the time dependence of the directed edge;
S1-3-3, carrying out weighted fusion on the spatial dependence and the time dependence, constructing a space-time dependence, and determining the space-time dependence as the edge weight.
In some specific embodiments, calculating the spatial dependence of the directed edge includes:
S1-3-1-1, anchoring a target node and an effective dependent node which are arranged on the downstream of the edge,
S1-3-1-2, extracting two corresponding power grid functional units according to node indexes of a target node and an effective dependent node;
s1-3-1-3, acquiring space coordinates of the two power grid functional units in an electronic map;
s1-3-1-4, calculating the space distance between two power grid functional units according to the space coordinates;
s1-3-1-5, inversely proportional the space distance to the space dependence degree between the target node and the effective dependent node;
In some specific embodiments, calculating the time dependence of the directed edge includes:
S1-3-2-1, anchoring the historical timestamp on a time axis;
s1-3-2-2, determining a time window with a fixed length on a time axis by taking a historical time stamp as a reference;
S1-3-2-3, acquiring running state parameters of a target node and an effective dependent node on a plurality of continuous time stamps in a time window, and respectively arranging the running state parameters into a target parameter sequence and a dependent parameter sequence based on the sequence of the plurality of continuous time stamps;
s1-3-2-4, calculating the average value of a target parameter sequence and a dependent parameter sequence;
S1-3-2-5, calculating non-normalized covariance and non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence according to the average value of the target parameter sequence and the dependent parameter sequence;
S1-3-2-6, calculating the time dependence between the target node and the effective dependent node according to the non-normalized covariance and the non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence;
the calculation formula of the time dependence is as follows:
;
Wherein, the The degree of time-dependence is indicated,Representing the i-th target node,Representing the j-th valid dependent node,Representing pearson the coefficient of the,Representing the absolute value of the pearson coefficient; an operational status parameter representing the ith target node at the kth consecutive time stamp, Representing the mean value of the sequence of target parameters,Representing the operational status parameters of the jth active dependent node at the kth consecutive time stamp,Representing the mean value of the dependent parameter sequence,Indicating that the covariance is not normalized,Indicating that the standard deviation product is not normalized.
S1-3-2-7, obtaining absolute values of the Pelson coefficients, and defining the absolute values as the time dependence.
In some specific embodiments thereof, the slicing of the dependency graph sequence into M spatiotemporal dependency samples comprises:
s3-1, setting the length of a preprocessed sample;
s3-2, based on the preprocessed sample length, sequentially intercepting the dependency graph subsequences from the dependency graph sequence;
Wherein the dependency graph subsequence contains a functional dependency graph of consecutive historical timestamps;
S3-3, regarding each dependency graph subsequence, taking the running state parameters of K nodes in the function dependency graph corresponding to the tail end time stamp as target tag values;
S3-4, combining each dependency graph subsequence and the corresponding target label value thereof to form the space-time dependency sample.
In some specific embodiments, determining control instructions for a networked energy storage inverter based on a spatiotemporal state prediction model includes:
S5-1, acquiring operation state parameters of K power grid functional units of a target acquisition area under a current time stamp, and constructing a function dependency graph of the current time stamp based on the operation state parameters;
S5-2, inputting the function dependency graph of the current time stamp into a space-time state prediction model, and outputting predicted state parameters of the target time stamp in K power grid function units;
S5-3, generating control instructions of the grid-built energy storage inverter according to the predicted state parameters of the target time stamps in the K power grid functional units.
The invention provides a control method of a network-structured energy storage inverter based on space-time optimization of a complex network, which has the following beneficial effects:
According to the method, the spatial distance between the downstream target node and the effective dependent node on the directed edge is calculated, and is inversely proportional to the spatial dependence degree between the target node and the effective dependent node, so that the adjacent relation of the power grid functional units on the geographic space can be reflected, the spatial distribution is introduced into the functional dependency graph in the construction process, the model has the perception capability on the geographic layout when the functional dependency relation is identified, and the difference of the influence of the power grid functional units at different spatial positions on the running state parameters can be differentiated.
Furthermore, the invention calculates the time dependence by adopting the absolute value of the pearson coefficient, can capture the dynamic response consistency of the running state parameters of the power grid functional units on the continuous time stamp, can identify the power grid functional unit pair with strong correlation in the time evolution process, and improves the representation of the dynamic behavior.
Furthermore, the invention combines the space dependence degree and the time dependence degree into the space-time dependence degree in a weighting way, and the side weight not only considers the adjacency of the power grid functional units on the geographic space, but also combines the dynamic response consistency of the power grid functional units on the time sequence, so that the functional dependence graph can comprehensively reflect the coupling in the power grid.
