CN115983714A - Static security assessment method and system for edge graph neural network power system - Google Patents
Static security assessment method and system for edge graph neural network power system Download PDFInfo
- Publication number
- CN115983714A CN115983714A CN202310064423.9A CN202310064423A CN115983714A CN 115983714 A CN115983714 A CN 115983714A CN 202310064423 A CN202310064423 A CN 202310064423A CN 115983714 A CN115983714 A CN 115983714A
- Authority
- CN
- China
- Prior art keywords
- power system
- attention
- edge
- network model
- model based
- 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.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a static security assessment method and a static security assessment system for a boundary graph neural network power system, which are used for acquiring parameters of a target power system; and inputting parameters of the target power system into a pre-established edge map attention network model based on the multi-head attention system, and obtaining a static safety evaluation result of the power system through an output result of the edge map attention network model based on the multi-head attention system. The method can simultaneously extract the operation characteristics and the structural characteristics of the power grid, and can adapt to the change of a network topology structure caused by N-1 faults and the uncertainty caused by the fluctuation of new energy; the evaluation calculation speed is high, and the accurate evaluation under a large number of expected accidents can be realized.
Description
Technical Field
The invention belongs to the field of static stability evaluation of power systems, and particularly relates to a static safety evaluation method and system for a boundary graph neural network power system.
Background
The traditional power grid safety early warning field mainly aims at some off-line methods based on model driving. Such as: the early safety early warning concept is based on a power grid safety early warning technology of a diagnosis-consultation mode, the potential safety hazard of a power grid is searched in a multidimensional way in the diagnosis, the safety level of the power grid is determined, comprehensive early warning is made in the diagnosis, and high automation is realized. In the prior art, a power grid safety early warning and decision support system is designed from three levels of time dimension, space dimension and object dimension. According to the prior art, a major power failure defense system is designed according to three defense lines of a power system, the safety early warning of a power grid is the basis of the system, and the three defense lines are mainly used for protection control after a fault and cannot early warn in advance. In the prior art, a power grid safety early warning system is constructed by comprehensively considering static safety problems, transient safety problems, voltage safety problems, relay protection fixed value checking and other problems based on an EMS and a DTS data platform. The online dynamic safety assessment and early warning system (PDSA) of the power system is also technically constructed, and the system can realize various types of online safety and stability analysis such as static stability, transient stability, voltage stability, small interference stability and the like. Some students developed a safety analysis, early warning and control system for a large power grid, and can preliminarily preview, analyze, early warn and pre-control various power grid faults and accidents.
With the continuous development of the smart power grid, on one hand, the scale of the power grid is larger and larger, and the operation mode of the power grid tends to be complicated and close to a stable operation boundary due to the fact that high-proportion new energy and alternating current and direct current are in series-parallel connection; on the other hand, a large number of measurement means and accumulation of multi-space-time scale data also bring new challenges to operation analysis and evaluation of the power grid, and the traditional power grid risk evaluation technology based on the model driving type has the following problems: (1) the key line or the transmission section is an important means for scheduling operators to perform 'dimension reduction monitoring' on the power grid, and in the traditional 'model driving type' power grid safety early warning, the transmission section is made off-line, cannot be updated on line, and is difficult to adapt to the complex and changeable operation mode of the current power grid. (2) The section limit transmission capacity is an important basis for a dispatcher to monitor the section, and the section flow needs to be controlled below the section limit transmission capacity. However, in the traditional model-driven type power grid safety early warning, the operation rule is relatively extensive, the selection of the safety characteristics of the power grid is lacked, the key factors influencing the power grid safety cannot be clearly expressed, and the early warning and pre-control of the power grid according to the key factors are more difficult to perform. (3) The real-time performance is the basic requirement of safety early warning, and the safety early warning of the model-driven power grid is difficult to realize real-time early warning and real-time or advanced early warning due to the limitation of computing capacity.
Therefore, research on data generation and preprocessing technology related to data-driven power grid static risk assessment in an intelligent power grid information physical environment, voltage out-of-limit and line blocking risk early warning and other problems is urgently needed.
Disclosure of Invention
The invention aims to provide a static safety assessment method and a static safety assessment system for a boundary graph neural network power system, which are used for overcoming the problems in the prior art, can simultaneously extract the operation characteristics and the structural characteristics of a power grid, and can adapt to the change of a network topological structure caused by N-1 faults and the uncertainty caused by the fluctuation of new energy; the evaluation calculation speed is high, and the accurate evaluation under massive and multiple expected accidents can be realized.
