CN117077780A

CN117077780A - Evaluation method for optimizing communication private network faults based on particle swarm optimization and knowledge graph

Info

Publication number: CN117077780A
Application number: CN202311351893.XA
Authority: CN
Inventors: 李旭旭; 李兴; 马玫; 王超; 卢金奎; 杨波; 樊雪婷; 姚文浩; 张乐; 谢欢
Original assignee: State Grid Sichuan Electric Power Co Ltd
Current assignee: State Grid Sichuan Electric Power Co Ltd
Priority date: 2023-10-19
Filing date: 2023-10-19
Publication date: 2023-11-17

Abstract

The invention relates to the technical field of communication private networks of transmission lines, and discloses an evaluation method for optimizing communication private network faults based on a particle swarm algorithm and a knowledge graph, which comprises the following steps: collecting related data of a transmission line communication private network, wherein the related data comprises unstructured data and structured data; text extraction is carried out on the unstructured data to form a first triplet taking an entity, a first attribute and a first attribute value as a combination and associated information among the entities, and a second triplet taking a model, a second attribute and a second attribute value as a combination is established according to the attributes and key models in the structured data; constructing an entity, a model, an attribute value and an element of associated information by combining the first triplet with the second triplet; representing the elements in a graph structure mode according to the relation network; and optimizing the training set data through a particle optimization algorithm. The invention can rapidly evaluate and predict faults.

Description

Evaluation method for optimizing communication private network faults based on particle swarm optimization and knowledge graph

Technical Field

The invention relates to the technical field of communication private networks of transmission lines, in particular to an evaluation method for optimizing communication private network faults based on a particle swarm algorithm and a knowledge graph.

Background

The special power transmission line communication network is used as an important component of a power transmission system, plays an important role in ensuring the normal operation and communication of 3 ten thousand or more kilometers of high-voltage transmission lines in China, has wide national coverage, is bad in operation environment, keeps real-time operation and online state for a long time, and has high requirements on the reliable operation of equipment. Based on this, when a fault occurs, the conventional way of diagnosing and locating the fault mainly includes based on manual experience, priori knowledge, fault analysis tree, etc., and in this process, the requirements on the expertise of the fault analysis personnel and the capability of the fault analysis experience are high.

Disclosure of Invention

The invention provides an evaluation method for optimizing communication private network faults based on a particle swarm optimization and a knowledge graph, which utilizes the knowledge graph of multi-source information, including a power transmission line topological graph, equipment parameters, historical fault data, an operation environment (such as weather) and the like, establishes a comprehensive and accurate line communication private network fault knowledge base, utilizes the knowledge base to combine the data information of the multi-source information knowledge graph, utilizes the particle swarm optimization algorithm to perform fault evaluation and prediction optimization, improves the speed and precision of power transmission line communication private network fault evaluation and prediction, and provides a corresponding solution.

The invention is realized by the following technical scheme:

an evaluation method for optimizing communication private network faults based on a particle swarm algorithm and a knowledge graph comprises the following steps:

s1, collecting related data of a transmission line communication private network from multiple aspects, wherein the related data comprises unstructured data and structured data, and the unstructured data is derived from a system architecture, a function description, an operation environment description, a historical fault analysis form, a text and a fault photo of the transmission line communication private network; the structured data is derived from transmission level line service environment information and information acquired by the running state of a communication network;

s2, extracting texts of the unstructured data to form a first triplet taking an entity, a first attribute and a first attribute value as a combination and associated information among the entities to form a relational network, and establishing a second triplet taking the model, a second attribute and a second attribute value as a combination according to the attributes and key models in the structured data, wherein the entity comprises a communication line, communication private network nodes and corresponding responsible personnel;

s3, constructing elements of an entity, a model, an attribute value and associated information according to the first triplet and the second triplet;

S4, combining a plurality of elements according to the consistency and identity judgment standards of the entities, and then representing the elements after the combination treatment in a graph structure mode according to the relation network, so as to form an initial power transmission line communication private network knowledge graph, store and manage the initial power transmission line communication private network knowledge graph, and provide corresponding query interfaces;

s5, acquiring data of the private network of the power transmission line in real time, inquiring through the knowledge graph of the communication private network of the initial power transmission line, taking the inquired result and corresponding faults as training set data, optimizing the training set data through a particle optimization algorithm, and updating the knowledge graph of the communication private network of the initial power transmission line according to the optimized result and the corresponding faults.

In S2, text extraction is performed on the unstructured data to form a first triplet with "entity, first attribute value" as a combination, where the specific process is as follows:

s2.1, entity identification: identifying an entity in the text by using an NER algorithm, and marking the identified entity as a corresponding type;

s2.2, extracting attributes: extracting related first attributes and corresponding first attribute values of each entity from the text to form a first triplet taking the entity, the first attribute and the first attribute value as a combination;

S2.3, establishing a relation: and establishing a relation among the entities according to the context information in the text.

As optimization, the first attribute comprises fault description, fault characteristics, fault analysis, fault occurrence time and fault reason of the entity;

the second attribute comprises the number of characteristic parameters counted and analyzed by manufacturers, specification models, version information, operation time, operation environment, lines and real-time operation state data of the entity.