Furthermore, the space-time dependency degree is determined as the edge weight, 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, the strength is determined by the spatial adjacency and the time synchronicity, the graph structure can dynamically reflect the comprehensive influence degree of the power grid functional units in a historical period, a higher-quality sequence sample is provided for the graph neural network model, the prediction capability of the model on the future state of the power grid is enhanced, and the grid-structured energy storage inverter is further supported to generate a control instruction based on prediction.
Drawings
FIG. 1 is a schematic flow chart of a method for controlling a grid-built energy storage inverter based on space-time optimization of a complex network;
FIG. 2 is a schematic diagram of a functional dependency graph construction process according to the present invention;
FIG. 3 is a schematic diagram of a determination flow of an active dependent node according to the present invention;
FIG. 4 is a schematic diagram of a calculation flow of the spatial dependence according to the present invention;
Fig. 5 is a schematic diagram of a calculation flow of the time dependence according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Embodiment 1 referring to fig. 1, the present invention provides a method for controlling a network-structured energy storage inverter based on space-time optimization of a complex network, comprising the following steps:
s1, positioning a historical time stamp on a time axis, and constructing a functional dependency graph of the historical time stamp, wherein the functional dependency graph comprises K nodes representing power grid functional units;
in the embodiment, the power grid functional units represent units corresponding to wind, light and diesel storage of a power grid system, and the K power grid functional units comprise photovoltaic power generation units, wind power generation units, diesel power generation units, energy storage units, load cluster units and power grid-connected units.
S2, sliding to construct a functional dependency graph of L historical time stamps according to a preset time step, and generating a dependency graph sequence based on the sequential arrangement of the historical time stamps;
S3, cutting the dependency graph sequence into M space-time dependency samples;
s4, performing iterative supervision training on the graph neural network model based on M space-time dependent samples to obtain a space-time state prediction model;
specifically, the iterative training step of the space-time state prediction model includes:
Sequentially inputting M space-time dependent samples into a graph neural network model;
Outputting a preliminary prediction result of the future running state of the power grid functional unit according to the dependency graph sequence in the input sample by the model;
comparing the preliminary prediction result with the actually collected running state parameters, and calculating a prediction error;
Reversely adjusting internal parameters of the graph neural network model based on the prediction error;
And repeating the process until the prediction result output by the model converges or reaches the preset training precision, and finishing the model training.
S5, determining a control instruction of the grid-built energy storage inverter based on the space-time state prediction model.
In the embodiment, a functional dependency graph representing a power grid functional unit and a dependency relationship thereof under a historical timestamp is constructed, a dependency graph sequence arranged in time sequence is generated, the dependency graph sequence is cut into space-time dependency samples, the space-time dependency graph is used for iterative supervision training of a graph neural network model to generate a space-time state prediction model, and a control instruction of a grid-built energy storage inverter is determined based on the model, so that space-time modeling and prediction driving from a historical running state of the power grid functional unit to future control behaviors are realized, the accuracy of predicting power grid voltage, frequency and power change trend is improved, the control instruction with a power grid supporting function is generated in advance by the grid-built energy storage inverter under a complex running condition, and the active regulation capability of power grid running is enhanced.
Embodiment 2 referring to fig. 2 to 5, the technical solution of embodiment 2 is different from embodiment 1 in that specific execution sub-steps of each application step in embodiment 1 are disclosed.
The specific implementation substeps of the control method step S1 are as follows:
S1-1, acquiring a functional node set containing K nodes under the historical timestamp;
S1-2, establishing a directed edge among K nodes of the functional node set,
S1-3, determining edge weights based on power grid functional units on the upstream and downstream sides of the directed edges;
s1-4, traversing the functional node set, and repeatedly establishing a directed edge and edge weights thereof until the functional dependency graphs corresponding to the K power grid functional units are established.
In this embodiment, the structured functional dependency graph realizes the structural expression of the connection direction and interaction strength of each power grid functional unit under the historical timestamp through the functional node set including K nodes, the directed edges between the nodes and the edge weights determined based on the upstream and downstream power grid functional units.
Further, the substep S1-1 specifically includes:
s1-1-1, defining a target acquisition area on an electronic map, and anchoring K power grid functional units in the target acquisition area;
s1-1-2, acquiring a plurality of running state parameters of K power grid functional units in the same historical time stamp;
In this embodiment, the operation state parameter represents a core operation state parameter of the power grid functional unit under a timestamp, including dimensions such as an environment input, an operation state, and a power response.