In order to achieve the purpose, the invention adopts the following technical scheme:
the static safety assessment method for the electric power system of the edge graph neural network comprises the following steps:
acquiring parameters of a target power system;
inputting parameters of a target power system into a pre-established edge map attention network model based on a multi-head attention system, and obtaining a static safety evaluation result of the power system through an output result of the edge map attention network model based on the multi-head attention system;
the edge map attention network model based on the multi-head attention mechanism comprises:
standardization layer: for normalizing the input parameters;
multilayer edge map attention layer: the system is used for extracting and updating the characteristics of the standardized parameters, and a multi-head attention mechanism is introduced into each layer of edge map attention layer;
full connection layer: and the method is used for processing the output result of the attention layer of the edge map to obtain the static safety evaluation result of the power system.
Further, the parameters of the target power system include system topology connections, loads, and generator operating data;
the system topological connection comprises various normal, overhaul, new commissioning branch resistance, reactance, susceptance and connection states;
the load and generator operation data comprise active and reactive power of the load and active and reactive power of the generator in different operation modes;
the operation modes refer to different conventional generator sets, new energy source sets and various tide operation modes of loads.
Further, the edge graph attention network model based on the multi-head attention mechanism is obtained by training through a training sample set, and the training sample set obtaining process is as follows:
based on a simulation model of the power system, randomly setting different power generation or load levels and different faults, and calculating the power flow by a Newton-Raphson method to obtain a label y consisting of a node voltage vector and a branch power vector;
corresponding the load of the power system and the generator operation data to the node characteristic h i Connecting the system topology of the power system with the corresponding edge feature f ij Said node characteristic h i And edge feature f ij Jointly forming input features;
constructing a training sample set as h based on the label y and the input features i ,f j |y k And j belongs to M, k belongs to S, S is the sample number, N is the node number, and M is the edge number.
Further, the training process specifically comprises:
randomly extracting a preset number of samples from a training sample set each time, setting a learning rate, training a model by adopting an Adam method, wherein a loss function is a mean square, and an expression is as follows:
wherein the content of the first and second substances,
calculating an output value for the model;
and training until the model loss function is smaller than a set threshold value, and obtaining the edge map attention network model based on the multi-head attention mechanism.
Further, the normalization is performed by using a z-score method, and the calculation formula is as follows:
wherein x is the initial value of the sample, x μ Is the mean value of the samples, x σ Is the standard deviation of the sample, x * Normalized values for the samples;
the calculation formula of the attention layer of each layer of the edge map is as follows:
h' i ＝Wh i +b
f' ij ＝LeakyReLU(A[h' i ||f ij ||h' j ])
h” i ＝∑ i∈N α ij h' i
wherein h is i 、f ij Respectively representing initial node characteristics, edge characteristics, h " i 、f ij "respectively represents the updated node feature, edge feature, α ij W, b, A, F are learnable weight matrices for attention coefficients;
the calculation expression of the multi-attention mechanism is as follows:
wherein P is the number of attention heads, alpha in,p Is the attention coefficient, h 'obtained by the p-th calculation' n,p The node characteristics after the p-th update.
Further, the obtaining of the static safety assessment result of the power system through the output result of the edge map attention network model based on the multi-head attention mechanism specifically includes:
comparing an output result of the edge map attention network model based on the multi-head attention mechanism with a preset threshold value, and realizing static safety evaluation of the power system according to the comparison result;
the output result of the edge graph attention network model based on the multi-head attention mechanism comprises node voltage and branch power;
and the comparison result is represented by 0 or 1, the node or branch set corresponding to all the variables of 1 is an unstable region, and the node or branch set corresponding to all the variables of 0 is a stable region.
The static safety evaluation system of the electric power system of the edge graph neural network comprises:
a parameter acquisition module: the method comprises the steps of obtaining parameters of a target power system;
an evaluation module: inputting parameters of a target power system into a pre-established edge map attention network model based on a multi-head attention system, and obtaining a static safety evaluation result of the power system through an output result of the edge map attention network model based on the multi-head attention system;
the edge map attention network model based on the multi-head attention mechanism comprises:
standardization layer: for normalizing the input parameters;
multilayer edge map attention layer: the system is used for carrying out feature extraction and updating transformation on the standardized parameters, and a multi-head attention mechanism is introduced into each layer of edge map attention layer;
full connection layer: and processing the result output by the attention layer of the edge map to obtain the static safety evaluation result of the power system.
