CN114943430A - Smart grid key node identification method based on deep reinforcement learning - Google Patents
Smart grid key node identification method based on deep reinforcement learning Download PDFInfo
- Publication number
- CN114943430A CN114943430A CN202210518642.5A CN202210518642A CN114943430A CN 114943430 A CN114943430 A CN 114943430A CN 202210518642 A CN202210518642 A CN 202210518642A CN 114943430 A CN114943430 A CN 114943430A
- Authority
- CN
- China
- Prior art keywords
- network
- reinforcement learning
- model
- power network
- nodes
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- 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
Abstract
The invention provides a smart grid key node identification method based on deep reinforcement learning, which comprises the following steps: establishing a topological structure of a power grid and a communication network according to power grid data, constructing a real power network by combining service attributes and physical characteristics of the power network, and constructing and generating a network model; constructing a full-order sub-graph feature classification model based on deep learning according to network features of a real power network and a generation network; outputting the reward value corresponding to each action through deep full connection learning according to the full-order sub-graph feature classification model; constructing a power network deep reinforcement learning model by using the obtained actions and the corresponding reward values; inputting the power network to be identified to the trained power network deep reinforcement learning model, outputting index values of all nodes, and determining key nodes according to the index values. The invention constructs a deep reinforcement learning-based power network key node identification model, and key nodes of different types of power networks can be accurately identified through the model.
Description
Technical Field
The invention relates to the technical field of smart grids, in particular to a smart grid key node identification method based on deep reinforcement learning.
Background
At present, a lot of researches are carried out on the identification of key nodes of a power grid by academic communities, and the main methods comprise the following steps: based on a graph theory and a complex network theory analysis method, the method mainly analyzes from the structure of a power grid and searches key nodes; the method is based on a detection method of the service attribute of the power grid, and the method analyzes and detects key nodes based on the operation parameters of the power grid, such as power flow distribution, reactance and power change of the power grid; in addition, there are also literature that combines the two approaches for research.
Although the methods obtain better key node detection effect on some networks, certain research results are also obtained. However, with the gradual development of large-scale and wide interconnection of power grids, the random and fluctuating new energy permeability is gradually improved, so that the power system is increasingly complex, and the fault risk is also increased continuously. This presents new challenges to grid key node identification: (1) modern power grids are larger and larger in scale and more complex in structure, and higher requirements are provided for the running speed and accuracy of the existing method; (2) the influence of various randomness and uncertainty factors is more complicated, such as diversified, intelligent and plug-and-play load random changes, which causes dynamic changes of a network structure, and key nodes are changed accordingly. Most of the existing methods are directed at a certain type of static network, and the dynamic change of a network structure is not considered. (3) The existing method lacks early warning on cascade reaction caused by accidents, if a certain fragile node in a power grid fails and is not controlled, the influence of the failure is continuously amplified by the connection between the power grids, cascade failure is generated, and large-area power failure of a regional power grid is caused. The above problem inevitably affects the accuracy and efficiency of the detection of the key nodes of the power grid, is a common problem of the detection of the key nodes of the power grid, and is a bottleneck of application and implementation of the key nodes of the power grid.
Therefore, the invention provides a smart grid key node identification method based on deep reinforcement learning.
Disclosure of Invention
In order to solve the problems, the invention provides a smart grid key node identification method based on deep reinforcement learning.
The invention provides the following technical scheme.
A smart grid key node identification method based on deep reinforcement learning comprises the following steps:
establishing a topological structure of a power grid and a communication network according to power grid data, constructing a real power network by combining service attributes and physical characteristics of the power network, and constructing and generating a network model;
constructing a full-order sub-graph feature classification model based on deep learning according to network features of a real power network and a generation network;
outputting the reward value corresponding to each action through deep full connection learning according to the full-order sub-graph feature classification model;
constructing a power network deep reinforcement learning model by using the obtained actions and the corresponding reward values;
inputting the power network to be identified to the trained power network deep reinforcement learning model, outputting index values of all nodes, and determining key nodes according to the index values.
Preferably, the construction of the full-order sub-graph feature classification model based on deep learning includes the following steps:
according to the topological structure of the network, a node2vec method is adopted to map the network to a low-dimensional vector space;
selecting a sub-graph node corresponding to the target node according to a diffusion rule, and acquiring a node set with a specific sequence according to the migration distance between nodes;
and constructing full-order subgraph features by using the subgraph adjacency matrix and the node vector matrix learned by the node2vec and a katz matrix based on global path similarity, wherein the full-order subgraph features respectively represent low-order, medium-order and high-order structural characteristics of nodes in the subgraph.