Specifically, the operation state parameters of the photovoltaic power generation unit are illumination intensity, assembly temperature and output power;
the running state parameters of the wind power generation unit are wind speed, component rotating speed and output power;
The running state parameters of the diesel power generation unit are fuel reserve, unit oil consumption and output power;
the operation state parameters of the energy storage unit are charge state, terminal voltage and charge and discharge power;
the operation state parameters of the load cluster unit are load type (industrial/commercial/residential), electricity consumption period and load power;
the operation state parameters of the power grid-connected unit are grid-connected point voltage, grid-connected point frequency and electricity price time period (peak/flat/valley);
s1-1-3, carrying out feature standardization on a plurality of operation state parameters, and constructing operation state features corresponding to the power grid functional units;
Specifically, feature standardization means dimension normalization processing is carried out on each running state parameter so as to eliminate numerical value differences among different physical quantities, wherein the feature standardization adopts a Z-score standardization or Min-Max normalization method;
The operation state feature represents dimensionless feature vectors which are obtained after standardized processing and are used for representing the operation state of the power grid functional unit.
S1-1-4, defining K power grid functional units as K nodes of a graph structure, wherein corresponding operation state features are node vectors, and forming a functional node set;
The K nodes of the functional node set are respectively provided with node indexes, and each node index corresponds to a power grid functional unit.
In this embodiment, the constructed functional node set maps K power grid functional units in the target acquisition area to K nodes in the graph structure, and uses the standardized running state characteristics as node vectors of the nodes, so that the running state parameters of the power grid functional units under the same historical timestamp are organized to have node indexes, and a unified data basis is provided for establishing directed edges and determining edge weights between the nodes of the functional node set.
Further, the substep S1-2 specifically includes:
S1-2-1, anchoring a target node in a functional node set, and selecting J candidate dependent nodes with candidate dependent relations with the target node to obtain J candidate node pairs, wherein J=K-1;
Specifically, in the functional node set, the candidate dependent nodes refer to other nodes corresponding to the remaining K-1 power grid functional units except the target node. It may have potential interaction possibilities with the target node in electrical topology, geographic proximity, or historical operational data.
S1-2-2, judging whether candidate dependent nodes in each candidate node pair belong to effective dependent nodes or not based on a pre-constructed knowledge fact graph;
the judging process of the effective dependent node is as follows:
S1-2-2-1, anchoring two power grid functional units corresponding to the candidate node pairs;
S1-2-2-2, labeling knowledge facts of two power grid functional units, and pairing to generate an entity pair to be judged;
S1-2-2-3, traversing each knowledge fact triplet in a pre-constructed knowledge fact map, extracting a head entity and a tail entity of the knowledge fact triplet, and pairing to generate a plurality of fact entity pairs;
S1-2-2-4, calculating a plurality of matching degrees of an entity pair to be judged and a plurality of fact entity pairs;
illustratively, the matching degree is determined by mapping the names of the power grid functional units in the entity pair to be determined and the fact entity pair into pre-trained electric power domain word vectors, and calculating cosine similarity between the corresponding word vectors;
The matching degree of the entity pair to be judged and a certain fact entity pair is the average value of the similarity of the head entity word vector and the tail entity word vector.
S1-2-2-5, selecting the maximum matching degree from a plurality of matching degrees, and comparing the maximum matching degree with a set threshold value;
S1-2-2-6, if the maximum matching degree is larger than a set threshold value, judging that the candidate dependent node in the candidate node pair belongs to an effective dependent node, and if not, judging that the candidate dependent node belongs to an ineffective dependent node.
Specifically, the knowledge fact map is a predefined rule-type knowledge base, and includes a physical connection rule and operation coupling relation between various power grid functional units in the power grid, for example:
(photovoltaic array, supplied to, load Cluster)
(Energy storage System, support, grid frequency)
(Diesel generator, backup Power supply, critical load)
(Wind-driven generator, controlled by wind speed)
Each knowledge fact is in the form of a triplet (head entity, relationship, tail entity) constituting the knowledge fact triplet.
In this embodiment, knowledge fact labeling is performed by anchoring two power grid functional units corresponding to the candidate node pair to generate a to-be-determined entity pair, a knowledge fact triplet in a knowledge fact map is traversed to generate a fact entity pair, the matching degree between the to-be-determined entity pair and the fact entity pair is calculated, the maximum matching degree is selected and compared with a set threshold value, whether the candidate dependent node belongs to an effective dependent node is determined according to a comparison result, and validity confirmation of the dependency relationship between the power grid functional units is completed.