Further, in the parameter acquiring module, the parameters of the target power system include system topology connection, load and generator operation data;
the system topological connection comprises various normal, overhaul, new commissioning branch resistance, reactance, susceptance and connection states;
the load and generator operation data comprise active and reactive power of the load and active and reactive power of the generator in different operation modes;
the operation modes refer to different conventional generator sets, new energy source sets and various tide operation modes of loads.
Further, in the evaluation module, the edge graph attention network model based on the multi-head attention mechanism is obtained by training through a training sample set, and the training sample set obtaining process is as follows:
based on a simulation model of the power system, randomly setting different power generation or load levels and different faults, and calculating the power flow through a Newton-Raphson method to obtain a label y consisting of a node voltage vector and a branch power vector;
corresponding the load of the power system and the generator operation data to the node characteristics h i Connecting the system topology of the power system with the corresponding edge characteristics f ij Said node characteristic h i And edge feature f ij Jointly forming input features;
constructing a training sample set as h based on the label y and the input features i ,f j |y k I belongs to N, j belongs to M, k belongs to S, S is the number of samples, N is the number of nodes, and M is the number of edges;
the training process specifically comprises the following steps:
randomly extracting a preset number of samples from a training sample set each time, setting a learning rate, training a model by adopting an Adam method, wherein a loss function is a mean square, and an expression is as follows:
wherein, the first and the second end of the pipe are connected with each other,
calculating an output value for the model;
and training until the model loss function is smaller than a set threshold value, and obtaining the edge map attention network model based on the multi-head attention mechanism.
Further, in the evaluation module, a static safety evaluation result of the power system is obtained through an output result of the edge map attention network model based on the multi-head attention mechanism, and specifically:
comparing an output result of the edge map attention network model based on the multi-head attention mechanism with a preset threshold value, and realizing static safety evaluation of the power system according to the comparison result;
the output result of the edge graph attention network model based on the multi-head attention mechanism comprises node voltage and branch power;
and the comparison result is represented by 0 or 1, the node or branch set corresponding to all the variables of 1 is an unstable area, and the node or branch set corresponding to all the variables of 0 is a stable area.
Compared with the prior art, the invention has the following beneficial technical effects:
according to the method, an accurate power grid mathematical model is not needed, mining is carried out only on the basis of simulation data and historical operation data, off-line learning and on-line learning can be carried out simultaneously, and the accuracy of the machine learning power grid static safety assessment method is continuously improved, so that the defect that the traditional N-1 static safety analysis cannot carry out on-line quick and accurate early warning on the system safety is overcome; and the power grid static operation risk early warning is carried out through expected accident analysis, decision support can be provided for power grid static operation risk prevention and dispatching control, the actual operation safety of the power grid is obviously improved, and the method has obvious economic and social benefits.
Drawings
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention, and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a schematic diagram of a static security assessment method for a boundary graph neural network power system according to the present invention;
FIG. 2 is a schematic diagram of an edge graph attention network model principle based on a multi-head attention mechanism according to the present invention;
FIG. 3 is a schematic diagram of a static security assessment system of a boundary graph neural network power system according to the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example one
The invention provides a static security assessment method for a boundary graph neural network power system, which comprises the following steps of:
acquiring parameters of a target power system; the parameters of the target power system comprise system topological connection, load and generator operation data; the system topological connection comprises various normal, overhaul, new commissioning branch resistance, reactance, susceptance and connection states; the load and generator operation data comprise active and reactive power of the load and active and reactive power of the generator in different operation modes;
the method comprises the steps of inputting parameters of a target power system into a pre-established edge graph attention network model based on a multi-head attention system, adding the multi-head attention system, fusing edge and node characteristics better, obtaining a power system static safety evaluation result through an output result of the edge graph attention network model based on the multi-head attention system, specifically, comparing the output result including node voltage and branch power with respective preset threshold values respectively, achieving power system static safety evaluation according to the comparison result, outputting the comparison result by 0 and 1, collecting nodes or lines corresponding to all variables of 0 as stable regions, and collecting nodes or lines corresponding to all variables of 1 as unstable regions.