As optimization, optimizing the training set data by a particle optimization algorithm specifically includes:

s5.1, regarding the transmission line communication private network fault as particles, randomly generating a certain number of particle groups, and initializing the positions, the speeds and related parameters of the particles, wherein the positions of the particles represent the current scheme for solving the transmission line communication private network fault or the parameter configuration of communication private network nodes, and the speeds of the particles represent the searching direction and the searching distance;

s5.2, calculating an fitness function value of each particle according to factors affecting the transmission line communication private network fault;

s5.3, updating the speed of each particle according to the current position, the historical optimal position and the global optimal position of each particle;

S5.4, updating the current position of each particle according to the latest speed and the current position of each particle;

s5.5, updating the historical optimal position and the global optimal position of each particle according to the current position and the fitness function value of each particle;

and S5.6, judging whether a termination condition is met, if so, outputting the global optimal position of the particles, namely an optimal solution, and otherwise, jumping to S5.2.

As optimization, the specific process of S5.1 is:

s5.1.1, initializing particle swarm size N, maximum iteration number max_iter, inertia weight w and acceleration coefficients c1 and c2;

s5.1.2 for each particle i ε [1, N ], the following is performed:

a. randomly generating an initial position xi and an initial speed vi;

b. updating the current position, the historical optimal position and the global optimal position of the particle i to be initial positions xi;

s5.1.3, obtaining initialized particle swarms;

the specific steps of S5.2 are as follows:

s5.2.1, inputting a current position vector x of the particle;

s5.2.2, calculating an fitness function value f (x) of the current position vector x of the particle according to the current position vector x through a set fault analysis tree;

s5.2.3, returning the fitness function value f (x) of the current position vector x of the particle;

S5.2.4 and circulating S5.2.1-S5.2.3 until the fitness function value of the current position vector of each particle in the particle swarm is obtained.

As optimization, the specific steps of calculating the fitness function value through the fault analysis tree are as follows:

a1, constructing the fault analysis tree: constructing a fault analysis tree according to the specific condition of the power transmission line communication private network fault, wherein the fault analysis tree is in a tree structure and is used for describing different fault conditions and influences of the fault on the performance of the communication private network system, wherein nodes of the tree represent fault types or states, namely particles, and edges represent the relationship between two nodes;

a2, based on the influence degree of a plurality of factors on the faults, giving different weights to each node to reflect the influence degree of different faults on the overall performance;

the method comprises the following steps:

1. determining influencing factors and weights: identifying and listing each factor influencing the particle fitness, setting corresponding weight for each factor according to the characteristics and the priority of the problem, wherein the weight reflects the importance degree of the factor on the whole fitness, and the weight needs to be 1 in total or distributed proportionally;

2. normalization factor: normalizing the influence factors of different ranges and units so that all the factors have the same weight when calculating the fitness function;

3. Multiplying the factors by the corresponding weights to obtain the weights of the nodes;

a3, defining an evaluation index in the fault analysis tree: defining a corresponding evaluation index for each node in the fault analysis tree, wherein the evaluation index is a performance parameter related to fault performance or state, and the evaluation index of each node is matched with the requirement and constraint of the fault performance or state;

a4, evaluating from the root node of the fault analysis tree: each node is evaluated step by step downwards from the root node of the fault analysis tree, and an evaluation value of the node is calculated according to the current position vector x and evaluation indexes related to the node, wherein the evaluation value is a real number and is used for representing the performance and the health degree of the current position;

a5, calculating an fitness function value according to the relation among the nodes: in the fault analysis tree, the relation between the nodes comprises parallel or serial, if the nodes are in parallel relation, the evaluation values of the two nodes are combined into a total fitness value in a summation or weighted average mode, and if the nodes are in serial relation, the fitness value depends on the evaluation values of all the nodes on the path.

A6, iterative calculation is carried out until the leaf nodes: according to the structure of the fault analysis tree, iterative calculation is carried out until leaf nodes are reached according to the relation among the nodes, and the finally obtained fitness function value f (x) reflects the performance evaluation of the current position vector x based on the fault analysis tree.

As optimization, the specific process of S5.3 is:

s5.3.1 for each particle i ε [1, N ], the following is performed:

a. calculating a difference dx between the current position xi and the historical optimal position pi;

b. calculating a difference dg between the current position xi and the global optimal position pg;

c. updating the particle velocity vi, the formula is: vi=w+vi+c1×rand () ×dx+c2×rand () ×dg;

where w, c1 and c2 represent the inertial weight, acceleration factor 1 and acceleration factor 2, respectively, and rand () represents a number randomly generated between 0, 1;

d. limiting the particle velocity vi to ensure that the velocity does not exceed a preset maximum vmax;

s5.3.2, returning updated particle velocity;

s5.3.3, repeating S5.3.1-S5.3.2 until all particle velocities are updated;

the specific steps of S5.4 are as follows:

s5.4.1 for each particle i ε [1, N ], the following is performed:

updating the current position xi of the particle, the formula is: xi '=xi+vi, xi being the current position of the particle before updating, xi' being the current position of the updated particle;

s5.4.2, returning the updated current position xi 'of the particle, and regarding the updated current position xi' of the particle as the current position xi in the subsequent operation;

s5.4.3, repeating S5.4.1-S5.4.2 until the current positions of all particles are updated.