Further, the substep S1-2 specifically further comprises
S1-2-3, if the knowledge fact map belongs to the effective dependent node, extracting a predefined entity relationship from the knowledge fact map as a dependent relationship between the target node and the effective dependent node;
s1-2-4, based on the dependency relationship, establishing a directed edge between the target node and the effective dependency node.
In the embodiment, by anchoring the target node among the K nodes of the functional node set and selecting the candidate dependent nodes, carrying out validity judgment on the candidate dependent nodes by combining a pre-constructed knowledge fact graph, determining a dependency relationship according to the entity relationship in the knowledge fact graph, and establishing a directed edge based on the dependency relationship, the establishment of the connection direction among the power grid functional units is realized.
Further, the substep S1-3 specifically includes:
s1-3-1, calculating the spatial dependence of the directed edge;
s1-3-2, calculating the time dependence of the directed edge;
S1-3-3, carrying out weighted fusion on the spatial dependence and the time dependence, constructing a space-time dependence, and determining the space-time dependence as the edge weight.
In this embodiment, the spatial dependency and the time dependency of the directed edges are calculated respectively, and the spatial dependency and the time dependency are weighted and fused into the edge weights, so that the edge weights synthesize the calculation results of the spatial dependency and the time dependency, and the quantized establishment of the connection strength between the power grid functional units is completed.
Illustratively, the step of calculating the spatial dependence includes:
S1-3-1-1, anchoring a target node and an effective dependent node which are arranged on the downstream of the edge,
S1-3-1-2, extracting two corresponding power grid functional units according to node indexes of a target node and an effective dependent node;
s1-3-1-3, acquiring space coordinates of the two power grid functional units in an electronic map;
s1-3-1-4, calculating the space distance between two power grid functional units according to the space coordinates;
for example, the spatial distance may be calculated based on a euclidean distance between two spatial coordinates.
S1-3-1-5, inversely proportional the space distance to the space dependence degree between the target node and the effective dependent node;
Illustratively, the inverse calculation of the spatial dependence is:
;
Wherein, the The degree of spatial dependence is represented and,Representing the spatial distance between the i-th target node and the j-th valid dependent node,Is the attenuation coefficient of the spatial distance, and in this embodiment, the attenuation coefficient is used to adjust the nonlinear influence intensity of the spatial distance on the spatial dependence. Specifically:
When the attenuation coefficient is The effect of spatial distance on spatial dependence is more pronounced when increased. That is, as the spatial distance increases, the weight drops faster.
When the attenuation coefficient isWhen reduced, the influence of the spatial distance on the spatial dependency is reduced. That is, the weight drops more slowly as the spatial distance increases.
In this embodiment, by anchoring the target node and the effective dependency node on the upstream and downstream of the directed edge, the corresponding power grid functional unit is extracted according to the node index, the spatial coordinates of the power grid functional unit on the electronic map are obtained, the spatial distance is calculated, and the spatial distance is inversely proportional to the spatial dependency, so that the expression of the spatial relationship between the power grid functional units is completed.
Illustratively, the process of calculating the time dependence includes:
S1-3-2-1, anchoring the historical timestamp on a time axis;
s1-3-2-2, determining a time window with a fixed length on a time axis by taking a historical time stamp as a reference;
illustratively, in this embodiment, the time window is The method is characterized in that a time sub-axis with the length T is taken as a time window by backtracking the historical time stamp T for T lengths, wherein the length T is preferably 1 hour.
S1-3-2-3, acquiring running state parameters of a target node and an effective dependent node on a plurality of continuous time stamps in a time window, and respectively arranging the running state parameters into a target parameter sequence and a dependent parameter sequence based on the sequence of the plurality of continuous time stamps;
Illustratively, the sequence of operating state parameters represents the core dynamic variables of the node, such as:
photovoltaic power generation unit→output power;
Energy storage unit- & gt charge and discharge power;
load cluster unit→load power;
The power grid-connected unit- & gt grid-connected comment rate;
of course, the core dynamic variable only needs to reflect the energy interaction behavior of the power grid functional unit, and the corresponding observable dynamic variable can be completely selected based on the actual system demand.
