CN110971471A - Power communication backbone network fault recovery method and device based on state perception - Google Patents
Power communication backbone network fault recovery method and device based on state perception Download PDFInfo
- Publication number
- CN110971471A CN110971471A CN201911394553.9A CN201911394553A CN110971471A CN 110971471 A CN110971471 A CN 110971471A CN 201911394553 A CN201911394553 A CN 201911394553A CN 110971471 A CN110971471 A CN 110971471A
- Authority
- CN
- China
- Prior art keywords
- nodes
- node
- saturated
- unsaturated
- max
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0654—Management of faults, events, alarms or notifications using network fault recovery
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/12—Discovery or management of network topologies
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a state-aware power communication backbone network fault recovery method and device, which are used for obtaining the number of nodes of which actual unsaturated nodes in the power communication backbone network are turned into saturated nodes at the current moment; predicting the number of nodes for converting the unsaturated nodes into the saturated nodes at the next moment according to the number of the nodes for converting the actual unsaturated nodes into the saturated nodes; and taking the number of the nodes of which the unsaturated nodes are converted into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and the failure recovery is carried out according to the final recovery strategy. The invention shows better performance in the power communication backbone network fault recovery problem.
Description
Technical Field
The invention relates to the technical field of intelligent power grids and network communication, in particular to a method and a device for recovering a fault of a power communication backbone network based on state perception.
Background
Modern power grids rely on a corresponding communication backbone and ensure safe operation and authority control of the power grid via the communication network. As an important infrastructure, the communication backbone network has different bandwidth, delay and reliability characteristics compared with the general communication network, and its corresponding fault diagnosis, location and recovery problems are also hot research. The mainstream method is to alleviate the channel congestion problem by establishing a fault model of the communication network and the cache capacity, and typically a fault detection and recovery method (DM-FDR) based on distributed deployment, and a fault detection and recovery method (GA-CFRM) based on a genetic algorithm.
However, existing methods do not guarantee that the smart grid service is completely isolated when a link failure occurs, thereby recovering from the failure with minimal cost. Reinforcement learning, one of the core technologies of artificial intelligence, has a significant effect in the aspect of effectively analyzing large-scale data. As a tool specially used for analyzing rapidly-growing large-scale heterogeneous data, the reinforcement learning method is very suitable for solving the problems of communication backbone network fault prediction and processing.
Disclosure of Invention
The invention provides a state-sensing-based power communication backbone network fault recovery method, which solves the problems that rapid isolation cannot be realized when a link fault occurs in the conventional power communication backbone network, and the fault recovery cost is high.
In order to achieve the above purpose, the invention adopts the following technical scheme: a power communication backbone network fault recovery method based on artificial intelligence and state perception comprises the following steps:
acquiring the number of nodes of which actual unsaturated nodes are turned into saturated nodes in the power communication backbone network at the current moment;
the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes is corrected and predicted through the weighting parameters to obtain the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes at the next moment;
and taking the number of the nodes of which the unsaturated nodes are converted into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and the failure recovery is carried out according to the final recovery strategy.
Further, the number of the nodes where the actual unsaturated node is inverted into the saturated node is as follows:
Nact(t)＝pr(k,q)·Nmax(t)
in the formula, k and q are respectively node numbers; p is a radical ofr(k, q) represents at least MmaxNode k rollover probability, n, when a packet is dropped at node qkIndicates the number of packets at node k, MmaxRepresenting the maximum number of packets that a node can receive, nqIndicating the number of packets at node q,
representing the maximum number of packets at node k,
represents the maximum number of packets at node q;
wherein m issizeIs the packet size; b is the channel transmission rate; Δ t is the time interval;
Nmax(t)＝Nv(t)
Nmax(t) represents the maximum number of inversion nodes at which the unsaturated node is inverted to the saturated node at time t, Nv(t) represents the number of nodes that can communicate with a node at time t.
Further, the number of the nodes at which the unsaturated node is turned into the saturated node at the next moment is Nact(t+Δt)：
Nact(t+Δt)＝pr(k,q)·Nmax(t+Δt)
Nmax(t+Δt)＝N′v(t+Δt)：
N′v(t + Δ t) is the average of the sum of the predicted value of the number of nodes with which the current node can communicate and the predicted values of the number of nodes with which all reachable nodes around the current node can communicate;
under the normalized condition, u + b + d is 1, and a simultaneous equation set of the following three equations is solved:
where v and s are nodes, respectively, known at tThe number of nodes with which the respective time of day can communicate is Nv(t) and Ns(t)，Nv(t + Δ t) is a predicted value of the number of nodes that can communicate with the current node v at time t + Δ t, Ns(t + Δ t) represents a predicted value of the number of nodes that can communicate with a node s at the time t + Δ t, the node s representing a certain node communicating with the current node v; b represents the weighted weight of the current node v flip state perception result; d represents the influence weight of the sensing result of the node s in the overturning state, u represents the empirical influence weight of other unknown factors on the link between the two nodes, and N is obtained by solvingv(t+Δt)，Ns(t+Δt)，b,d,u。
Further, the reinforcement learning algorithm proposes a new network topology as a recovery strategy when a certain link e fails, establishes a new more effective recovery strategy according to the obtained feedback, and performs failure recovery according to a final recovery strategy, and the specific steps include:
1) when a certain link E belongs to E and fails, taking out the service affected by the failure from the new network topology structure; e represents the aggregate of links;
2) recalculating a risk value of each service in the power communication under a new network topology structure, sequencing the risk values according to the sizes, removing links with risks above a threshold value, and then solving the shortest path between nodes to serve as a service path;
3) and if the service path exists, performing fault recovery through the service path, otherwise, continuously updating the network structure, repeatedly calculating the risk value, and solving the path until the shortest path is solved to realize fault recovery.
A power communication backbone network fault recovery device based on artificial intelligence and state perception comprises:
the actual overturning node acquisition module is used for acquiring the number of nodes of which actual unsaturated nodes in the power communication backbone network are overturned into saturated nodes at the current moment;
the next-time turning node prediction module is used for correcting and predicting the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes through weighting parameters to obtain the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes at the next time;
and the fault recovery module is used for taking the number of the nodes of which the unsaturated nodes are turned into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and fault recovery is carried out according to a final recovery strategy.
Further, the number of the nodes where the actual unsaturated node is inverted into the saturated node is as follows:
Nact(t)＝pr(k,q)·Nmax(t)
in the formula, k and q are respectively node numbers; p is a radical ofr(k, q) represents at least MmaxNode k rollover probability, n, when a packet is dropped at node qkIndicates the number of packets at node k, MmaxRepresenting the maximum number of packets that a node can receive, nqIndicating the number of packets at node q,
representing the maximum number of packets at node k,
represents the maximum number of packets at node q;
wherein m issizeIs the packet size; b is the channel transmission rate; Δ t is the time interval;
Nmax(t)＝Nv(t)
Nmax(t) represents the maximum number of inversion nodes at which the unsaturated node is inverted to the saturated node at time t, Nv(t) represents the number of nodes that can communicate with a node at time t.
Further, the number of the nodes at which the unsaturated node is turned into the saturated node at the next moment is Nact(t+Δt)：
Nact(t+Δt)＝pr(k,q)·Nmax(t+Δt)
Nmax(t+Δt)＝Nv(t+Δt)：
N′v(t + Δ t) is the average of the sum of the predicted value of the number of nodes with which the current node can communicate and the predicted values of the number of nodes with which all reachable nodes around the current node can communicate;
under the normalized condition, u + b + d is 1, and a simultaneous equation set of the following three equations is solved:
where v and s are nodes, respectively, and the number of nodes with which communication is possible at time t is known to be Nv(t) and Ns(t)，Nv(t + Δ t) is a predicted value of the number of nodes that can communicate with the current node v at time t + Δ t, Ns(t + Δ t) represents a predicted value of the number of nodes that can communicate with a node s at the time t + Δ t, the node s representing a certain node communicating with the current node v; b represents the weighted weight of the current node v flip state perception result; d represents the influence weight of the sensing result of the node s in the overturning state, u represents the empirical influence weight of other unknown factors on the link between the two nodes, and N is obtained by solvings(t+Δt)，Ns(t+Δt)，b,d,u。
Further, the reinforcement learning algorithm proposes a new network topology as a recovery strategy when a certain link e fails, establishes a new more effective recovery strategy according to the obtained feedback, and performs failure recovery according to a final recovery strategy, and the specific steps include:
1) when a certain link E belongs to E and fails, taking out the service affected by the failure from the new network topology structure; e represents the aggregate of links;
2) recalculating a risk value of each service in the power communication under a new network topology structure, sequencing the risk values according to the sizes, removing links with risks above a threshold value, and then solving the shortest path between nodes to serve as a service path;
3) and if the service path exists, performing fault recovery through the service path, otherwise, continuously updating the network structure, repeatedly calculating the risk value, and solving the path until the shortest path is solved to realize fault recovery.