Referring to fig. 2, the edge graph attention network model based on the multi-head attention mechanism is obtained by training through a training sample set, and the training sample set obtaining process is as follows:
based on a simulation model of the power system, different power generation or load levels and different faults (line topological structures) are randomly set, and the load flow is calculated by a Newton-Raphson method to obtain a label y consisting of a node voltage vector and a branch power vector;
corresponding the load of the power system and the generator operation data to the node characteristics h i Connecting the system topology of the power system with the corresponding edge characteristics f ij Said node characteristic h i And edge feature f ij Jointly forming input features; node characteristic h i Including active and reactive power of the load and active and reactive power of the generator, edge characteristic f ij The circuit comprises a branch resistance, a reactance, a susceptance and a connection state;
training sample construction based on label y and input featuresSet as { h i ,f j |y k J belongs to M, k belongs to S, S is the sample number, N is the node number, and M is the edge number;
when a training sample set is used for training a side graph attention network model based on a multi-head attention mechanism, samples of a preset number are randomly extracted from the training sample set every time, a learning rate is set, an Adam method is adopted for training the model, a loss function is a mean square, and an expression is as follows:
wherein the content of the first and second substances,
calculating an output value for the model;
and training until the model loss function is smaller than a set threshold value, and obtaining the edge map attention network model based on the multi-head attention mechanism.
The edge map attention network model based on the multi-head attention mechanism comprises:
standardization layer: for normalizing the input parameters;
specifically, the normalization is performed by using the z-score method, and the calculation formula is as follows:
wherein x is the initial value of the sample, x μ Is the sample mean, x σ Is the standard deviation of the sample, x * Normalized values for the samples;
multilayer edge map attention layer: the system is used for carrying out feature extraction and updating transformation on the standardized parameters, and a multi-head attention mechanism is introduced into each layer of edge map attention layer;
the calculation formula of the attention layer of each layer of the edge map is as follows:
h i ＝Wh i +b
f' ij ＝LeakyReLU(A[h' i ||f ij ||h' j ])
h” i ＝∑ i∈N α ij h' i
wherein h is i 、f ij Respectively represent initial node characteristics, edge characteristics, h' i 、f ij "respectively represents the updated node feature, edge feature, α ij W, b, A, F are learnable weight matrices for attention coefficients;
the computational expression of the multi-attention mechanism is:
wherein P is the number of attention heads, alpha in,p Is the attention coefficient, h 'obtained by the p-th calculation' n,p The node characteristics after the p-th update.
Full connection layer: the system is used for processing the output result of the attention layer of the edge map to obtain a static safety evaluation result of the power system; the dimension of the output result is (1 XN) | (1 XM) | | which represents the vector splicing operation.
The boundary graph attention network model based on the multi-head attention mechanism does not need mechanism modeling on a power system, the trained model can analyze expected accidents, online application is supported, and the problem of low calculation speed of the traditional method is solved. The method can cope with diversity and uncertainty of operation modes of the power system, particularly avoids a complex modeling process of a mechanism model under the condition of topology change caused by faults, adaptively extracts topological features and operation features, has better generalization capability in static stability evaluation, can fully fuse and extract the features of nodes and edges by introducing a multi-head attention mechanism, and has higher static stability evaluation precision.
Example two
The invention also provides a static safety evaluation system of the boundary graph neural network power system, which comprises the following steps:
a parameter acquisition module: the method comprises the steps of obtaining parameters of a target power system; parameters of the target power system include system topology connections, loads, and generator operating data; the system topological connection comprises various normal, overhaul, new commissioning branch resistance, reactance, susceptance and connection states; the load and generator operation data comprise active and reactive power of the load and active and reactive power of the generator in different operation modes; the operation modes refer to different conventional generator sets, new energy source sets and various tide operation modes of loads.
An evaluation module: inputting parameters of a target power system into a pre-established edge map attention network model based on a multi-head attention system, and obtaining a static safety evaluation result of the power system through an output result of the edge map attention network model based on the multi-head attention system; the method comprises the following specific steps: comparing an output result of the edge map attention network model based on the multi-head attention mechanism with a preset threshold value, and realizing static safety evaluation of the power system according to the comparison result; the output result of the edge graph attention network model based on the multi-head attention mechanism comprises node voltage and branch power; and the comparison result is represented by 0 or 1, the node or branch set corresponding to all the variables of 1 is an unstable region, and the node or branch set corresponding to all the variables of 0 is a stable region.