S1-3-2-4, calculating the average value of a target parameter sequence and a dependent parameter sequence;
S1-3-2-5, calculating non-normalized covariance and non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence according to the average value of the target parameter sequence and the dependent parameter sequence;
S1-3-2-6, calculating the time dependence between the target node and the effective dependent node according to the non-normalized covariance and the non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence;
the calculation formula of the time dependence is as follows:
;
Wherein, the The degree of time-dependence is indicated,Representing the i-th target node,Representing the j-th valid dependent node,Representing pearson the coefficient of the,Representing the absolute value of the pearson coefficient; an operational status parameter representing the ith target node at the kth consecutive time stamp, Representing the mean value of the sequence of target parameters,Representing the operational status parameters of the jth active dependent node at the kth consecutive time stamp,Representing the mean value of the dependent parameter sequence,Indicating that the covariance is not normalized,Indicating that the standard deviation product is not normalized.
S1-3-2-7, obtaining absolute values of the Pelson coefficients, and defining the absolute values as the time dependence.
In this embodiment, a fixed-length time window is determined by taking a historical time stamp as a reference, running state parameters of a target node and an effective dependent node on a plurality of continuous time stamps in the time window are obtained, a target parameter sequence and a dependent parameter sequence are constructed, a mean value, a non-normalized covariance and a non-normalized standard deviation of the two sequences are calculated, correlation is calculated based on a pearson coefficient formula, an absolute value is taken as a time dependency, and quantitative representation of a running state parameter change relation between power grid functional units on the continuous time stamps is completed.
In this embodiment, the specific implementation substeps of the control method step S3 are as follows:
s3-1, setting the length of a preprocessed sample;
s3-2, based on the preprocessed sample length, sequentially intercepting the dependency graph subsequences from the dependency graph sequence;
Wherein the dependency graph subsequence contains a functional dependency graph of consecutive historical timestamps;
S3-3, regarding each dependency graph subsequence, taking the running state parameters of K nodes in the function dependency graph corresponding to the tail end time stamp as target tag values;
S3-4, combining each dependency graph subsequence and the corresponding target label value thereof to form the space-time dependency sample.
In this embodiment, by setting the length of the sample to be preprocessed, the dependency graph subsequence of the functional dependency graph including the continuous historical timestamps is sequentially intercepted from the dependency graph sequence, the running state parameters of K nodes in the functional dependency graph corresponding to the timestamps at the tail end of each dependency graph subsequence are used as the target label values, and the dependency graph subsequence and the corresponding target label values are combined, so that the space-time dependency sample applicable to model input is obtained.
In this embodiment, the specific implementation substeps of the control method step S5 are as follows:
S5-1, acquiring operation state parameters of K power grid functional units of a target acquisition area under a current time stamp, and constructing a function dependency graph of the current time stamp based on the operation state parameters;
In this embodiment, the function dependency graph of the current timestamp is constructed according to the same method as the historical timestamp, and includes obtaining a current running state parameter, constructing a function node set, and building directed edges and edge weights based on a knowledge fact graph.
S5-2, inputting the function dependency graph of the current time stamp into a space-time state prediction model, and outputting predicted state parameters of the target time stamp in K power grid function units;
The prediction state parameter refers to an operation state parameter representing a future target time stamp of a power grid functional unit, and the operation state parameter comprises, but is not limited to, output power, terminal voltage, state of charge, grid connection point frequency, environment input quantity and the like. Therefore, the running state parameters of the present embodiment can be classified into "historical running state parameters", "current running state parameters" and "predicted state parameters" according to the time period in which they are located, and their data bodies are consistent, and only the time attributes are different.
S5-3, generating control instructions of the grid-built energy storage inverter according to the predicted state parameters of the target time stamps in the K power grid functional units;
Specifically, the generation of the control instruction includes:
If the comparison prediction state parameters find that the grid-connected point frequency is lower than the set threshold value, generating a control instruction for actively supporting the grid-connected point frequency, so that the energy storage unit enters a discharge state and the inertia response of the power grid is enhanced;
If the comparison of the predicted state parameters finds that the load power is greatly increased and the photovoltaic/wind power output is insufficient, a control instruction for maintaining the voltage stability of the bus is generated, so that the energy storage unit releases active power in advance to provide voltage support;
if the comparison prediction state parameter finds that the electricity price period is about to enter a valley period and the state of charge is lower than a set threshold value, a control instruction for low-cost energy storage is generated, so that the energy storage unit is switched to a charging mode to absorb low-price electric energy;
If the comparison prediction state parameters find that the wind-solar power output is significantly higher than the load demand and the energy storage is not full, generating a control instruction for peak clipping and valley filling, so that the energy storage unit absorbs redundant electric energy to balance the system power;
The control instruction comprises an active/reactive power set value, virtual inertia, a sagging coefficient and an operation mode signal, is dynamically configured according to a control framework of the grid-built inverter, and is issued to the grid-built energy storage inverter to drive the grid-built energy storage inverter to finish operation mode switching before a target time stamp.