Preferably, the method further comprises the following steps: the power network deep reinforcement learning model is trained according to a plurality of index values of the nodes in an online and offline combined mode through a real power network and a generation network.
Preferably, the online learning input is a real power network, and the index value of the online learning input comprises characteristic information of topological structure characteristics, physical attributes and business attributes of the network.
Preferably, the offline learning comprises the following steps:
inputting a training network in a data set, adopting a removal strategy for each node to obtain a current state, and calculating actions and rewards by using a full-order sub-graph feature classification model; the data set adopts the established generating network model to generate a large number of power networks with topological structure characteristics and physical attributes similar to those of a real network;
the training network outputs the index value of each node through the learning of the deep reinforcement learning model, and the key node is judged according to the index value.
Preferably, the method further comprises the following steps:
attacking the structural characteristic attributes of the power network, wherein the attacking nodes comprise key nodes, intermediate nodes, large-scale nodes and random nodes;
and constructing a power network robustness evaluation model, measuring the robustness of the power network through the model, and giving early warning prompts of different levels through the quantitative evaluation value.
The invention has the beneficial effects that:
(1) innovation method for identifying key nodes of power network
The deep reinforcement learning technology is applied to the key node identification of the large-scale interconnected network for the first time, and the problems that the traditional method is not suitable for the large-scale interconnected power grid and the key node identification efficiency is low are solved.
(2) Power grid robustness analysis method innovation
A plurality of attack means such as key nodes, intermediate nodes, large-scale nodes and random nodes are adopted, the robustness of the power network is evaluated from different dimensions such as a topological structure, physical attributes and property attributes, and early warning information of different levels is given according to the quantitative evaluation value.
Drawings
FIG. 1 is a scheme flow chart of a smart grid key node identification method based on deep reinforcement learning according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a full-order sub-graph feature classification model based on deep learning according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an offline power network deep reinforcement learning model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an on-line power network deep reinforcement learning model according to an embodiment of the present invention;
FIG. 5 is a graph of a power network robustness analysis of an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Example 1
According to the intelligent power grid key node identification method based on deep reinforcement learning, disclosed by the invention, as shown in FIG. 1, an intelligent power grid is taken as a research object, the problem of key node identification in a real power network is analyzed, a complex network theory is combined, and a brand-new intelligent network key node identification framework is provided based on the deep reinforcement learning method; accurately identifying key nodes in the power network through offline and online training of the model; and attacking the key nodes by adopting different attack methods, analyzing the robustness of the power network, and giving power network early warning according to the analysis result. The method mainly comprises the following points:
(1) a power network key node identification model based on deep reinforcement learning is constructed, and key nodes of different types of power networks can be accurately identified through the model.
(2) And analyzing the robustness of the power network from multiple dimensions based on the identified key nodes, and giving early warning information.
Specifically, the method comprises the following steps:
s1: and establishing a topological structure of the power grid and the communication network according to the power grid data, constructing a real power network by combining the service attribute and the physical characteristic of the power network, and constructing and generating a network model.
S2: and constructing a full-order sub-graph feature classification model based on deep learning according to the network features of the real power network and the generation network. Specifically, the method comprises the following steps:
as shown in fig. 2, the main idea of the model is also to organically combine the sub-graph mode with the physical and business attribute features of the grid. Firstly, according to the topological structure of the network, a node2vec method is adopted to map the network to a low-dimensional vector space; selecting a sub-graph node corresponding to a target node according to a diffusion rule, and acquiring a node set with a specific sequence according to the migration distance between nodes; and constructing full-order subgraph features by using the subgraph adjacency matrix and the node vector matrix learned by the node2vec and a katz matrix based on global path similarity, wherein the full-order subgraph features respectively represent low-order, medium-order and high-order structural characteristics of nodes in the subgraph. The processing method adopted for the physical attributes is to represent and learn the attribute matrix of the node as an important feature. Through the deep full-link learning, the reward value corresponding to each action in the reinforcement learning is output, as shown in fig. 2.
S3: outputting the reward value corresponding to each action through deep full connection learning according to the full-order sub-graph feature classification model; and constructing a power network deep reinforcement learning model by using the obtained actions and the corresponding reward values.