The method comprises the following steps that a multi-head attention mechanism-based edge map attention network model is obtained by training through a training sample set, and the training sample set is obtained in the following process:
based on a simulation model of the power system, randomly setting different power generation or load levels and different faults, and calculating the power flow through a Newton-Raphson method to obtain a label y consisting of a node voltage vector and a branch power vector;
corresponding the load of the power system and the generator operation data to the node characteristic h i Connecting the system topology of the power system with the corresponding edge feature f ij Said node characteristic h i And edge feature f ij Jointly forming input features;
constructing a training sample set as h based on the label y and the input features i ,f j |y k I belongs to N, j belongs to M, k belongs to S, S is the number of samples, N is the number of nodes, and M is the number of edges;
the training process specifically comprises the following steps:
randomly extracting a preset number of samples from a training sample set each time, setting a learning rate, training a model by adopting an Adam method, wherein a loss function is a mean square, and an expression is as follows:
wherein the content of the first and second substances,
calculating an output value for the model;
and training until the model loss function is smaller than a set threshold value, and obtaining the edge map attention network model based on the multi-head attention mechanism.
The edge map attention network model based on the multi-head attention mechanism comprises:
standardization layer: for normalizing the input parameters;
the normalization is performed by using a z-score method, and the calculation formula is as follows:
wherein x is the initial value of the sample, x μ Is the mean value of the samples, x σ Is the standard deviation of the sample, x * Normalized values for the samples;
multilayer edge map attention layer: the system is used for carrying out feature extraction and updating transformation on the standardized parameters, and a multi-head attention mechanism is introduced into each layer of edge map attention layer;
the calculation formula of each edge map attention layer is as follows:
h i '＝Wh i +b
f ij '＝LeakyReLU(A[h i '||f ij ||h' j ])
h i ”＝∑ i∈N α ij h i '
wherein h is i 、f ij Respectively representing initial node characteristics, edge characteristics, h " i 、f ij "respectively represents the updated node feature, edge feature, alpha ij W, b, A, F are learnable weight matrices for attention coefficients;
the computational expression of the multi-attention mechanism is as follows:
wherein P is the number of attention heads, α in,p Is the attention coefficient, h 'obtained by the p-th calculation' n,p The node characteristics after the p-th update.
Full connection layer: and processing the output result of the attention layer of the edge map to obtain the static safety evaluation result of the power system, wherein the dimension of the output result is (1 XN) | (1 XM) |, and | represents the vector splicing operation.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: although the present invention has been described in detail with reference to the above embodiments, those skilled in the art will appreciate that various changes, modifications and equivalents can be made in the embodiments of the invention without departing from the scope of the invention as defined by the appended claims.
Claims (10)
1. The static safety assessment method of the boundary graph neural network power system is characterized by comprising the following steps:
acquiring parameters of a target power system;
inputting parameters of a target power system into a pre-established edge map attention network model based on a multi-head attention system, and obtaining a static safety evaluation result of the power system through an output result of the edge map attention network model based on the multi-head attention system;
the edge map attention network model based on the multi-head attention mechanism comprises:
standardization layer: for normalizing the input parameters;
multilayer edge map attention layer: the system is used for carrying out feature extraction and updating transformation on the standardized parameters, and a multi-head attention mechanism is introduced into each layer of edge map attention layer;
full connection layer: and the method is used for processing the output result of the attention layer of the edge map to obtain the static safety evaluation result of the power system.
2. The charpy neural network power system static security assessment method of claim 1, wherein the parameters of the target power system include system topology connections, loads, and generator operating data;
the system topological connection comprises various normal, overhaul, new commissioning branch resistance, reactance, susceptance and connection states;
the load and generator operation data comprise active and reactive power of the load and active and reactive power of the generator in different operation modes;
the operation modes refer to different conventional generator sets, new energy source sets and various tide operation modes of loads.
3. The edge graph neural network power system static security assessment method according to claim 2, wherein the edge graph attention network model based on the multi-head attention mechanism is obtained by training through a training sample set, and the training sample set is obtained through the following process:
based on a simulation model of the power system, randomly setting different power generation or load levels and different faults, and calculating the power flow through a Newton-Raphson method to obtain a label y consisting of a node voltage vector and a branch power vector;
corresponding the load of the power system and the generator operation data to the node characteristics h i Connecting the system topology of the power system with the corresponding edge characteristics f ij Said node characteristic h i And edge feature f ij Jointly forming input features;
constructing a training sample set as h based on the label y and the input features i ,f j |y k And j belongs to M, k belongs to S, S is the sample number, N is the node number, and M is the edge number.
4. The method for static security assessment of a boundary graph neural network power system according to claim 3, wherein the training process specifically comprises:
randomly extracting a preset number of samples from a training sample set each time, setting a learning rate, training a model by adopting an Adam method, wherein a loss function is a mean square, and an expression is as follows:
wherein the content of the first and second substances,
calculating an output value for the model;
and training until the model loss function is smaller than a set threshold value, and obtaining an edge graph attention network model based on a multi-head attention mechanism.
5. The method for static security assessment of a sidegram neural network power system as claimed in claim 1, wherein said normalization is by using z-score method, and the calculation formula is:
wherein x is the initial value of the sample, x μ Is a sampleMean value, x σ Is the standard deviation of the sample, x * Normalized values for the samples;
the calculation formula of the attention layer of each layer of the edge map is as follows:
h i '＝Wh i +b
f ij '＝LeakyReLU(A[h i '||f ij ||h' j ])
h i ”＝∑ i∈N α ij h i '
wherein h is i 、f ij Respectively represent initial node characteristics, edge characteristics, h' i 、f ij "respectively represents the updated node feature, edge feature, alpha ij W, b, A, F are learnable weight matrices for attention coefficients;
the computational expression of the multi-attention mechanism is as follows:
wherein P is the number of attention heads, alpha in,p Attention factor, h 'obtained for the p th calculation' n,p The node characteristics after the p-th update.
6. The boundary graph neural network power system static security assessment method according to claim 1, wherein the power system static security assessment result is obtained through an output result of the boundary graph attention network model based on a multi-head attention mechanism, and specifically comprises:
comparing an output result of the edge map attention network model based on the multi-head attention mechanism with a preset threshold value, and realizing static safety evaluation of the power system according to the comparison result;
the output result of the edge graph attention network model based on the multi-head attention mechanism comprises node voltage and branch power;
and the comparison result is represented by 0 or 1, the node or branch set corresponding to all the variables of 1 is an unstable region, and the node or branch set corresponding to all the variables of 0 is a stable region.
7. The static safety assessment system of the boundary graph neural network power system is characterized by comprising:
a parameter acquisition module: the method comprises the steps of obtaining parameters of a target power system;
an evaluation module: inputting parameters of a target power system into a pre-established edge map attention network model based on a multi-head attention system, and obtaining a static safety evaluation result of the power system through an output result of the edge map attention network model based on the multi-head attention system;
the edge map attention network model based on the multi-head attention mechanism comprises:
standardization layer: for normalizing the input parameters;
multilayer edge map attention layer: the system is used for carrying out feature extraction and updating transformation on the standardized parameters, and a multi-head attention mechanism is introduced into each layer of edge map attention layer;
full connection layer: and processing the result output by the attention layer of the edge map to obtain a static safety evaluation result of the power system.
8. The system according to claim 7, wherein in the parameter obtaining module, the parameters of the target power system include system topology connection, load and generator operation data;
the system topological connection comprises various normal, overhaul, new commissioning branch resistance, reactance, susceptance and connection states;
the load and generator operation data comprise active and reactive power of the load and active and reactive power of the generator in different operation modes;
the operation modes refer to different conventional generator sets, new energy source sets and various tide operation modes of loads.
9. The system for evaluating the static security of the edge graph neural network power system according to claim 8, wherein in the evaluation module, an edge graph attention network model based on a multi-head attention mechanism is obtained by training through a training sample set, and the training sample set is obtained by the following steps:
based on a simulation model of the power system, randomly setting different power generation or load levels and different faults, and calculating the power flow through a Newton-Raphson method to obtain a label y consisting of a node voltage vector and a branch power vector;
corresponding the load of the power system and the generator operation data to the node characteristics h i Connecting the system topology of the power system with the corresponding edge feature f ij Said node characteristic h i And edge feature f ij Jointly forming input features;
constructing a training sample set as h based on the label y and the input features i ,f j |y k J belongs to M, k belongs to S, S is the sample number, N is the node number, and M is the edge number;
the training process specifically comprises the following steps:
randomly extracting a preset number of samples from a training sample set each time, setting a learning rate, training a model by adopting an Adam method, wherein a loss function is a mean square, and an expression is as follows:
wherein the content of the first and second substances,
calculating an output value for the model;
and training until the model loss function is smaller than a set threshold value, and obtaining an edge graph attention network model based on a multi-head attention mechanism.
10. The system according to claim 7, wherein the evaluation module obtains the power system static security evaluation result through an output result of the edge graph attention network model based on a multi-head attention mechanism, and specifically comprises:
comparing an output result of the edge map attention network model based on the multi-head attention mechanism with a preset threshold value, and realizing static safety evaluation of the power system according to the comparison result;
the output result of the edge graph attention network model based on the multi-head attention mechanism comprises node voltage and branch power;
and the comparison result is represented by 0 or 1, the node or branch set corresponding to all the variables of 1 is an unstable region, and the node or branch set corresponding to all the variables of 0 is a stable region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310064423.9A CN115983714A (en) | 2023-01-15 | 2023-01-15 | Static security assessment method and system for edge graph neural network power system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310064423.9A CN115983714A (en) | 2023-01-15 | 2023-01-15 | Static security assessment method and system for edge graph neural network power system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115983714A true CN115983714A (en) | 2023-04-18 |
Family
ID=85970277
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310064423.9A Pending CN115983714A (en) | 2023-01-15 | 2023-01-15 | Static security assessment method and system for edge graph neural network power system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115983714A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117640218A (en) * | 2023-12-04 | 2024-03-01 | 北京浩然五洲软件技术有限公司 | Power network safety simulation method and system |
-
2023
- 2023-01-15 CN CN202310064423.9A patent/CN115983714A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117640218A (en) * | 2023-12-04 | 2024-03-01 | 北京浩然五洲软件技术有限公司 | Power network safety simulation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110311376B (en) | Dynamic safety assessment comprehensive model and space-time visualization method for power system | |
| Xie et al. | A review of machine learning applications in power system resilience | |
| Xu et al. | Real-time transient stability assessment model using extreme learning machine | |
| Bahmanyar et al. | Power system voltage stability monitoring using artificial neural networks with a reduced set of inputs | |
| CN104463706B (en) | A kind of method and system that voltage dip business reasons are detected for power network | |
| CN112508442B (en) | Transient stability assessment method and system based on automatic and interpretable machine learning | |
| CN114723285B (en) | Power grid equipment safety evaluation prediction method | |
| CN116245033B (en) | Artificial intelligent driven power system analysis method and intelligent software platform | |
| CN109428327B (en) | Power grid key branch and leading stable mode identification method and system based on response | |
| Jiao | Application and prospect of artificial intelligence in smart grid | |
| LU500551B1 (en) | Virtual load dominant parameter identification method based on incremental learning | |
| Zheng et al. | Real-time transient stability assessment based on deep recurrent neural network | |
| CN112241606A (en) | Cooperative decision-making method for operation and maintenance of ship intelligent equipment based on CPS decision-making module | |
| CN112862065A (en) | Power grid branch parameter identification method based on multitask graph neural network | |
| CN115983714A (en) | Static security assessment method and system for edge graph neural network power system | |
| Zhukov et al. | On-line power systems security assessment using data stream random forest algorithm modification | |
| CN114997566A (en) | Power grid blocking risk assessment method and system considering node connectivity loss | |
| CN113758652B (en) | Oil leakage detection method and device for converter transformer, computer equipment and storage medium | |
| CN112949201A (en) | Wind speed prediction method and device, electronic equipment and storage medium | |
| CN116541748A (en) | Power grid fault diagnosis method and system based on artificial intelligence technology | |
| Olaru et al. | Data mining tools and application in power system engineering | |
| CN116204771A (en) | Power system transient stability key feature selection method, device and product | |
| CN115271070A (en) | Voltage stability margin evaluation method, equipment and medium based on improved Relief-F algorithm | |
| Wang et al. | Design of an Underground Transmission Line Condition Fault Monitoring System for Power Grids Based on Data Analysis Algorithms | |
| CN110956010B (en) | Large-scale new energy access power grid stability identification method based on gradient lifting tree |
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 |