As optimization, the specific steps of S5.5 are:

s5.5.1 for each particle i ε [1, N ], the following is performed:

a. if the current fitness function value f (x) is better than the historical optimal fitness function value f (pi), the current position xi is updated to the historical optimal position pi, namely: pi=xi, otherwise, maintaining the original historical optimal position pi;

b. if the current fitness function value f (x) is better than the global optimal fitness function value f (pg), the current position xi is updated to the global optimal position pg, namely: pg=xi, otherwise, the original global optimal position pg is maintained;

s5.5.2, returning updated historical optimal positions and global optimal positions;

s5.5.3, repeating S5.5.1-S5.5.2 until all the particles' historical optimal positions and global optimal positions are updated.

As optimization, the judgment as to whether the termination condition is satisfied specifically includes:

and judging whether the current iteration number reaches the maximum iteration number max_iter or whether the searched global optimal position reaches a preset threshold value.

In S5, as optimization, the specific process of updating the initial power transmission line communication private network knowledge graph according to the optimized result is as follows:

s5.7, determining an entity to be updated, and attribute and associated information thereof according to an output result of the particle swarm optimization algorithm;

S5.8, carrying out standardized processing on the entity and relation to be updated;

s5.9, updating the initial power transmission line communication private network knowledge graph according to the new entity and relation information to obtain a new power transmission line communication private network knowledge graph;

and S5.10, verifying and updating the new power transmission line communication private network knowledge graph.

Compared with the prior art, the invention has the following advantages and beneficial effects:

(1) Integrity and accuracy: by adopting the knowledge graph of the multi-source information, a comprehensive and accurate line communication private network fault knowledge base is established, so that faults can be comprehensively estimated and predicted, and an accurate solution is provided.

(2) The service efficiency is high: the particle swarm optimization algorithm is integrated into the field of power transmission line communication private network fault evaluation and prediction, so that faults can be evaluated and predicted rapidly, and the use efficiency is improved.

(3) The applicability is strong: the invention is suitable for various transmission line communication private network faults and has wide application prospect.

Drawings

In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings that are needed in the examples will be briefly described below, it being understood that the following drawings only illustrate some examples of the present invention and therefore should not be considered as limiting the scope, and that other related drawings may be obtained from these drawings without inventive effort for a person skilled in the art. In the drawings:

Fig. 1 is a flowchart of an evaluation method for optimizing communication private network faults based on a particle swarm algorithm and a knowledge graph.

Description of the embodiments

For the purpose of making apparent the objects, technical solutions and advantages of the present invention, the present invention will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present invention and the descriptions thereof are for illustrating the present invention only and are not to be construed as limiting the present invention.

An evaluation method for optimizing communication private network faults based on a particle swarm algorithm and a knowledge graph of embodiment 1, as shown in fig. 1, comprises:

the system architecture is generally in the form of pictures, the functional description can be text and pictures, the test run environment description is generally text, and the historical fault analysis form is generally a table.

S2, extracting texts from the unstructured data to form a first triplet taking an entity, a first attribute and a first attribute value as a combination and associated information among the entities to form a relational network, and establishing a second triplet taking the model, a second attribute and a second attribute value as a combination according to the attributes and key models in the structured data, wherein the entity comprises a communication line and communication private network nodes;

extracting text from the unstructured data, extracting entity and attribute information from the unstructured data for standardization processing, and performing rule matching, keyword extraction or other natural language processing technologies to form a first triplet taking an entity, a first attribute and a first attribute value as a combination, wherein the specific process is as follows:

s2.1, entity identification: identifying an entity in the text by using an NER algorithm, and marking the identified entity as a corresponding type, wherein the type of the entity comprises, but is not limited to, a system, a module, equipment and a name;

after the algorithm identifies the entity, the corresponding mapping relation is resolved through the text, and the mapping relation is marked by the computer, which is the prior art and will not be described again here.

S2.2, extracting attributes: extracting a first attribute related to each entity and a corresponding first attribute value from the text, wherein if the entity is a name, the attribute is a 'job', the attribute value is a specific job name, if the entity is equipment, the attribute comprises a 'fault name', 'fault grade', and the like, the attribute value is a specific fault name and fault grade, for example, the attribute is a 'fault name', and the corresponding attribute value is a 'motor fault'.

S2.3, establishing a relation: and establishing an association relation between the entities according to the context information in the text. For example, by analyzing the syntactic structure or dependency relationship in the text, the relationship between different entities is determined and the association relationship is established.

And determining the relation between different entities by analyzing the syntactic structure or the dependency relationship in the text, and constructing a relation network in the multi-source information knowledge graph according to the relation and the association between the entities. For example, the relation (such as connection relation between devices, working relation between personnel, etc.) in the text can be extracted by a relation extraction technology based on semantics, and finally represented in a knowledge graph.

The first attribute comprises fault description, fault characteristics, fault analysis, fault occurrence time and fault reasons of the entity; extracting a first triplet containing the entity, attribute and attribute value related to the knowledge map ontology such as line, system, component, fault description, fault feature, fault analysis, fault occurrence time, fault reason and the like from unstructured information such as prior knowledge related to fault diagnosis and prediction, fault analysis tree text and the like;

the second attribute comprises the number of characteristic parameters counted and analyzed by manufacturers, specification models, version information, operation time, operation environment, lines and real-time operation state data of the entity;

and extracting the model, attribute and attribute value triplets related to the knowledge graph body including related real-time running state data, statistics and analysis of the quantity of characteristic parameters and the like from the dynamic real-time structured information of the on-site communication terminal and the communication module.