In summary, in this embodiment, by constructing a functional dependency graph to express the connection direction and the dependency strength between the power grid functional units, the evolution process of the system state on the continuous time stamp is reflected by using the space-time dependency sample, the operation state parameter of the target time stamp is generated based on the prediction model, and the corresponding control command is generated, so that the generation of the control command is expanded from the operation state of a single time point to the composite representation covering the spatial connection relationship and the time evolution relationship between the power grid functional units, the control behavior is changed from responsive regulation to predictive intervention based on prediction, the prospective regulation and control on the power grid voltage, frequency and power balance is realized, and the active regulation capability of the grid-constructed energy storage inverter in the complex network environment is supported.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center by wired (e.g., infrared, wireless, microwave, etc.).
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application.
Claims (10)
1. A method for controlling a network-structured energy storage inverter based on space-time optimization of a complex network is characterized by comprising the following steps:
positioning a historical time stamp on a time axis, and constructing a functional dependency graph of the historical time stamp, wherein the functional dependency graph comprises K nodes representing power grid functional units;
sliding to construct a functional dependency graph of L historical timestamps according to 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 space-time dependency samples;
performing iterative supervision training on the graph neural network model based on M space-time dependent samples to obtain a space-time state prediction model;
And determining a control instruction of the grid-built energy storage inverter based on the space-time state prediction model.
2. The method for controlling a network-structured energy storage inverter based on complex network space-time optimization of claim 1, wherein constructing the functional dependency graph of the historical time stamp comprises:
Acquiring a functional node set containing K nodes under the historical timestamp;
A directed edge is established between K nodes of the set of functional nodes,
Determining edge weights based on the power grid functional units on the upstream and downstream sides;
Traversing the functional node set, and repeatedly establishing a directed edge and edge weights thereof until the functional dependency graphs corresponding to the K power grid functional units are established.
3. The method for controlling a network-structured energy storage inverter based on complex network space-time optimization according to claim 2, wherein obtaining the set of functional nodes including K nodes under the historical timestamp comprises:
a target acquisition area is defined on the electronic map, and K power grid functional units in the target acquisition area are anchored;
acquiring a plurality of running state parameters of K power grid functional units in the same historical time stamp;
Carrying out feature standardization on a plurality of operation state parameters, and constructing the operation state parameters as operation state features corresponding to the power grid functional units;
Defining K power grid functional units as K nodes of a graph structure, wherein the corresponding running state features are node vectors, and forming a functional node set, wherein node indexes are built on the K nodes of the functional node set, and each node index corresponds to the power grid functional unit.
4. The method for controlling a network-structured energy storage inverter based on complex network space-time optimization according to claim 2, wherein establishing a directed edge between K nodes of the functional node set comprises:
Anchoring a target node in the functional node set, and selecting J candidate dependent nodes with candidate dependent relations with the target node to obtain J candidate node pairs, wherein J=K-1;
for each candidate node pair, judging whether the candidate dependent node belongs to a valid dependent node or not based on a pre-constructed knowledge fact graph;
if the knowledge fact map belongs to the effective dependent nodes, extracting a predefined entity relationship from the knowledge fact map as a dependent relationship between the target node and the effective dependent nodes;
Based on the dependency relationship, a directed edge is established between the target node and the active dependency node.
5. The method for controlling a networked energy storage inverter based on complex network space-time optimization of claim 4, wherein determining whether the candidate dependent nodes belong to valid dependent nodes based on the pre-constructed knowledge-fact graph comprises:
anchoring two power grid functional units corresponding to the candidate node pairs;
carrying out knowledge fact labeling on the two power grid functional units, and pairing to generate an entity pair to be judged;
traversing each knowledge fact triplet in a pre-constructed knowledge fact map, extracting a head entity and a tail entity of the knowledge fact triplet, and pairing to generate a plurality of fact entity pairs;
Calculating a plurality of matching degrees of the entity pair to be judged and a plurality of fact entity pairs;
selecting the maximum matching degree from the plurality of matching degrees, and comparing the maximum matching degree with a set threshold value;
and if the maximum matching degree is greater than the set threshold value, judging that the candidate dependent node in the candidate node pair belongs to the effective dependent node, and if not, judging that the candidate dependent node belongs to the ineffective dependent node.
6. The method for controlling a grid-built energy storage inverter based on complex network space-time optimization of claim 2, wherein determining the edge weight based on the power grid functional units on the upstream and downstream of the directed edge comprises:
calculating the spatial dependence of the directed edge;
Calculating the time dependence of the directed edge;
and carrying out weighted fusion on the spatial dependence and the time dependence, constructing a space-time dependence, and determining the space-time dependence as the edge weight.