The invention achieves the following beneficial effects: the invention combines the state perception of the node and the reinforcement learning method to obtain an automatic fault recovery method, and the automatic fault recovery method is applied to the power communication backbone network, thereby realizing an automatic fault recovery method, greatly reducing the human intervention and improving the reliability of network service. When a link fault occurs in the power communication backbone network, the fast isolation is realized, and the fault recovery cost is low;
compared with the traditional DM-FDR and GA-CFRM methods, the method has better performance in the problem of power communication backbone network fault recovery.
Drawings
FIG. 1 is a schematic diagram of an infrastructure network and corresponding serving network;
FIG. 2 is a schematic diagram of an enhanced learning algorithm architecture for a power communications backbone;
fig. 3 is a flow chart of the RL-FRA algorithm.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
Referring to fig. 1, there is shown a basic power communication backbone network, consisting of actual physical nodes a-F, and various service networks of network loads. On the network, in different topologiesRunning corresponding web services, including1-d1Service network 1 and a formed by nodes2-e2The nodes form a service network 2.
Example 1:
a power communication backbone network fault recovery method based on artificial intelligence and state perception comprises the following steps:
step 1, acquiring the number of nodes of which actual unsaturated nodes are turned into saturated nodes in a power communication backbone network at the current moment t;
the occupancy degree of the node cache is a direct index reflecting the communication state of the network. However, for the power communication backbone network, the state of a single node is not enough to reflect the congestion information of all the nodes around. If only a certain link is blocked, the reasonable forwarding modification can effectively relieve the blockage. Nodes in the network may be classified into saturated nodes and unsaturated nodes according to their cache occupancy states. When the node judges the state of the node, the corresponding blocking state can be judged according to the return information of other nodes in the communication range. Ideally, the upper limit of the number of the unsaturated nodes turned into the saturated nodes can be determined by the information of the peripheral nodes received within a period of time:
Nmax(t)＝Nv(t) (1)
in the formula (1), Nmax(t) represents the maximum number of inversion nodes at which the unsaturated node is inverted to the saturated node at time t, Nv(t) represents the number of nodes that can communicate with the current node v at time t, i.e. the node congestion status. Consider the maximum capacity for data transmission on a link between two nodes over a time interval Δ t of B x Δ t, where B is the channel transmission rate. If the residual cache capacity on a node is larger than B multiplied by delta t, the probability of the node being turned into a saturated node is 0, otherwise, the node is in a transition state, and the maximum number M of data packets which can be received by the node in the transition state at the momentmaxComprises the following steps:
wherein m issizeIs the packet size.
According to the routing principle, when between any two nodes k and q, at least MmaxWhen a data packet is discarded at the node q, the node k will turn over, and the turning probability of the node k is as follows:
in the formula, k and q are respectively node numbers; p is a radical ofr(k, q) represents at least MmaxNode k rollover probability, n, when a packet is dropped at node qkIndicates the number of packets at node k, MmaxRepresenting the maximum number of packets that a node can receive, nqIndicating the number of packets at node q,
representing the maximum number of packets at node k,
represents the maximum number of packets at node q;
therefore, the number of unsaturated node flips actually occurred is as shown in equation (4):
Nact(t)＝pr(k,q)·Nmax(t) (4)
in the formula, Nact(t) represents the number of flip nodes that actually occur. The more the number of actually-occurring turning nodes is, the higher the congestion degree of the nodes is; therefore, the congestion degree of the node can be automatically evaluated according to the number of actually occurring overturning nodes (namely the overturning state perception result of the node).
Step 2, correcting and predicting the number of the nodes which are actually converted into the saturated nodes through weighting parameters to obtain the number of the nodes which are converted into the saturated nodes at the next moment;
the state information of the network nodes has strong dynamic characteristics, so that the number N of the nodes capable of communicating with the current node v only at the moment tv(t) the precise state of the node v at the next time t + Δ t cannot be accurately predictedI.e. the number of nodes N that can communicate with the current node v at time t + Deltatv(t + Δ t), the present invention improves the prediction accuracy of the node state by quantifying the uncertainty of the node state. For nodes v and s, it is known that the number of nodes with which each can communicate at time t is Nv(t) and Ns(t) at time t + Δ t for the corresponding Nv(t + Δ t) and Ns(t + Δ t) from Nv(t) and Ns(t) and weighted array (b, d, u) prediction. Wherein b represents the sensing result of the current node v for the flip state in step 1 (i.e. the number of nodes N capable of communicating with the current node v at time t)v(t)) a weighted weight positively correlated to the congestion status of the node at the next time; d represents the influence weight of the sensing result of the peer node s in a communication state with the node v by turning the state, and u represents the experience influence weight of other unknown factors on the link between the two nodes. Three weight factors meet the normalization requirement. In order to correct the influence of other factors, the node state N at the moment t + delta t is improvedv(t + Δ t) and NsPrediction accuracy of (t + Δ t), normalizing u to:
the other two weighting factors are calculated according to the following formula:
to N between two nodesv(t + Δ t) and NsThe solution of (t + Δ t) can solve the simultaneous equations of equations (5), (6) and (7) under the normalized condition (i.e., u + b + d ═ 1). After all nodes and peripheral nodes are pairwise simultaneous equations, N is given to each nodev(t+Δt)、Ns(t + Δ t) and the solution of u, b, d is iteratively solved in a mean field manner, where N isvN in (t + Δ t)v(t) initialization, NsN in (t + Δ t)s(t) Initialization, u, b, d are initialized with 1, 0, respectively.
State information N corresponding to all current nodes v peripheral reachable nodess(t) is sensed and the prediction is modified according to the weighting parameters to obtain Ns(t + Δ t), and finally predicting value N 'at t + Δ t moment of current node v'v(t + Δ t) is a predicted value N of the number of nodes with which the current node can communicatev(t + Deltat) and the predicted value N of the number of nodes with which all the peripheral reachable nodes can communicatesAverage of the sum of (t + Δ t). Predicting value N 'by the number of nodes with which the current node can communicate at the next moment'v(t + Deltat) correction of Nv(t) to predict more accurately the number N of nodes at the next time when the unsaturated node is inverted to the saturated nodeact(t+Δt)。
Predicting the number N of turning nodes at the moment of t + delta tact(t+Δt)：
Nact(t+Δt)＝pr(k,q)·Nmax(t+Δt) (8)
Nmax(t+Δt)＝N′v(t+Δt) (9)
Correcting and predicting the number of the turning nodes actually generated in the step 1 through weighting parameters to obtain the number of the turning nodes at the next moment; n is a radical ofmax(t + Δ t) represents the maximum number of inversion nodes where the unsaturated node is inverted to the saturated node at time t + Δ t;
and 3, taking the number of the nodes of which the unsaturated nodes are converted into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and the failure recovery is carried out according to the final recovery strategy.
The reinforcement learning algorithm for the power communication backbone is shown in fig. 2.
The input of the algorithm is the number of the predicted overturning nodes at the next moment, and the output is a new learned network topology structure as a recovery strategy. Reinforcement learning is a method for strategy learning by a terminal in the field of machine learning. In learning, the learning end gives a possible new network topology as the trying action a in case of a specific failure. And removing the high-risk link from the network structure, and finally calculating the reasonability of the network structure to draw a conclusion on the action a as the supervision information s for enhancing learning. The learning end judges whether the current action can obtain the forward gain r according to the recursive slave information s of the reinforcement learning algorithm. Finally, the learning end can learn a series of effective network topology structure transformation strategies through learning, and a new replaceable structure is guaranteed to be recovered from a fault topology structure in one step.
When a link e fails, the reinforcement learning algorithm proposes a new network topology structure as a recovery strategy, establishes a new more effective recovery strategy according to the obtained feedback, and performs failure recovery according to a final recovery strategy, as shown in fig. 3, and the specific steps include:
1) when a certain link E belongs to E and fails, taking out the service affected by the failure from the new network topology structure; assume that the communication network topology is represented by G' (V, E), where V represents the aggregate set of node devices and E represents the aggregate set of links.
2) Recalculating a risk value of each service in the power communication under a new network topology structure, sequencing the risk values according to the sizes, removing links with risks above a threshold value, and then solving the shortest path between nodes to serve as a service path;
3) if the service path exists, fault recovery is carried out through the service path, otherwise, the network structure is continuously updated, the risk value is repeatedly calculated, and the path is solved until the shortest path is solved, so that fault recovery is realized; .
Example 2:
a power communication backbone network fault recovery device based on artificial intelligence and state perception comprises:
the actual overturning node acquisition module is used for acquiring the number of nodes of which actual unsaturated nodes in the power communication backbone network are overturned into saturated nodes at the current moment;
the next-time turning node prediction module is used for correcting and predicting the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes through weighting parameters to obtain the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes at the next time;
and the fault recovery module is used for taking the number of the nodes of which the unsaturated nodes are turned into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and fault recovery is carried out according to a final recovery strategy.
Further, the number of the nodes where the actual unsaturated node is inverted into the saturated node is as follows:
Nact(t)＝pr(k,q)·Nmax(t)
in the formula, k and q are respectively node numbers; p is a radical ofr(k, q) represents at least MmaxNode k rollover probability, n, when a packet is dropped at node qkIndicates the number of packets at node k, MmaxRepresenting the maximum number of packets that a node can receive, nqIndicating the number of packets at node q,
representing the maximum number of packets at node k,
represents the maximum number of packets at node q;
wherein m issizeIs the packet size; b is the channel transmission rate; Δ t is the time interval;
Nmax(t)＝Nv(t)
Nmax(t) represents the maximum number of inversion nodes at which the unsaturated node is inverted to the saturated node at time t, Nv(t) represents the number of nodes that can communicate with a node at time t.
Further, the number of the nodes at which the unsaturated node is turned into the saturated node at the next moment is Nact(t+Δt)：
Nact(t+Δt)＝pr(k,q)·Nmax(t+Δt)
Nmax(t+Δt)＝Nv(t+Δt)：
N′v(t + Δ t) is the average of the sum of the predicted value of the number of nodes with which the current node can communicate and the predicted values of the number of nodes with which all reachable nodes around the current node can communicate;
under the normalized condition, u + b + d is 1, and a simultaneous equation set of the following three equations is solved:
where v and s are nodes, respectively, and the number of nodes with which communication is possible at time t is known to be Nv(t) and Ns(t)，Nv(t + Δ t) is the number of nodes that can communicate with the current node v at time t + Δ t, Ns(t + Δ t) represents the number of nodes that can communicate with node s at time t + Δ t, node s representing some node that communicates with current node v; b represents the weighted weight of the current node v flip state perception result; d represents the influence weight brought by saturated nodes in the sensing result of the turning state of the node s, u represents the empirical influence weight of other unknown factors on the link between the two nodes, and N is obtained by solvings(t+Δt)，Ns(t+Δt)，b,d,u。
Further, the reinforcement learning algorithm proposes a new network topology as a recovery strategy when a certain link e fails, establishes a new more effective recovery strategy according to the obtained feedback, and performs failure recovery according to a final recovery strategy, and the specific steps include:
1) when a certain link E belongs to E and fails, taking out the service affected by the failure from the new network topology structure; e denotes the aggregate of links.
2) Recalculating a risk value of each service in the power communication under a new network topology structure, sequencing the risk values according to the sizes, removing links with risks above a threshold value, and then solving the shortest path between nodes to serve as a service path;
3) and if the service path exists, performing fault recovery through the service path, otherwise, continuously updating the network structure, repeatedly calculating the risk value, and solving the path until the shortest path is solved to realize fault recovery.
Simulation results in a virtual environment show that compared with the traditional DM-FDR method and the GA-CFRM method, the recovery capability of the method (RL-FRA) for the service is improved in networks with different scales, and the recovery time is shortened, and the advantage is more obvious particularly when the network scale is enlarged to more than 500 nodes.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application 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 application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams 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.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (8)
1. A power communication backbone network fault recovery method based on artificial intelligence and state perception is characterized in that: the method comprises the following steps:
acquiring the number of nodes of which actual unsaturated nodes are turned into saturated nodes in the power communication backbone network at the current moment;
the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes is corrected and predicted through the weighting parameters to obtain the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes at the next moment;
and taking the number of the nodes of which the unsaturated nodes are converted into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and the failure recovery is carried out according to the final recovery strategy.
2. The method for recovering the fault of the power communication backbone network based on artificial intelligence and state perception as claimed in claim 1, wherein: the number of the nodes for actually inverting the unsaturated nodes into the saturated nodes is as follows:
Nact(t)＝pr(k,q)·Nmax(t)
in the formula, k and q are respectively node numbers; p is a radical ofr(k, q) represents at least MmaxNode k rollover probability, n, when a packet is dropped at node qkIndicates the number of packets at node k, MmaxRepresenting the maximum number of packets that a node can receive, nqIndicating the number of packets at node q,
representing the maximum number of packets at node k,
represents the maximum number of packets at node q;
wherein m issizeIs the packet size; b is the channel transmission rate; Δ t is the time interval;
Nmax(t)＝Nv(t)
Nmax(t) represents the maximum number of inversion nodes at which the unsaturated node is inverted to the saturated node at time t, Nv(t) represents the number of nodes that can communicate with a node at time t.
3. The method for recovering the fault of the power communication backbone network based on artificial intelligence and state perception as claimed in claim 2, wherein: the number of the nodes of which the unsaturated nodes are turned into the saturated nodes at the next moment is Nact(t+Δt)：
Nact(t+Δt)＝pr(k,q)·Nmax(t+Δt)
Nmax(t+Δt)＝Nv'(t+Δt)：
Nv' (t + Δ t) is the average of the sum of the predicted value of the number of nodes with which the current node can communicate and the predicted value of the number of nodes with which all reachable nodes around the current node can communicate;
under the normalized condition, u + b + d is 1, and a simultaneous equation set of the following three equations is solved:
where v and s are nodes, respectively, and the number of nodes with which communication is possible at time t is known to be Nv(t) and Ns(t)，Nv(t + Δ t) is a predicted value of the number of nodes that can communicate with the current node v at time t + Δ t, Ns(t + Δ t) represents a predicted value of the number of nodes that can communicate with a node s at the time t + Δ t, the node s representing a certain node communicating with the current node v; b represents the weighted weight of the current node v flip state perception result; d represents the influence weight of the sensing result of the node s in the overturning state, u represents the empirical influence weight of other unknown factors on the link between the two nodes, and N is obtained by solvingv(t+Δt)，Ns(t+Δt)，b,d,u。
4. The method for recovering the fault of the power communication backbone network based on artificial intelligence and state perception as claimed in claim 1, wherein: when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and the failure recovery is carried out according to the final recovery strategy, which specifically comprises the following steps:
1) when a certain link E belongs to E and fails, taking out the service affected by the failure from the new network topology structure; e represents the aggregate of links;
2) recalculating a risk value of each service in the power communication under a new network topology structure, sequencing the risk values according to the sizes, removing links with risks above a threshold value, and then solving the shortest path between nodes to serve as a service path;
3) and if the service path exists, performing fault recovery through the service path, otherwise, continuously updating the network structure, repeatedly calculating the risk value, and solving the path until the shortest path is solved to realize fault recovery.
5. A power communication backbone network fault recovery device based on artificial intelligence and state perception is characterized in that: the method comprises the following steps:
the actual overturning node acquisition module is used for acquiring the number of nodes of which actual unsaturated nodes in the power communication backbone network are overturned into saturated nodes at the current moment;
the next-time turning node prediction module is used for correcting and predicting the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes through weighting parameters to obtain the number of the nodes of which the actual unsaturated nodes are turned into the saturated nodes at the next time;
and the fault recovery module is used for taking the number of the nodes of which the unsaturated nodes are turned into the saturated nodes at the next moment as the input of the reinforcement learning algorithm, when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and fault recovery is carried out according to a final recovery strategy.
6. The device for recovering the fault of the power communication backbone network based on artificial intelligence and state perception of claim 5, wherein: the number of the nodes for actually inverting the unsaturated nodes into the saturated nodes is as follows:
Nact(t)＝pr(k,q)·Nmax(t)
in the formula, k and q are respectively node numbers; p is a radical ofr(k, q) represents at least MmaxNode k rollover probability, n, when a packet is dropped at node qkIndicates the number of packets at node k, MmaxRepresenting the maximum number of packets that a node can receive, nqIndicating the number of packets at node q,
representing the maximum number of packets at node k,
represents the maximum number of packets at node q;
wherein m issizeIs the packet size; b is the channel transmission rate; Δ t is the time interval;
Nmax(t)＝Nv(t)
Nmax(t) represents the maximum number of inversion nodes at which the unsaturated node is inverted to the saturated node at time t, Nv(t) represents the number of nodes that can communicate with a node at time t.
7. The device for recovering the fault of the power communication backbone network based on artificial intelligence and state perception of claim 5, wherein: the unsaturated node is turned into a saturated node at the next momentThe number of nodes is Nact(t+Δt)：
Nact(t+Δt)＝pr(k,q)·Nmax(t+Δt)
Nmax(t+Δt)＝Nv(t+Δt)：
N′v(t + Δ t) is the average of the sum of the predicted value of the number of nodes with which the current node can communicate and the predicted values of the number of nodes with which all reachable nodes around the current node can communicate;
under the normalized condition, u + b + d is 1, and a simultaneous equation set of the following three equations is solved:
where v and s are nodes, respectively, and the number of nodes with which communication is possible at time t is known to be Nv(t) and Ns(t)，Nv(t + Δ t) is a predicted value of the number of nodes that can communicate with the current node v at time t + Δ t, Ns(t + Δ t) represents a predicted value of the number of nodes that can communicate with a node s at the time t + Δ t, the node s representing a certain node communicating with the current node v; b represents the weighted weight of the current node v flip state perception result; d represents the influence weight of the sensing result of the node s in the overturning state, u represents the empirical influence weight of other unknown factors on the link between the two nodes, and N is obtained by solvings(t+Δt)，Ns(t+Δt)，b,d,u。
8. The device for recovering the fault of the power communication backbone network based on artificial intelligence and state perception of claim 5, wherein: when a certain link e fails, the reinforcement learning algorithm provides a new network topology structure as a recovery strategy, a new more effective recovery strategy is established according to the obtained feedback, and the failure recovery is carried out according to the final recovery strategy, which specifically comprises the following steps:
1) when a certain link E belongs to E and fails, taking out the service affected by the failure from the new network topology structure; e represents the aggregate of links;
2) recalculating a risk value of each service in the power communication under a new network topology structure, sequencing the risk values according to the sizes, removing links with risks above a threshold value, and then solving the shortest path between nodes to serve as a service path;
3) and if the service path exists, performing fault recovery through the service path, otherwise, continuously updating the network structure, repeatedly calculating the risk value, and solving the path until the shortest path is solved to realize fault recovery.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911394553.9A CN110971471B (en) | 2019-12-30 | 2019-12-30 | Power communication backbone network fault recovery method and device based on state perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911394553.9A CN110971471B (en) | 2019-12-30 | 2019-12-30 | Power communication backbone network fault recovery method and device based on state perception |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110971471A true CN110971471A (en) | 2020-04-07 |
| CN110971471B CN110971471B (en) | 2022-03-29 |
Family
ID=70037320
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911394553.9A Active CN110971471B (en) | 2019-12-30 | 2019-12-30 | Power communication backbone network fault recovery method and device based on state perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110971471B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114629779A (en) * | 2021-11-04 | 2022-06-14 | 国网浙江省电力有限公司湖州供电公司 | Active recovery strategy for power communication backbone network under extreme fault |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103906130A (en) * | 2014-03-18 | 2014-07-02 | 重庆邮电大学 | Congestion control method with node state estimation function |
| CN104578427A (en) * | 2015-01-27 | 2015-04-29 | 国家电网公司 | Fault self-healing method for power distribution network containing microgrid power source |
| CN104934968A (en) * | 2015-06-04 | 2015-09-23 | 国家电网公司 | Multi-agent based distribution network disaster responding recovery coordinate control method and multi-agent based distribution network disaster responding recovery coordinate control device |
| CN108112042A (en) * | 2016-11-25 | 2018-06-01 | 中兴通讯股份有限公司 | A kind of method, apparatus and system for keeping service RL |
| CN108573303A (en) * | 2018-04-25 | 2018-09-25 | 北京航空航天大学 | It is a kind of that recovery policy is improved based on the complex network local failure for improving intensified learning certainly |
| US20190102676A1 (en) * | 2017-09-11 | 2019-04-04 | Sas Institute Inc. | Methods and systems for reinforcement learning |
| CN109831229A (en) * | 2019-02-28 | 2019-05-31 | 中国电力科学研究院有限公司 | A kind of method and system for restoring business in powerline network based on intensified learning |
| CN110061982A (en) * | 2019-04-02 | 2019-07-26 | 广州大学 | A kind of confrontation intelligence attack safe transmission method based on intensified learning |
-
2019
- 2019-12-30 CN CN201911394553.9A patent/CN110971471B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103906130A (en) * | 2014-03-18 | 2014-07-02 | 重庆邮电大学 | Congestion control method with node state estimation function |
| CN104578427A (en) * | 2015-01-27 | 2015-04-29 | 国家电网公司 | Fault self-healing method for power distribution network containing microgrid power source |
| CN104934968A (en) * | 2015-06-04 | 2015-09-23 | 国家电网公司 | Multi-agent based distribution network disaster responding recovery coordinate control method and multi-agent based distribution network disaster responding recovery coordinate control device |
| CN108112042A (en) * | 2016-11-25 | 2018-06-01 | 中兴通讯股份有限公司 | A kind of method, apparatus and system for keeping service RL |
| US20190102676A1 (en) * | 2017-09-11 | 2019-04-04 | Sas Institute Inc. | Methods and systems for reinforcement learning |
| CN108573303A (en) * | 2018-04-25 | 2018-09-25 | 北京航空航天大学 | It is a kind of that recovery policy is improved based on the complex network local failure for improving intensified learning certainly |
| CN109831229A (en) * | 2019-02-28 | 2019-05-31 | 中国电力科学研究院有限公司 | A kind of method and system for restoring business in powerline network based on intensified learning |
| CN110061982A (en) * | 2019-04-02 | 2019-07-26 | 广州大学 | A kind of confrontation intelligence attack safe transmission method based on intensified learning |
Non-Patent Citations (2)
| Title |
|---|
| 肖平等: "模糊联想记忆网络的增强学习算法", 《中国图像图形学报》 * |
| 贾惠彬等: "基于强化学习的电力通信网故障恢复方法", 《中国电力》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114629779A (en) * | 2021-11-04 | 2022-06-14 | 国网浙江省电力有限公司湖州供电公司 | Active recovery strategy for power communication backbone network under extreme fault |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110971471B (en) | 2022-03-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112181666B (en) | Equipment assessment and federal learning importance aggregation method based on edge intelligence | |
| CN107634911B (en) | Adaptive congestion control method based on deep learning in information center network | |
| CN109039942B (en) | Network load balancing system and balancing method based on deep reinforcement learning | |
| CN108076158B (en) | Minimum load route selection method and system based on naive Bayes classifier | |
| CN107733877B (en) | Management method and system for wireless communication architecture of Internet of things | |
| CN112422443B (en) | Adaptive control method, storage medium, equipment and system of congestion algorithm | |
| CN107040961B (en) | Congestion control method for wireless sensor network | |
| US20230299872A1 (en) | Neural Network-Based Communication Method and Related Apparatus | |
| CN108696453B (en) | Lightweight SDN service flow notification method and system | |
| CN107155197B (en) | Distributed storage method and device based on multi-hop cooperation and electronic equipment | |
| CN113452676B (en) | Detector distribution method and Internet of things detection system | |
| CN110971471B (en) | Power communication backbone network fault recovery method and device based on state perception | |
| CN111781822A (en) | Privacy protection grouping consistency control method of multi-agent system | |
| CN111191955B (en) | Power CPS risk area prediction method based on dependent Markov chain | |
| CN113411351B (en) | DDoS attack elastic defense method based on NFV and deep learning | |
| WO2022262127A1 (en) | Smart internet-of-things comprehensive awareness system and method | |
| CN108880909B (en) | Network energy saving method and device based on reinforcement learning | |
| CN106332131B (en) | A kind of clone's nodal test method and system of wireless sensor network | |
| WO2023159812A1 (en) | Method and apparatus for detecting ami network intrusion, and medium | |
| CN115865682A (en) | SDN link detection processing method, controller, system and medium | |
| CN115567466A (en) | Time-sensitive traffic online scheduling method and device based on deep reinforcement learning | |
| CN112399483B (en) | Constraint route control system based on GNs graph predictor | |
| CN114422438A (en) | Link adjusting method and device of power communication network | |
| CN116506916A (en) | Routing path determining method, node, terminal and storage medium | |
| CN112437407A (en) | Method and system for calculating initialized safety state of power internet of things wireless terminal equipment based on mixed information entropy |
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 |