The deep reinforcement learning model adopts an offline learning and online supplementing mode. The main training is done offline, as shown in fig. 3, starting to enter the entire network, apply a removal strategy to the nodes, get the current state, and calculate the actions and rewards taken using the proposed depth model. The data set adopts the established generative model to generate a large number of power networks, and the networks have similar topological structure characteristics and physical properties with the real networks. Because real power grid data are less, the generation network training is adopted for most of the training time, and 80% of data of the known power grid data set are used as training data. The input of the training is individual networks, and through the learning of the deep reinforcement learning model, the index value of each node is output, and the higher the index value is, the more important the nodes are explained. Evaluation indexes in training have important influence on the identification of key nodes, so that the method is intended to select a plurality of indexes for comprehensive comparison. As shown in fig. 4, the online learning input is a real power network, including characteristic information such as a topological structure characteristic, a physical attribute, and a service attribute of the network. Because the number of real large-scale interconnected power grid nodes is large, the online learning model needs to consider to perform corresponding optimization processing on the processing of the network.
S4: inputting the power network to be identified to the trained power network deep reinforcement learning model, outputting index values of all nodes, and determining key nodes according to the index values.
S5: the robustness of the power network is not only related to the topological structure and physical properties of the network, but also related to the accuracy degree of an attack mode, so that the robustness of the power network is evaluated by adopting various attack strategies. The power network robustness analysis study is shown in fig. 5. Firstly, constructing a robustness evaluation model of the power network by attacking structural characteristic attributes of the power network (such as key nodes, intermediate nodes, large nodes, random nodes and the like); then, measuring the robustness of the power network through the model; and finally, quantifying the evaluation value to give early warning prompts of different levels.
The present invention is not limited to the above preferred embodiments, and any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A smart grid key node identification method based on deep reinforcement learning is characterized by comprising the following steps:
establishing a topological structure of a power grid and a communication network according to the power grid data; constructing a real power network according to the service attribute and the physical characteristic of the power network; constructing and generating a network model;
constructing a full-order sub-graph feature classification model based on deep learning according to network features of a real power network and a generation network;
outputting the reward value corresponding to each action through deep full connection learning according to the full-order sub-graph feature classification model;
constructing a power network deep reinforcement learning model by using the obtained actions and the corresponding reward values;
and inputting the power network to be identified to the trained power network deep reinforcement learning model, outputting index values of all nodes, and determining key nodes according to the index values.
2. The smart grid key node identification method based on deep reinforcement learning as claimed in claim 1, wherein the construction of the full-order sub-graph feature classification model based on deep learning comprises the following steps:
according to the topological structure of the network, a node2vec method is adopted to map the network to a low-dimensional vector space;
selecting a sub-graph node corresponding to a target node according to a diffusion rule, and acquiring a node set with a specific sequence according to the migration distance between nodes;
and constructing full-order subgraph features by using the subgraph adjacency matrix and the node vector matrix learned by the node2vec and a katz matrix based on global path similarity, wherein the full-order subgraph features respectively represent low-order, medium-order and high-order structural characteristics of nodes in the subgraph.
3. The smart grid key node identification method based on deep reinforcement learning according to claim 1, further comprising: the power network deep reinforcement learning model is trained according to a plurality of index values of the nodes in an online and offline combined mode through a real power network and a generation network.
4. The intelligent power grid key node identification method based on deep reinforcement learning of claim 3, wherein the online learning input is a real power network, and the index value comprises characteristic information of topological structure characteristics, physical attributes and business attributes of the network.
5. The smart grid key node identification method based on deep reinforcement learning according to claim 4, wherein the offline learning comprises the following steps:
inputting a training network in a data set, adopting a removal strategy for each node to obtain a current state, and calculating actions and rewards by using a full-order sub-graph feature classification model; the data set adopts the established generating network model to generate a large number of power networks with topological structure characteristics and physical attributes similar to those of a real network;
the training network outputs the index value of each node through the learning of the deep reinforcement learning model, and the key node is judged according to the index value.
6. The smart grid key node identification method based on deep reinforcement learning according to claim 1, further comprising:
attacking the structural characteristic attributes of the power network, wherein the structural characteristic attributes comprise key nodes, intermediate nodes, large-scale nodes and random nodes;
and constructing a power network robustness evaluation model, measuring the robustness of the power network through the model, and giving early warning prompts of different levels through the quantitative evaluation value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210518642.5A CN114943430B (en) | 2022-05-13 | 2022-05-13 | Smart grid key node identification method based on deep reinforcement learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210518642.5A CN114943430B (en) | 2022-05-13 | 2022-05-13 | Smart grid key node identification method based on deep reinforcement learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114943430A true CN114943430A (en) | 2022-08-26 |
| CN114943430B CN114943430B (en) | 2023-04-28 |
Family
ID=82907219
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210518642.5A Active CN114943430B (en) | 2022-05-13 | 2022-05-13 | Smart grid key node identification method based on deep reinforcement learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114943430B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105303460A (en) * | 2015-10-30 | 2016-02-03 | 国家电网公司 | Identification method of key nodes and key branches in power grid |
| CN112487658A (en) * | 2020-12-14 | 2021-03-12 | 重庆邮电大学 | Method, device and system for identifying key nodes of power grid |
| CN113285828A (en) * | 2021-05-19 | 2021-08-20 | 湖南经研电力设计有限公司 | Complex network key node identification method and power grid key node identification method |
| CN114266475A (en) * | 2021-12-22 | 2022-04-01 | 深圳供电局有限公司 | Power network key node identification method based on multi-attribute decision |
-
2022
- 2022-05-13 CN CN202210518642.5A patent/CN114943430B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105303460A (en) * | 2015-10-30 | 2016-02-03 | 国家电网公司 | Identification method of key nodes and key branches in power grid |
| CN112487658A (en) * | 2020-12-14 | 2021-03-12 | 重庆邮电大学 | Method, device and system for identifying key nodes of power grid |
| CN113285828A (en) * | 2021-05-19 | 2021-08-20 | 湖南经研电力设计有限公司 | Complex network key node identification method and power grid key node identification method |
| CN114266475A (en) * | 2021-12-22 | 2022-04-01 | 深圳供电局有限公司 | Power network key node identification method based on multi-attribute decision |
Non-Patent Citations (1)
| Title |
|---|
| 汪黎明: "基于深度强化学习的复杂网络关键节点识别", 《中国优秀硕士学位论文全文数据库》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114943430B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Boosting slime mould algorithm for parameter identification of photovoltaic models | |
| Jain et al. | Data mining techniques: a survey paper | |
| Liang et al. | Applying genetic algorithm and ant colony optimization algorithm into marine investigation path planning model | |
| CN107133695A (en) | A kind of wind power forecasting method and system | |
| CN104881735A (en) | System and method of smart power grid big data mining for supporting smart city operation management | |
| Gong et al. | Novel heuristic density-based method for community detection in networks | |
| Mohammed et al. | An innovative synthesis of optmization techniques (FDIRE-GSK) for generation electrical renewable energy from natural resources | |
| CN113065974A (en) | Link prediction method based on dynamic network representation learning | |
| CN103455610A (en) | Network community detecting method based on multi-objective memetic computation | |
| CN114004155A (en) | Transient stability assessment method and device considering topological structure characteristics of power system | |
| CN111984702A (en) | Method, device, equipment and storage medium for analyzing spatial evolution of village and town settlement | |
| CN114781688A (en) | Method, device, equipment and storage medium for identifying abnormal data of business expansion project | |
| Guo et al. | Distribution network topology identification based on gradient boosting decision tree and attribute weighted naive Bayes | |
| Cao et al. | Fast and explainable warm-start point learning for AC Optimal Power Flow using decision tree | |
| Tunali | Large-scale network community detection using similarity-guided merge and refinement | |
| Kan et al. | Comprehensive evaluation of tourism resources based on multispecies evolutionary genetic algorithm-enabled neural networks | |
| CN114943430B (en) | Smart grid key node identification method based on deep reinforcement learning | |
| Ahmad et al. | Energy demand forecasting of buildings using random neural networks | |
| CN114676887A (en) | River water quality prediction method based on graph convolution STG-LSTM | |
| Kashyap et al. | Multi-objective metaheuristic search algorithms for service composition in IoT | |
| Weng et al. | Boosted backtracking search optimization with information exchange for photovoltaic system evaluation | |
| CN112651183A (en) | Reliability evaluation method for quantum distributed countermeasure unified deep hash network | |
| Zhang | Traffic Sign Detection Based on YOLO v3 | |
| Tian et al. | Time-delay prediction method based on improved genetic algorithm optimized echo state networks | |
| Chai et al. | Production Forecast of Coalbed Methane Based on GA Optimized BP Neural Network |
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 |