7. The method for controlling a networked energy storage inverter based on complex network space-time optimization of claim 6, wherein calculating the spatial dependence of the directed edge comprises:
The target nodes and the valid dependent nodes downstream on the edge are anchored,
Extracting two corresponding power grid functional units according to node indexes of the target node and the effective dependent node;
acquiring space coordinates of the two power grid functional units in an electronic map;
calculating the space distance between the two power grid functional units according to the space coordinates;
And inversely proportional to the spatial dependence between the target node and the effective dependent node.
8. The method for controlling a networked energy storage inverter based on complex network space-time optimization of claim 6, wherein calculating the time dependence of the directed edge comprises:
anchoring the historical timestamp on a timeline;
determining a time window with a fixed length on a time axis by taking the historical time stamp as a reference;
acquiring running state parameters of the target node and the effective dependent node on a plurality of continuous time stamps in a time window, and respectively arranging the running state parameters into a target parameter sequence and a dependent parameter sequence based on the sequence of the continuous time stamps;
Calculating the average value of the target parameter sequence and the dependent parameter sequence;
Calculating the non-normalized covariance and the non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence according to the average value of the target parameter sequence and the dependent parameter sequence;
calculating the time dependence between the target node and the effective dependent node according to the non-normalized covariance and the non-normalized standard deviation of the target parameter sequence and the dependent parameter sequence;
the calculation formula of the time dependence is as follows:
;
Wherein, the The degree of time-dependence is indicated,Representing the i-th target node,Representing the j-th valid dependent node,Representing pearson the coefficient of the,Representing the absolute value of the pearson coefficient; an operational status parameter representing the ith target node at the kth consecutive time stamp, Representing the mean value of the sequence of target parameters,Representing the operational status parameters of the jth active dependent node at the kth consecutive time stamp,Representing the mean value of the dependent parameter sequence,Indicating that the covariance is not normalized,Indicating that the standard deviation product is not normalized;
The absolute value of the pearson coefficient is obtained and defined as the time dependence.
9. The method for controlling a network-structured energy storage inverter based on complex network space-time optimization according to claim 1, wherein the step of cutting the dependency graph sequence into M space-time dependency samples comprises the steps of:
setting the length of a preprocessed sample;
Based on the preprocessed sample length, sequentially intercepting a dependency graph subsequence from the dependency graph sequence;
Wherein the dependency graph subsequence contains a functional dependency graph of consecutive historical timestamps;
For each dependency graph subsequence, taking the running state parameters of K nodes in the function dependency graph corresponding to the tail end time stamp as target tag values;
And combining each dependency graph subsequence and the corresponding target label value thereof to form the space-time dependency sample.
10. The method for controlling a networked energy storage inverter based on complex network space-time optimization of claim 9, wherein determining the control command of the networked energy storage inverter based on the space-time state prediction model comprises:
acquiring running state parameters of K power grid functional units of a target acquisition area under a current time stamp, and constructing a function dependency graph of the current time stamp based on the running state parameters;
Inputting the function dependency graph of the current time stamp into a space-time state prediction model, and outputting predicted state parameters of the target time stamp in K power grid function units;
And generating control instructions of the grid-built energy storage inverter according to the predicted state parameters of the target time stamps in the K power grid functional units.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511335563.0A CN120834603B (en) | 2025-09-18 | 2025-09-18 | A Control Method for Grid-Based Energy Storage Inverters Based on Spatiotemporal Optimization of Complex Networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511335563.0A CN120834603B (en) | 2025-09-18 | 2025-09-18 | A Control Method for Grid-Based Energy Storage Inverters Based on Spatiotemporal Optimization of Complex Networks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN120834603A true CN120834603A (en) | 2025-10-24 |
| CN120834603B CN120834603B (en) | 2025-12-16 |
Family
ID=97401045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202511335563.0A Active CN120834603B (en) | 2025-09-18 | 2025-09-18 | A Control Method for Grid-Based Energy Storage Inverters Based on Spatiotemporal Optimization of Complex Networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120834603B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN121012087A (en) * | 2025-10-28 | 2025-11-25 | 珠海市贡峰新能源开发有限公司 | AI-based prediction-based bidirectional inverter-coordinated converter method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117675615A (en) * | 2023-12-01 | 2024-03-08 | 北京大学深圳研究生院 | A dynamic network traffic control method and system based on pre-training |
| CN118157202A (en) * | 2024-02-20 | 2024-06-07 | 国网湖北省电力有限公司电力科学研究院 | Network-type converter stability control method, system and storage medium |
| CN119169144A (en) * | 2024-11-20 | 2024-12-20 | 国网山东省电力公司信息通信公司 | A method and system for constructing a power information system operation service link topology |
| CN119578648A (en) * | 2024-11-29 | 2025-03-07 | 常州大学 | Ozone concentration prediction method, device, equipment and medium based on spatiotemporal graph convolutional network model |
| CN120320405A (en) * | 2025-06-17 | 2025-07-15 | 华北电力大学（保定） | A method and system for adaptively adjusting inverter grid-following and grid-building control modes |
-
2025
- 2025-09-18 CN CN202511335563.0A patent/CN120834603B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117675615A (en) * | 2023-12-01 | 2024-03-08 | 北京大学深圳研究生院 | A dynamic network traffic control method and system based on pre-training |
| CN118157202A (en) * | 2024-02-20 | 2024-06-07 | 国网湖北省电力有限公司电力科学研究院 | Network-type converter stability control method, system and storage medium |
| CN119169144A (en) * | 2024-11-20 | 2024-12-20 | 国网山东省电力公司信息通信公司 | A method and system for constructing a power information system operation service link topology |
| CN119578648A (en) * | 2024-11-29 | 2025-03-07 | 常州大学 | Ozone concentration prediction method, device, equipment and medium based on spatiotemporal graph convolutional network model |
| CN120320405A (en) * | 2025-06-17 | 2025-07-15 | 华北电力大学（保定） | A method and system for adaptively adjusting inverter grid-following and grid-building control modes |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN121012087A (en) * | 2025-10-28 | 2025-11-25 | 珠海市贡峰新能源开发有限公司 | AI-based prediction-based bidirectional inverter-coordinated converter method and system |
| CN121012087B (en) * | 2025-10-28 | 2026-03-27 | 珠海市贡峰新能源开发有限公司 | Bidirectional inversion cooperative conversion method and system based on AI prediction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN120834603B (en) | 2025-12-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN118411003B (en) | Load control method, system, device and storage medium for multi-class power device | |
| CN116614177B (en) | Optical fiber state multidimensional parameter monitoring system | |
| CN120834603B (en) | A Control Method for Grid-Based Energy Storage Inverters Based on Spatiotemporal Optimization of Complex Networks | |
| CN120377317B (en) | Multi-source new energy-oriented grid-connected power fluctuation active stabilization control method and system | |
| CN118825965A (en) | A load dynamic response modeling method and system integrating multiple working conditions, processes and techniques | |
| CN120728593B (en) | Power load forecasting method based on dynamic correlation of industrial user groups | |
| Nichiforov et al. | Efficient load forecasting model assessment for embedded building energy management systems | |
| CN120763590B (en) | Distribution load prediction method and system | |
| CN118983784A (en) | Distributed photovoltaic cluster control simulation method, system, device and storage medium | |
| Yu et al. | A Method for Multisource Heterogeneous Data Fusion and Modeling in New Power Systems | |
| CN113222216A (en) | Method, device and system for predicting cooling, heating and power loads | |
| Qiuxian et al. | Research on economic distribution strategy of unit combination load based on robust optimization | |
| CN119312115B (en) | A method for short-term load forecasting of buildings | |
| CN120728589B (en) | A Method and System for Load Twin Modeling and Prediction Based on Multi-Source Heterogeneous Feature Fusion | |
| CN121258283B (en) | A method and system for load balancing in industrial parks to promote the consumption of new energy sources. | |
| Zhang et al. | Electric Power Load Prediction based on Temporal Semantic Information and LSTM | |
| Li et al. | Research on the Improvement of Operating Efficiency of Electric Meter Calibration System by Clustering Algorithm | |
| Pan et al. | Predicting of Power Quality Steady State Index Based on Chaotic Theory Using Least Squares Support Vector Machine | |
| CN120875426A (en) | Building energy management system and method suitable for multiple energy sources | |
| CN118644019A (en) | An energy management method and related equipment applied to cloud-edge collaboration | |
| Pan et al. | Calculation method of line loss in low-voltage substation based on K-Means++ and Elman neural network | |
| Zhang | Application Research on Photovoltaic Electricity Plant Output Prediction Based on Random Forest Model | |
| CN120073648A (en) | Short-term power load prediction method and system | |
| Jian et al. | Ultra-Short-term Power Prediction of the Wind Farm Based on Multivariate Data Combination | |
| Tao et al. | Application research of deep learning-based BIM Technology in Intelligent Construction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |