CN-121984880-A - Optimization method of operation and maintenance strategy of power communication network

CN121984880ACN 121984880 ACN121984880 ACN 121984880ACN-121984880-A

Abstract

The application relates to the technical field of power system communication, in particular to an optimization method of an operation and maintenance strategy of a power communication network, the method utilizes a sliding window and a graph neural network to perform space-time alignment and feature reconstruction on multi-source heterogeneous telemetry data of an electric power communication network, and constructs a state tensor reflecting physical and logical panoramas of the network. And introducing a dynamic scoring mechanism comprising network pressure sensing and marginal utility analysis, fusing a high-risk strategy by combining a logarithmic barrier function, and selecting an optimal strategy considering both performance and safety from the candidate set. Finally, a final execution instruction is generated through the verification of the power business hard rule filter, and the real network feedback after the execution is stored in the experience playback buffer area to update the model parameters, so that the problems that the traditional static strategy cannot be dynamically adapted to the network pressure and the direct trial-and-error risk is high are solved.

Inventors

WU LIJIE
SHENG LEI
ZHANG NINGNING
YUAN ZHENGZHENG
FENG HAO
WANG CHUNYING
AN ZHIYUAN
Quan Yizhan
WANG LEI
DONG JIAOJIAO
ZHANG YUJIA
LIU YAN

Assignees

国网河南省电力公司信息通信分公司

Dates

Publication Date: 20260505
Application Date: 20260214

Claims (8)

1. An optimization method for an operation and maintenance strategy of an electric power communication network is characterized by comprising the following steps: s1, acquiring an original telemetry data stream of an electric power communication network; S2, carrying out data cleaning alignment and graph structural feature extraction on an original telemetry data stream to obtain a network panoramic state tensor; s3, inputting the network panoramic state tensor into a preset deep reinforcement learning action network to perform reasoning so as to obtain a candidate operation and maintenance strategy set; S4, performing policy previewing and evaluation on the candidate operation and maintenance policy set under the digital twin environment to obtain a policy evaluation score list; S5, selecting a target strategy from the candidate operation and maintenance strategy set based on the strategy evaluation score list; and S6, performing hard constraint verification on the target strategy based on the power business safety rule to obtain an optimal execution instruction passing the verification.
2. The method for optimizing operation and maintenance policies of a power communication network according to claim 1, wherein the original telemetry data stream comprises a port optical power value, a real-time signal-to-noise ratio, a port traffic utilization, a bit error rate of a service layer, topology information of a current routing table, logic routing table information of the current service layer, and global constraints.
3. The method for optimizing an operation and maintenance strategy of an electric power communication network according to claim 1, wherein step S2 comprises: performing sliding window based data cleansing and time sequence alignment on the original telemetry data stream to obtain a preprocessing standardized data matrix; Constructing a topological graph based on the physical connection relation of the power communication network and generating an adjacency matrix; Carrying out topological feature extraction based on a graph rolling network on the preprocessing standardized data matrix and the adjacent matrix to obtain a space topological feature matrix; logic routing table information of a current service layer in an original telemetry data stream and global constraint are encoded into service constraint feature vectors through an embedded layer; and performing multi-mode splicing and dimensional remolding processing on the space topology feature matrix and the service constraint feature vector to obtain a network panoramic state tensor.
4. The method for optimizing operation and maintenance policies of a power communication network according to claim 1, wherein step S3 comprises: performing feature mapping and inference based on a depth neural network on the network panoramic state tensor to obtain an original action log probability vector; Introducing a temperature coefficient to carry out numerical value smooth adjustment on the original action log probability vector and carrying out probability space transformation processing by utilizing a normalized exponential function so as to obtain action probability distribution; And carrying out strategy instantiation and decoding based on Top-K sampling on the action probability distribution to obtain a candidate operation and maintenance strategy set.
5. The method for optimizing operation and maintenance policies of a power communication network according to claim 1, wherein step S4 comprises: analyzing physical topology parameters in the network panoramic state tensor, loading the physical topology parameters into a virtual simulation engine for state synchronous mapping to obtain a twin simulation environment object in a synchronous initial state; In the twin simulation environment object in the synchronized initial state, the candidate operation and maintenance strategy set is simulated in a multithread parallel mode to obtain a simulation result index set; and carrying out strategy quantization scoring on the multidimensional performance data in the simulation result index set to obtain a strategy evaluation score list.
6. The method of optimizing power communication network operation and maintenance policies according to claim 5, wherein the multi-dimensional performance data comprises virtual average latency, virtual maximum link utilization, and potential traffic outage probabilities.
7. The method for optimizing operation and maintenance policies of a power communication network according to claim 1, wherein step S5 comprises: based on the strategy evaluation score list, performing primary optimal strategy selection based on scores on the candidate operation and maintenance strategy set to obtain an ordered strategy queue to be executed; Performing a security filter based on the power business hard rule on the ordered policy queue to obtain a verified compliance policy; and formatting and compiling the checked compliance strategy to obtain candidate execution instructions.
8. The method for optimizing operation and maintenance policies of a power communication network according to claim 5, wherein performing policy quantization scoring on the multidimensional performance data in the simulation result index set to obtain a policy evaluation score list comprises: Constructing a dynamic weight vector based on the simulation result index set; Performing performance marginal utility mapping on the simulation result index set and the dynamic weight vector to obtain a performance marginal utility score; And carrying out risk fusion and fusion scoring on the performance marginal utility score and the potential business interruption probability in the simulation result index set to obtain a strategy evaluation score list.

Description

Optimization method of operation and maintenance strategy of power communication network Technical Field The application relates to the technical field of power system communication, in particular to an optimization method of an operation and maintenance strategy of a power communication network. Background The electric power communication network is used as a core neural network of the intelligent power grid and carries key production services such as relay protection, safety automatic devices, scheduling automation and the like, and the reliability and instantaneity of the operation of the electric power communication network are directly related to the safety and stability of the physical power grid. With the deep advancement of the western electric east delivery strategy and the construction of a novel electric power system, the topology structure of an electric power backbone communication network is increasingly complex, and the service flow presents the characteristics of high burst, large bandwidth and differentiated service quality requirements. When facing the dynamic scenes such as optical cable aging, natural disasters or sudden business congestion, how to quickly generate and execute the optimal operation and maintenance strategy (such as route reconstruction and wavelength scheduling) becomes a key technical challenge for guaranteeing the smoothness of the life line of the power grid. However, existing power communication network operation and maintenance means mainly depend on manual experience or preset schemes based on static rules, and have significant hysteresis and limitation. Firstly, the traditional operation and maintenance strategy usually adopts a linear weighting model based on fixed weight to evaluate the network state, and the stiff evaluation system cannot sense the current health pressure of the network, namely, the network stability should pay attention to the resource efficiency, and the survivability should be guaranteed in a high risk period (such as alarm frequent). The prior art lacks the ability to dynamically adjust the decision center of gravity according to the environmental pressure, and often neglects the marginal utility decreasing rule of the performance optimization, so that the resource waste is caused on the excessive optimization of non-key indexes meeting the SLA requirement, even the risk of overcoming the service safety for reducing the tiny delay appears, and a ticket overrule mechanism for the high-risk strategy is lacking. Secondly, most of the existing automatic operation and maintenance are open-loop control, a real-time feedback and learning mechanism for the action execution effect is lacking, and the routing trial-and-error is directly performed by applying an intelligent algorithm in a production network, so that the risk of service interruption is extremely high, and the severe requirement of power service on hard constraint is difficult to meet. Therefore, an optimization method scheme of the operation and maintenance strategy of the power communication network is desired. Disclosure of Invention The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides an optimization method of an operation and maintenance strategy of an electric power communication network, which comprises the following steps: s1, acquiring an original telemetry data stream of an electric power communication network; S2, carrying out data cleaning alignment and graph structural feature extraction on an original telemetry data stream to obtain a network panoramic state tensor; s3, inputting the network panoramic state tensor into a preset deep reinforcement learning action network to perform reasoning so as to obtain a candidate operation and maintenance strategy set; S4, performing policy previewing and evaluation on the candidate operation and maintenance policy set under the digital twin environment to obtain a policy evaluation score list; S5, selecting a target strategy from the candidate operation and maintenance strategy set based on the strategy evaluation score list; and S6, performing hard constraint verification on the target strategy based on the power business safety rule to obtain an optimal execution instruction passing the verification. Compared with the prior art, the application provides an optimization method of the operation and maintenance strategy of the power communication network. Firstly, panoramic state sensing is carried out on multi-source telemetry data through a graph neural network, a candidate strategy set is generated by utilizing a deep reinforcement learning model, and parallel simulation previewing is carried out in a digital twin environment. Based on the method, the evaluation weight is dynamically adjusted by utilizing the network entropy pressurizing force index, and the nonlinear quantitative evaluation is carried out on the strategy by combining the performance margi