The real-time running state data comprises network connection state, bandwidth utilization rate, data transmission speed, fault event log, routing table and forwarding information, service quality index, equipment health state, alarm information, running log and the like, and the characteristic parameter quantity comprises network link quality parameter, equipment health state parameter, data flow parameter, routing table and forwarding parameter, fault event parameter, alarm parameter and the like.

S3, constructing elements of an entity, a model, an attribute value and associated information according to the first triplet and the second triplet; such as merging the commonalities, adding additional information or context information, if necessary, via "association information", S4.

For example, a first triplet (entity, first attribute value):

entity: a communication line;

first attribute: a line type;

first attribute value: a backbone fiber;

examples: the first triplet is a communication line, a line type and a trunk optical fiber;

second triplet (model, second attribute value):

and (3) model: a communication line fault assessment model;

second attribute: the type of fault;

second attribute value: loss of signal;

Examples: the second triplet is a communication line fault evaluation model, a fault type and a signal loss;

here, "element" example: (communication line, communication line failure assessment model, line type, backbone fiber, failure type, signal loss).

The entity of the first triplet is a communication line, and the model of the second triplet is a fault evaluation model related to the communication line, so that the second triplet is also a specific communication line, and therefore, the elements after combination are the communication line, the communication line fault evaluation model, the line type, the trunk optical fiber, the fault type and the signal loss. S4, combining a plurality of elements according to the consistency and identity judgment standards of the entities, and then representing the elements after the combination treatment in a graph structure mode according to the relation network, so as to form an initial power transmission line communication private network knowledge graph, store and manage the initial power transmission line communication private network knowledge graph, and provide corresponding query interfaces;

The particle swarm optimization algorithm can be applied to the evaluation and prediction of the transmission line communication private network faults. The transmission line communication private network needs to manage hundreds of communication private network nodes at the same time, and relates to a large amount of data and equipment, and a private network system can collect and store a large amount of real-time data and perform fault prediction and optimal scheduling by using a data analysis technology. The transmission line communication private network fault can be regarded as an optimization problem in a multidimensional space, each search point can be regarded as a particle, the position of the particle represents the state of the current solution, and the speed represents the search direction and distance. The position and the speed of the particles are continuously updated and adjusted, so that the optimal solution is searched and iteratively optimized, and further the evaluation and the prediction of the transmission line communication private network faults are realized.

The training set data is optimized by a particle optimization algorithm specifically as follows:

s5.1, regarding the transmission line communication private network faults as particles, randomly generating particle groups with the number of the particles being N times of the number of the faults,initializing the position, the speed and related parameters of particles, wherein the position of the particles represents the current scheme for solving the communication private network fault of the transmission line or the parameter configuration of the communication private network node, and the speed of the particles represents the searching direction and distance;

s5.1.2 for each particle i ε [1, N ], the following is performed:

a. randomly generating an initial position xi and an initial speed vi; for example, may be randomly generated by a hash function.

s5.1.3, obtaining initialized particle swarms;

the specific steps of S5.2 are as follows:

s5.2.1, inputting a current position vector x of the particle;

determining a fitness function: an fitness function for evaluating the particle solution quality is determined. The fitness function should be related to transmission line communication private network failure conditions and be able to measure the performance of the solution.

The specific implementation and calculation process may vary depending on the details of the problem and the complexity of the fault analysis tree, so the specific steps for setting the fault analysis tree according to the cause and influence of the fault are as follows:

(1) Constructing a fault analysis tree: and constructing a fault analysis tree according to the specific condition of the transmission line communication private network fault. The fault analysis tree is a tree structure that describes the different fault conditions and the impact of the fault on system performance. The nodes of the tree represent fault types or states, i.e. particles, and the edges represent the relationship between the two nodes.

(2) Consider the extent of impact of a fault: when the fitness function value is calculated through the fault analysis tree, different weights can be given to each node so as to reflect the influence degree of different faults on the overall performance. For example, more severe faults may be given higher weights.

The method comprises the following steps:

3. multiplying the factors by the corresponding weights to obtain the weights of the nodes.

(3) Defining an evaluation index in a fault analysis tree: in the fault analysis tree, a respective evaluation index is defined for each node. These metrics may be performance parameters related to the performance or status of the failure, such as link reliability, network delay, data transfer speed, etc., and the evaluation metrics of each node should match the requirements and constraints of the performance or status of the failure.

(4) Evaluation starting from the root node of the tree: each node is evaluated step by step down starting from the root node of the fault analysis tree. And calculating the evaluation value of the node according to the current position vector x and the evaluation index related to the node. The evaluation value may be a real number for indicating the performance and health of the current location.

For example, assuming that there is a problem node, an evaluation index to evaluate the degree of influence thereof is an "influence factor", i.e., an evaluation value, may be defined as the degree of influence of the node on the time required for problem solving. During the data processing, data sets have been obtained which relate to the time required for the problem to be solved, for example the average solving time in different situations. According to the specific case, it is possible to choose to define the "influence factor" calculation formula as:

Influence factor = (solution time-optimal solution time)/optimal solution time;

wherein the solution time represents the time required for actually solving the problem, and the optimal solution time represents the shortest time required for solving the problem under ideal conditions. The influence degree of the node on the problem solving time can be quantified through the value of the influence factor obtained through calculation. If the impact factor is positive, indicating that the impact of the node on the solution time is positive; if the impact factor is negative, then it is indicated that the node's impact on the solution time is negative; if the impact factor is zero, it indicates that the node has no impact on the solution time.

(5) Calculating an fitness function value according to the relation among the nodes: in the fault analysis tree, the relationships between nodes may be parallel or serial. If there are parallel relationships between nodes, the evaluation values may be combined into one overall fitness value, for example by summing, weighted averaging, etc. If there is a serial relationship between nodes, the fitness value depends on the evaluation values of all nodes on the path.

When there is a serial relationship between nodes, the fitness value is calculated from the evaluation values of all the nodes on the path. The following is an example of a fault analysis tree:

Assume that there is a fault analysis tree comprising three nodes A, B and C arranged in a serial relationship, namely A→B→C. The evaluation value of node a is a_eval, the evaluation value of node B is b_eval, and the evaluation value of node C is c_eval. To calculate the fitness value of the entire fault analysis tree, the evaluation values of all nodes on the path need to be considered. In a serial relationship, fitness value may be understood as an evaluation of the effect from a starting node to a terminating node. The evaluation values of all nodes on the path have a comprehensive effect on the effect evaluation.

A simple calculation is to multiply the evaluation values of all nodes on the path, i.e. fitness value=a_eval b_eval c_eval. This calculation assumes that the evaluation values between the nodes are independent of each other and have no overlapping influence, so that their product can be taken as a comprehensive evaluation value of the entire path. The evaluation value of each node may represent the effect it contributes to or the ability to solve a problem and integrate them by multiplication. It should be noted that the actual situation may be more complex, and the relationships between nodes and the evaluation values may have more complex calculation modes, involving operations such as weighting, accumulation, etc., and the specific calculation mode is determined according to the requirements of the specific situation and problem.

(6) Iterative computation until leaf nodes: and according to the structure of the fault analysis tree, iteratively calculating until the leaf nodes according to the relation among the nodes. Finally, the obtained fitness function value f (x) reflects the performance evaluation of the current position vector x based on the fault analysis tree.

Consider a number of objective functions: in some cases, there are other objective functions that need to be optimized in addition to fitness function values, and these objective functions can be fused into a fault analysis tree and processed and calculated appropriately according to specific needs.

As for the iteration stop condition: it is determined when to stop the iterative computation and update the fitness function value of the particle. The stop may be determined based on convergence, the number of iterations, or the satisfaction of a predetermined condition, etc.

Meanwhile, according to the actual problem needs, the related parameters of the fault analysis tree and the weights of the evaluation indexes may need to be adjusted to obtain a better fitness function value calculation result, which may need to determine the optimal parameter setting through experiments and optimization.

Examples: in some problems, there may be other objective functions that need to be optimized in addition to fitness function values. Multiple objective functions refer to optimization objectives that involve more than one, each objective function describing a particular aspect or requirement in a problem.

Fusing multiple objective functions can help integrate optimization requirements in different aspects to get a more comprehensive, balanced solution. Fusing, processing and calculating a plurality of objective functions, specifically as follows:

1. fusion:

fusing multiple objective functions may employ the following approach:

(1) Weighted summation: each objective function is assigned a weight and the values of the individual objective functions are then weighted and summed to obtain the value of the composite objective function.

(2) Ideal point method: the multiple objective functions are converted into a single objective function, ideal points are found under the new single objective function, and then the distance between the actual solution and the ideal points is taken as the value of the integrated objective function.

2. And (3) treatment:

in fusing objective functions, priority and weight between each objective function need to be considered. Different weights can be set according to the requirements of specific problems so as to reflect the importance degrees of different objective functions. In addition, the range of the objective function can be limited by setting constraint conditions.

3. And (3) calculating:

different methods may be used for calculating the values of the plurality of objective functions, depending on the definition of the objective function and the manner of data processing. For each objective function, there may be an independent calculation method, or part of the calculation process may be shared.

For example, in a fault analysis tree, if one objective function is minimizing solution time and the other objective function is minimizing cost, a weighted summation approach may be used. Assuming that the evaluation value obtained at the solution time is t_eval and the evaluation value obtained at the cost is c_eval, the value of the integrated objective function may be defined as: complex objective function = w1×t_eval+w2×c_eval, where w1 and w2 are weights to solve for time and cost, respectively.

The number of particles in a population is not fixed and the optimum number is determined by a general rule, usually considering several factors such as: problem complexity, computational resources, convergence speed and efficiency requirements, experience and experimentation, and the like. If the optimum number is between 2-10 times the number of failed particles, e.g. 10 failed particles, as determined by rule of thumb, the number of particle clusters may be chosen to be 20-100.

The step of determining the fitness function is as follows:

1. determining influencing factors and weights: individual factors affecting the fitness of the particles are identified and listed, and a corresponding weight is set for each factor. For a transmission line communication private network fault, possible influencing factors comprise link reliability, network delay, data transmission speed and the like, a weight is determined for each factor according to the characteristics and the priority of the problem, the importance degree of the whole fitness is reflected, and the total weight is required to be 1 or distributed in proportion;

2. Normalization factor: for different ranges and units of influencing factors, normalization is required in order to fuse them into one fitness function, which can scale all factors to the same scale so that they have the same weight when computing the fitness function.

3. Calculating a fitness value: and calculating the fitness value of each particle according to the fitness function and the normalized factors. And multiplying each influence factor by the corresponding weight, and summing or combining according to the definition of the fitness function to finally obtain a numerical value which represents the fitness value of the particle solution.

s5.3.1 for each particle i ε [1, N ], the following is performed:

In the first calculation dx and dg are initialized to 0, since at this point no change in the position of the particles has occurred yet. Subsequently, as the iteration proceeds, dx and dg will change as the location is updated, reflecting the change in the particle in the search space.

in order to limit the velocity of the particles not to exceed the preset maximum value vmax, some processing may be performed when updating the velocity of the particles. The following is a common method:

1. calculating a speed update amount: and calculating a value of the speed to be updated according to the updating rule of the particle swarm optimization algorithm. Typically, this is determined by the current speed, the individual's historical optimal location, the global optimal location of the population, and some coefficients.

The following are several coefficients in common use:

(1) Learning factors (c 1 and c 2): the learning factor represents the relative importance of the particle to the individual's historical optimal location and the global optimal location of the population at the update rate. c1 represents the degree of importance of the individual historical optimal position, and c2 represents the degree of importance of the global optimal position of the group. Typically, c1 and c2 range from 0 to 2.

(2) Acceleration coefficient (w): the acceleration coefficient is used to control the inertia of the particles at the update speed. A larger acceleration coefficient increases the impulse of the particles, making them more dependent on their own velocity variation. Smaller acceleration factors attenuate the impulse of the particles, allowing them to be more referenced to the information of the optimal locations of individuals and populations. In general, the acceleration factor w has a value ranging from 0 to 1.

Illustrating:

assuming that the current speed of a certain particle is V, the individual history optimal position is P_best, and the group global optimal position is G_best, according to the updating rule of the particle swarm optimization algorithm, the speed updating quantity can be calculated by the following formula:

speed update amount = w v+c1 x range () (p_best-current_position) +c2 x range () (g_best-current_position);

where w is the acceleration factor, c1 and c2 are learning factors, random () is a random number generation function between [0, 1], and current_position is the current position of the particle. Through the above formula, the speed update amount can be calculated according to the information of the current speed, the individual history optimal position, the group global optimal position and the like.

2. Judging whether the speed exceeds the maximum value: for each dimension of the speed value it is checked whether it exceeds a maximum value vmax, if so, it is set to the maximum value to ensure that the speed does not exceed the set threshold.

3. Updating the particle velocity: and updating the speed value according to the calculated speed updating quantity. In this step, it is ensured that the component exceeding the maximum speed is limited and the other components are kept unchanged.

By means of the above steps, the speed can be limited when updating the particle speed to ensure that it does not exceed the preset maximum vmax. Therefore, the moving range of particles can be controlled, possible optimal solutions can be skipped due to the fact that the searching speed is too high, and meanwhile stability and convergence of an algorithm can be improved.

S5.3.2, returning updated particle velocity;

s5.3.3, repeating S5.3.1-S5.3.2 until all particle velocities are updated;

the method comprises the following specific steps:

s5.4.1 for each particle i ε [1, N ], the following is performed:

in practice, the purpose of updating the particle position is to move the particles to a more optimal solution space. The position of the particles can be updated by a simple addition operation to move in a more excellent direction according to the current speed and position of the particles. It should be noted that boundary conditions and constraints need to be considered in updating the particle positions to ensure the validity and feasibility of the algorithm. The frequency and manner of updating the particle location may be set and optimized according to the particular problem and scenario. In each iteration process, the positions and speeds of all particles are updated until the preset maximum iteration times are reached or a stop condition is met.

S5.5.1 for each particle i ε [1, N ], the following is performed:

The historical optimal fitness function value f (pi) is obtained by comparing the fitness function value of the particle during the previous iteration with the current optimal value. The method comprises the following specific steps:

(1) In the initial stage, the historical optimal fitness function value f (pi) for each particle is initialized to the current fitness function value f (x), and because this is the first iteration, the historical optimal fitness function value is equal to the current fitness function value.

(2) In a subsequent iteration process, the optimal value is updated for each particle by comparing the current fitness function value f (x) with its historical optimal fitness function value f (pi).

(3) If the current fitness function value f (x) is better than the historical optimal fitness function value f (pi), the historical optimal fitness function value f (pi) is updated to the current fitness function value f (x), so that the historical optimal fitness function value f (pi) of the particle can store the best fitness function value encountered.

(4) In the subsequent iteration, each particle will compare with its historical optimal fitness function value according to its fitness function value, and continuously update the historical optimal fitness function value, and the optimal fitness function value in the whole particle swarm will be maintained as a global optimal fitness function value, which may be called f (pg).

In this way, the particle swarm algorithm can continually search and update the historical optimal fitness function value to find a globally optimal solution. In each iteration, by comparing the current fitness function value with the historical optimal fitness function value, it can be determined whether the current solution is better than the past optimal solution, and updated accordingly.

In each iteration process, the historical optimal positions and the global optimal positions of all particles are updated to reflect the optimal solution in the current search space. If the fitness function value of the current position is better than the historical optimal fitness function value, updating the current position to be the historical optimal position; and if the fitness function value of the current position is better than the global optimal fitness function value, updating the current position to be the global optimal position. It should be noted that in updating the historical optimal position and the global optimal position, the influence of a plurality of factors, such as the number of particles, the number of iterations, the construction of the fitness function, and the like, need to be considered. Meanwhile, the historical optimal position and the global optimal position are required to be stored and managed so as to support subsequent particle updating operation. In practical applications, sufficient experimentation and verification is required to determine the best historical and global optimal location update policies and methods.

The method specifically comprises the following steps:

and judging whether the current iteration number reaches the maximum iteration number max_iter or whether the searched global optimal position reaches a preset threshold value threshold.

In each iteration process, the historical optimal position and the global optimal position are updated according to the current particle position and the fitness function value, and meanwhile whether the termination condition is met or not needs to be judged. If the iteration number reaches the maximum iteration number or the searched optimal solution reaches a preset threshold, the optimal solution is considered to be found, the algorithm can be ended, and a result is output; otherwise, the next iteration is continued until the termination condition is satisfied. It should be noted that in practical applications, the setting of the termination condition needs to be designed and optimized according to specific problems and scenarios. Generally, the maximum iteration number and the optimal solution threshold can be set according to the difficulty and the scale of the problem so as to achieve the optimal searching effect. Meanwhile, the convergence and stability of the algorithm are also required to be considered, and the phenomenon of premature convergence or oscillation is avoided. In practical applications, sufficient experimentation and verification is required to determine the optimal termination condition setting method and parameter settings.

Through the steps, the evaluation and prediction of the transmission line communication private network fault can be realized. It should be noted that the particle swarm optimization algorithm needs to be designed and optimized in combination with specific scenes and problems to ensure the feasibility and effectiveness of the algorithm. In practice, sufficient experimentation and verification is required to determine the best methods and parameter combinations. Meanwhile, the particle swarm optimization algorithm can be combined with other optimization algorithms, so that the accuracy and the efficiency of evaluation and prediction are further improved.

And the particle swarm optimization algorithm outputs a result to update a multisource information knowledge graph, so that the diagnosis and prediction of the transmission line communication private network faults are rapid and convenient: the method solves the problems of low fault diagnosis speed and no potential fault prediction in the existing power transmission line communication private network, and can effectively correlate and inquire a system and a module which have faults by utilizing a knowledge graph intelligent sub-graph matching algorithm, and predict possible faults of the same operating environment, the same system and the same module in advance.

The specific process for updating the initial transmission line communication private network knowledge graph according to the optimized result is as follows:

s5.7, determining the entity to be updated and the attribute and the associated information thereof according to the output result of the particle swarm optimization algorithm.

The output result of the particle swarm optimization algorithm is usually an optimal solution or a near optimal solution. For example, in an application of equipment failure prediction, the goal of the optimization may be to find the best parameter configuration or model weights to predict the failure to the greatest degree of accuracy. Thus, the specific manner in which the entities and relationships that need to be updated are determined based on the output results of the algorithm will vary depending on the optimization objectives. In the example of device failure prediction, based on the output results of the algorithm, the entity attributes (e.g., device status, sensor data, etc.) and relationships associated with the failure prediction may be updated for further analysis, decision-making, and scheme-making.

S5.8, carrying out standardization processing on the entity and relation which need to be updated. For example, information such as device names, serviceman names, etc. is standardized to ensure consistency and queriability of entity and attribute information in a knowledge graph.

And S5.9, updating the initial power transmission line communication private network knowledge graph according to the new entity and relation information to obtain a new power transmission line communication private network knowledge graph. For example, new entity and relationship information may be added to the knowledge-graph using techniques such as graph databases, and indexing and querying operations may be performed.

And S5.10, verifying and updating the new power transmission line communication private network knowledge graph. For example, the updated knowledge graph can be verified by means of data quality inspection and the like, so that the accuracy and reliability of the knowledge graph are ensured. If an error or inconsistent place is found, timely correction and updating are needed.

The correction and updating of the knowledge graph requires the following steps:

(1) Where the positioning is wrong or inconsistent: errors or inconsistencies in the knowledge-graph are located by data quality inspection, expert evaluation, or other methods. This may involve looking up entities, relationships or attributes that do not match or meet specifications with existing data.

(2) Analyzing the cause of the error or inconsistency: for errors or inconsistencies found, an analysis is performed to determine the cause thereof. The process may include looking up data source problems, human entry errors, algorithm-generated errors, and the like.

(3) Correcting errors or inconsistent points: and correcting the error or inconsistent points according to the analysis result. Specific operations may include deleting invalid or erroneous data, modifying values or labels of attributes or relationships, correcting names or descriptions of corresponding entities, and so forth.

(4) Updating related entities and relationships: after correcting the error or inconsistent points, related entities and relationships need to be updated to ensure the integrity and consistency of the knowledge-graph. This may involve updating attributes of the associated entity, redefining the associated information, adding new associated information, etc.

(5) And verifying the updated knowledge graph: and verifying the corrected and updated knowledge graph to ensure the correction accuracy and integrity. The updated knowledge-graph may be validated using data quality inspection, expert evaluation, logical reasoning, and the like.

(6) Periodic maintenance and updating: knowledge maps are a dynamic data set that requires periodic maintenance and updating. The timeliness and accuracy of the knowledge graph are maintained through regular data quality inspection, expert evaluation, user feedback and the like.

Meanwhile, the scale and the complexity of the knowledge graph are required to be considered so as to ensure the efficiency and the stability of the updating operation. In practical applications, sufficient experimentation and verification is required to determine the best update strategy and method.

Considering the scale and complexity of the knowledge graph is an important factor in ensuring the efficiency and stability of the updating operation. The following methods and strategies can be used to handle large-scale and complex knowledge-graph updates:

(1) Incremental update: aiming at the large-scale knowledge graph, the efficiency can be improved by adopting an incremental updating mode. Incremental updating refers to updating only the changed part, and does not need to completely recalculate the whole knowledge graph. By detecting changes, incremental calculations, and update operations, excessive time and resources are avoided.

(2) Batch processing: aiming at complex knowledge graph updating, the updating operation is processed in batches, so that the efficiency and the stability can be improved. The update task is divided into a plurality of small tasks and is executed batch by batch, so that the burden of single update operation can be reduced, and the problem caused by processing a large amount of data at one time is avoided.

(3) Parallel processing: the knowledge graph updating speed can be accelerated by using parallel computing and distributed processing technology. Dividing the update task into a plurality of sub-tasks and executing on a plurality of computing resources simultaneously may increase the efficiency of the update operation.

(4) Index optimization: optimizing the index of the knowledge graph can improve the performance of the updating operation. Through reasonable index design and optimization, the time complexity of query and update operations can be reduced, and the data access efficiency is improved.

(5) Data partitioning and slicing: the large-scale knowledge graph is divided into a plurality of small data partitions or fragments, so that parallel processing and incremental updating can be facilitated, each partition or fragment can be independently updated, mutual conflict and dependence are reduced, and updating efficiency and stability are improved.

(6) Monitoring and fault recovery mechanisms: aiming at large-scale and complex knowledge map updating, a monitoring and fault recovery mechanism is required to be implemented, the execution condition of the updating operation is monitored in time, and the problems and errors are treated and repaired in time so as to ensure the stability and reliability of the updating operation.

The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.

Claims

1. The method for evaluating the communication private network fault based on the particle swarm optimization and the knowledge graph optimization is characterized by comprising the following steps:

2. The method for evaluating a communication private network fault based on a particle swarm optimization and a knowledge graph according to claim 1, wherein in S2, text extraction is performed on the unstructured data to form a first triplet with "entity, first attribute value" as a combination, and the specific process is as follows:

3. The method for evaluating a fault of a communication private network based on a particle swarm optimization and a knowledge graph according to claim 1, wherein the first attribute comprises a fault description, a fault feature, a fault analysis, a fault occurrence time and a fault cause of an entity;

4. The method for evaluating the failure of the communication private network based on the particle swarm optimization and the knowledge-graph optimization according to claim 1, wherein the optimization of the training set data by the particle optimization algorithm is specifically:

5. The method for evaluating the failure of the communication private network based on the particle swarm optimization and the knowledge-graph optimization according to claim 4, wherein the specific process of S5.1 is as follows:

s5.1.2 for each particle i ε [1, N ], the following is performed:

a. randomly generating an initial position xi and an initial speed vi;

s5.1.3, obtaining initialized particle swarms;

The specific steps of S5.2 are as follows:

s5.2.1, inputting a current position vector x of the particle;

6. The method for evaluating the fault of the optimized communication private network based on the particle swarm optimization and the knowledge-graph according to claim 5, wherein the specific steps of calculating the fitness function value through the fault analysis tree are as follows:

The method comprises the following steps:

A5, calculating an fitness function value according to the relation among the nodes: in the fault analysis tree, the relation between the nodes comprises parallel or serial, if the nodes are in parallel relation, the evaluation values of the two nodes are combined into an overall fitness value in a summation or weighted average mode, and if the nodes are in serial relation, the fitness value depends on the evaluation values of all the nodes on the path;

7. The method for evaluating the failure of the communication private network based on the particle swarm optimization and the knowledge-graph optimization according to claim 6, wherein the specific process of S5.3 is as follows:

s5.3.1 for each particle i ε [1, N ], the following is performed:

s5.3.2, returning updated particle velocity;

s5.3.3, repeating S5.3.1-S5.3.2 until all particle velocities are updated;

the specific steps of S5.4 are as follows:

s5.4.1 for each particle i ε [1, N ], the following is performed:

8. The method for evaluating the failure of the communication private network based on the particle swarm optimization and the knowledge-graph optimization according to claim 7, wherein the specific steps of S5.5 are as follows:

s5.5.1 for each particle i ε [1, N ], the following is performed:

9. The method for evaluating a failure of a communication private network based on a particle swarm optimization and a knowledge-graph optimization according to claim 8, wherein the determining whether the termination condition is satisfied specifically comprises:

10. The method for evaluating the fault of the communication private network based on the particle swarm optimization and the knowledge graph according to claim 9, wherein in S5, the specific process of updating the knowledge graph of the communication private network of the initial transmission line according to the optimized result is as follows: