CN118278914A - Method for realizing equipment fault rush repair based on GIS (geographic information System) - Google Patents

Abstract

The invention discloses a method for realizing fault rush-repair based on GIS system equipment, which comprises the steps of collecting and preprocessing the existing fault data and constructing a fault database; constructing a device fault prediction model, training the device fault prediction model by taking fault data as input data, and outputting fault prediction information, wherein the fault prediction information comprises fault positions and occurrence probabilities of each future period; constructing and calling a GIS module based on the fault prediction information, planning the position of the optimal rush-repair base station, and forming a rush-repair base station topology network; and acquiring real-time fault information, calculating the nearest repair base station and the optimal repair route by adopting a GIS module, forming a repair scheduling data packet, and transmitting the repair scheduling data packet to the repair base station. According to the invention, through integrating the GIS system and the machine learning model, a comprehensive equipment fault rush-repair flow is provided, future fault events are effectively predicted, an optimal rush-repair route is calculated, and meanwhile, the stability and the safety of the system are ensured.

Description

Method for realizing equipment fault rush repair based on GIS (geographic information System)

Technical Field

The invention relates to a related technology of electric power automation, in particular to a method for realizing equipment fault rush repair based on a GIS (geographic information system).

Background

The railway power system is an important component of railway transportation, provides power and signals for a train, and ensures safe and efficient operation of the railway. However, railway power systems are also faced with various threats such as cable breakage, equipment overload, internal short circuits, etc., which cause interruption of power supply, affect normal operation of trains, and even cause accidents. Therefore, the method is an important task for the maintenance and management of the railway power system. The main method for processing the faults of the power equipment at present is to rely on manual line inspection to identify fault sections, manually fill in operation ticket contents and dispatch field personnel to arrive at the field to carry out fault emergency repair. This conventional failure handling approach suffers from a number of problems and drawbacks, mainly: the manual line inspection to identify the faulty section requires a lot of time and human resources, which causes delay in the fault handling time, further causing train delays and passenger dissatisfaction. The manual filling of the contents of the operation ticket has potential error risks, such as wrong judgment in a failure zone, detouring or arrival at a wrong place, misoperation and the like. This not only increases the complexity of the fault handling but may also pose a threat to the safety and availability of the railway system. When the field personnel arrive at the field to carry out fault emergency repair, the manual navigation or the search is needed, so that the response speed and the operation efficiency are reduced, and the fault recovery time is prolonged.

In order to solve these problems and defects, in recent years, some researchers have proposed a device fault repair method based on GIS (geographic information system). The GIS-based equipment fault rush-repair method is an innovative method for monitoring the state of power equipment in real time by utilizing a GIS system, accurately identifying fault points and rapidly deploying maintenance teams. The method provides a new view for fault management of railway power equipment and makes an important contribution to sustainable operation of modern railway systems. However, this method also has some problems or drawbacks, requiring further investigation and improvement, mainly: algorithms generated by fault diagnosis and operation methods need to consider various factors, such as the topology structure, fault type, fault influence range, operation rules and the like of the power system, and research and optimization on the aspects are lacking at present. The GIS system needs to interact with the SCADA system, the communication system and the navigation system, and needs to ensure the stability and the safety of the system so as to avoid the problems of data delay or leakage and the like. For the security guarantee method, the current research is relatively few.

There is therefore a need to develop innovations to address the above-mentioned problems with the prior art.

Disclosure of Invention

The invention aims to provide a method for realizing fault rush-repair based on GIS system equipment, which aims to solve the problems existing in the prior art.

The technical scheme is that the method for realizing the fault rush repair of the equipment based on the GIS system comprises the following steps:

S1, acquiring and preprocessing existing fault data, and constructing a fault database;

s2, constructing an equipment fault prediction model, training the equipment fault prediction model by taking fault data as input data, and outputting fault prediction information, wherein the fault prediction information comprises fault positions and occurrence probabilities of each future period;

step S3, constructing and calling a GIS module based on the fault prediction information, planning the position of the optimal repair base station, and forming a repair base station topology network;

and S4, acquiring real-time fault information, calculating the nearest repair base station and the optimal repair route by adopting a GIS module, forming a repair scheduling data packet, and transmitting the repair scheduling data packet to the repair base station.

The invention has the beneficial effects that through integrating the GIS system and the advanced machine learning model, a comprehensive equipment fault rush-repair flow is provided, future fault events are effectively predicted, the optimal rush-repair route is calculated, and meanwhile, the stability and the safety of the system are ensured.

Drawings

Fig. 1 is a flow chart of the present invention.

Fig. 2 is a flowchart of step S1 of the present invention.

Fig. 3 is a flow chart of step S2 of the present invention.

Fig. 4 is a flowchart of step S3 of the present invention.

Fig. 5 is a flowchart of step S4 of the present invention.

Fig. 6 is a flow chart of an embodiment of the present invention.

Detailed Description

As shown in fig. 1, a method for implementing fault rush-repair based on a GIS system device is provided, which includes the following steps:

S1, acquiring and preprocessing existing fault data, and constructing a fault database;

s2, constructing an equipment fault prediction model, training the equipment fault prediction model by taking fault data as input data, and outputting fault prediction information, wherein the fault prediction information comprises fault positions and occurrence probabilities of each future period;

step S3, constructing and calling a GIS module based on the fault prediction information, planning the position of the optimal repair base station, and forming a repair base station topology network;

and S4, acquiring real-time fault information, calculating the nearest repair base station and the optimal repair route by adopting a GIS module, forming a repair scheduling data packet, and transmitting the repair scheduling data packet to the repair base station.

In this embodiment, by collecting existing fault data and performing preprocessing, accuracy and availability of the data can be ensured, and the fault database is constructed for storing and managing fault information, which is important for subsequent fault analysis and prediction; future fault events, including fault location and probability of occurrence, can be predicted by training a model using historical fault data, this predictive capability enabling the utility to be prepared in advance, thereby reducing the impact of faults on users; the GIS module is utilized to plan the optimal rush-repair base station position and the topology network, so that the resource allocation can be optimized, the quick and effective response can be ensured when faults occur, and the response speed and the service quality of the rush-repair work are improved; the real-time fault information is acquired and the optimal repair route is calculated, so that the repair work efficiency is further improved, and the repair scheduling data packet obtained through the GIS module calculation can ensure that a repair team reaches a fault site at the highest speed.

According to the embodiment, through big data analysis and machine learning technology, the prediction model can accurately predict the time and place of occurrence of faults, so that measures are taken in advance; through the use of the GIS module, rush repair resources, including personnel and equipment, are planned more reasonably, so that the rush repair resources can be rapidly put into use when needed; through real-time data processing and optimized rush repair route planning, a rush repair team can reach a fault site more quickly, and the power failure time is reduced; the systematic rush-repair flow can improve the reliability of service and the satisfaction of users, and enhance the public image of the power company; unnecessary maintenance work and cost can be reduced and the operation efficiency can be improved by predicting faults and optimizing the rush-repair flow.

As shown in fig. 2, according to an aspect of the present application, the step S1 is further:

Step S11, collecting fault information including fault time, fault type, fault position and fault reason through a distribution network automation system;

Step S12, data cleaning is carried out on fault information, repeated, wrong and invalid data are removed, and valid data are reserved;

s13, carrying out data analysis on fault information, and extracting fault characteristic variables including fault frequency, fault duration, fault influence range and fault occurrence environment factors;

And step S14, storing the fault information and the fault characteristic variable in a fault database, and providing data support for subsequent fault prediction and rush repair.

In a further embodiment, when the system detects a fault, the time, type (e.g., short circuit, overload, etc.), location, and possible cause (e.g., equipment aging, external disturbances, etc.) of the fault is automatically recorded; the system will transmit these fault information to the central processing server where the data cleaning process includes identifying and deleting duplicate records, correcting erroneous information such as erroneous time stamps or location coordinates, and eliminating false alarms due to sensor faults or other non-grid problems; the cleaned data can be used for data analysis, calculating the frequency of faults, analyzing the duration of the faults, evaluating the influence range of the faults (such as how many users are influenced and how large area of power failure is caused), and considering the environmental factors (such as weather conditions, temperature and the like) when the faults occur; the analyzed fault information and characteristic variables are stored in a fault database. This database can be accessed by engineers and decision makers of the utility company for the creation of preventive measures and emergency plans. If the data shows that the frequency of failure in a certain area is abnormally high, the company may decide to upgrade or replace the equipment therein. If the analysis indicates that a certain type of fault is associated with a particular weather condition, the company may be ready in advance for similar weather forecast.

In the embodiment, the power grid state is monitored in real time through a distribution network automation system, so that an important basis is provided for rapid positioning and processing of faults; the collected fault information is subjected to data cleaning, so that the rest data are ensured to be accurate and reliable; through deep analysis of fault information, feature variables of faults are extracted, so that understanding of the occurrence mode and cause of the faults is facilitated, and scientific basis is provided for establishment of preventive measures; the cleaned and analyzed fault information and characteristic variables thereof are stored in a fault database, and data support is provided for fault prediction and rush repair decision, so that a power grid operator can more accurately predict future possible faults based on historical data and a statistical model, and a corresponding coping strategy is formulated.

Further, step S14 specifically includes:

step S141, inputting the fault information and the feature variable x= { X ₁,x₂,...,x_n } into the graph database, where X _i represents the i-th fault entity.

In step S142, for each fault entity x _i, attribute information a _i＝{a₁,a₂,...,a_m is extracted, such as fault time, fault type, etc.

Step S143, for each pair of fault entities xi and x _j, calculating similarity si _j between the fault entities xi and x _j, wherein the similarity includes cosine similarity and Jaccard similarity;

step S144, according to the similarity matrix s= (S _ij) n×n, using a graph embedding algorithm, including Node2Vec, learning the low-dimensional vector representation z= { Z ₁,z₂,...,z_n } of the fault entity:

minZ∑i＝1ⁿΣj＝1ⁿ(sij-zi^Tzj)²。

step S145, grouping fault entities in a vector space by using a clustering algorithm comprising K-Means and DBSCAN, and finding potential fault modes and rules.

In the present embodiment, x= { X ₁,x₂,...,x_n }, which represents a set of fault variables extracted from the historical fault data, each X _i represents one fault variable such as a fault code, a fault component, a fault phenomenon, and the like.

In this embodiment, a graph database is established, fault information and feature variables are input into the graph database, and each fault entity has a set of attribute information, such as fault time, fault type, influence range, etc.; calculating the similarity between each pair of fault entities by using cosine similarity and Jaccard similarity, and constructing a similarity matrix, wherein the matrix is used as the input of a graph embedding algorithm, a Node2Vec algorithm is used for learning low-dimensional vector representations of the fault entities, the vector representations capture complex relations between the fault entities, and high-dimensional data are converted into a low-dimensional space which is easier to process; finally, K-Means and DBSCAN clustering algorithms are applied in this vector space, which are able to group fault entities according to their vector representation, revealing different fault patterns, which information can help us predict and prevent future faults, optimize maintenance plans, reduce outage time, and improve quality of service when it is found that a certain type of fault always occurs in a certain time period or on a certain device.

According to the embodiment, through the use of the graph database, a large number of fault entities and attributes thereof can be effectively managed and inquired, and the data processing efficiency is improved; the similarity between fault entities is evaluated from different angles by calculating cosine similarity and Jaccard similarity, so that a more comprehensive visual angle is provided for subsequent analysis; using graph embedding algorithms such as Node2Vec, failed entities can be mapped into a low-dimensional vector space, helping to reveal potential relationships between failed entities; through clustering algorithms such as K-Means and DBSCAN, faulty entities can be grouped in vector space to discover potential failure modes and laws. Therefore, the fault data analysis method and the fault data analysis device can help understand and analyze fault data, can provide actual operation and maintenance decision support, and can improve stability and efficiency of the power system, and the fault data analysis method and the fault data analysis device are of great significance in improving the technical level of the power automation field.

As shown in fig. 3, according to an aspect of the present application, the step S2 is further:

s21, selecting a preset machine learning algorithm, and constructing an equipment failure prediction model;

S22, obtaining fault data from a fault database, dividing the data into a training set and a testing set, training and verifying a device fault prediction model, optimizing model parameters, and improving model accuracy;

And S23, predicting equipment faults within a certain period of time in the future by using the trained equipment fault prediction model, and outputting fault prediction information comprising the fault position and occurrence probability of each future period of time.

In the embodiment, firstly, fault data including information such as fault time, type and influence range are extracted from a fault database, then the data are divided into a training set and a testing set, and the data are divided according to the proportion of 70% of the training set and 30% of the testing set; training the selected machine learning algorithm by using a training set, optimizing model parameters by using technologies such as cross verification and the like, improving the accuracy and generalization capability of the model, verifying the model by using a testing set, and evaluating the prediction performance of the model; and finally, predicting equipment faults in a certain period of time in the future by using the trained model.

According to the embodiment, the equipment fault prediction model is constructed by applying the machine learning algorithm, so that potential equipment faults in the power system can be effectively predicted. The prediction model can help operation and maintenance personnel to identify possible faults in advance, so that preventive measures are taken, the occurrence of unexpected power failure is reduced, and the stability and reliability of the power system are improved. Through analysis and study of historical fault data, the model can predict fault positions and occurrence probabilities in a certain period of time, and data support is provided for maintenance and optimization of the power system.

In another embodiment of the present application, the step S21 of constructing the equipment failure prediction model specifically includes:

step S211, the failure dataset d= { (x ₁,y₁),(x₂,y₂),...,(x_N,y_N) } is divided into K subsets D ₁,D₂,...,D_K.

In step S212, a base model f _k(x), such as a random forest, a support vector machine, etc., is trained on each subset D _k.

In step S213, a meta-model g (x), such as logistic regression, neural network, etc., is trained using all the prediction results { f _1(x),f_2(x),...,f_K(x) } of the basic models as new features.

In step S214, the prediction output of the integrated model is F (x) =g (F _1(x),f_2(x),...,f_K(x)).

In this embodiment, by integrating multiple basic models, more data features and modes can be captured, which can more accurately reflect the complexity of fault occurrence than a single model, each basic model may only learn limited information from the data, but when they merge, their prediction results will be complementary, thereby improving overall prediction accuracy; the sensitivity of different basic models to noise and abnormal values in data is different, and the sensitivity to the abnormal values can be reduced by combining the models through the integrated learning, so that the stability of the models in the face of different types of data is improved; the present embodiment allows the use of various types of basic models, such as random forests, support vector machines, etc., which means that the most appropriate model can be selected according to the specific problem, and in addition, as new data is acquired, more basic models can be easily added to improve the performance of the meta model; a single model may overfit training data, but integrating multiple models may reduce this risk; the meta model can better generalize unseen data by learning the prediction result of the basic model as a characteristic, and more accurate fault prediction means that an operation and maintenance team can more effectively allocate resources for preventive maintenance, thereby reducing the occurrence of unexpected power failure and improving the overall operation efficiency of the power system.

In another embodiment of the present application, step S22 further includes introducing a data enhancement technique to solve the problem of unbalanced fault data, specifically:

Step S221, for each minority class sample x _i, randomly selecting one sample x _i' from k nearest neighbor samples.

Step S222, randomly selecting a point on the line between x _i and x _i' as a new synthesized sample:

Step S223, x _new＝x_i+λ(x_i'-x_i), λ ε [0,1].

Steps S221 and S222 are repeated until the minority class sample count reaches equilibrium.

In the present embodiment, in the power system, the fault data tends to be unbalanced, because the state data of the normal operation is far more than the state data of the fault, and the unbalance may cause the model to be biased to the majority class, so that the prediction capability of the minority class is reduced. Therefore, the introduction of data enhancement techniques is of paramount importance. According to the embodiment, the synthetic sample is created, so that the feature space of the minority class is enlarged, and the model can learn the features of the minority class better. Not only is the accuracy of the model improved, but the generalization capability of the model is also improved, as it increases the sample diversity when the model is trained.

As shown in fig. 4, according to an aspect of the present application, the step S3 is further:

step S31, determining the position of equipment with possible faults, the probability and the influence degree of the faults according to the fault prediction information;

Step S32, geographic information of equipment positions and surrounding environments, including roads, traffic and buildings, is acquired by utilizing a GIS system;

step S33, calculating the optimal position of the first-aid repair base station by utilizing a GIS system according to the equipment position and the geographic information, so that the first-aid repair base station can cover all equipment which possibly fails and is nearest to the failed equipment;

And step S34, forming a topology network of the rush-repair base stations, namely a connection relation among the rush-repair base stations, by utilizing a GIS system according to the positions of the rush-repair base stations so as to facilitate information transmission and coordination among the rush-repair base stations.

In this embodiment, based on the information provided by the fault prediction model, determining the position of the equipment which is likely to generate faults, which involves analyzing the probability and the influence degree of the faults so as to give priority to the equipment which has the greatest influence on the stability of the power grid, so that the operation and maintenance team can be assisted in identifying key equipment and high-risk areas, and thus, preparation is made in advance; the GIS system is utilized to collect geographic information around the equipment position, including roads, traffic, buildings and the like, and detailed terrain and infrastructure information is provided for the rush repair team through the collection, cleaning and integration of the geographic data, which is important for planning the rush repair route and distributing resources; analyzing the collected geographic information by using a GIS system, and calculating the optimal position of the rush-repair base station by combining the position of the fault equipment, so that the rush-repair base station can be ensured to quickly respond to any possible fault and cover all key areas; based on the positions of the rush-repair base stations, a topological network between the rush-repair base stations is constructed by utilizing a GIS system, and the information transmission and coordination between the rush-repair base stations are promoted through network design and connection relation establishment and optimization, so that the efficiency and effect of the rush-repair work are improved.

The embodiment provides powerful support for equipment fault first-aid repair in the field of power automation through accurate data processing and efficient technical application. The method not only improves the speed and accuracy of fault response, but also optimizes the configuration and use of resources, and finally aims to reduce the power failure time and ensure the reliability and safety of power supply.

In another embodiment of the present application, step S33 specifically includes: constructing a multi-objective optimization model NSGA-II:

Wherein, the optimization objective is:

rush repair response time (T): the time from the occurrence of the fault to the arrival of the repair team at the point of the fault, the goal is to minimize the average response time,

Where t _i is the response time of the ith failure point and N is the total number of failure points.

Cost benefit (C): the overall cost of building and maintaining the rush repair base station, the goal is to minimize the overall cost,

C＝C _{Establishment of} +C _{Maintenance of}

Wherein, C _{Establishment of} is the cost of establishing the rush repair base station, and C _{Maintenance of} is the cost of maintaining the rush repair base station.

Balance of resource allocation (R): the allocation of rush repair resources among different base stations is aimed at optimizing the resource allocation to achieve efficient utilization,

Wherein r _i is the resource amount of the ith rush-repair base station, M is the total number of the rush-repair base stations, and r-is the average resource amount.

Step S331, randomly generating a certain number of solutions, wherein each solution represents a configuration scheme of the rush repair base station;

the population P0 is initialized, and each individual represents a repair base station layout scheme and is represented by a binary code string.

Step S332, calculating fitness value for each solution according to the above-defined objective function;

For each individual x in the population, its fitness value f _1(x),f_2(x),...,f_M(x) over M targets is calculated.

Step S333, performing non-dominant sorting according to the fitness value of the solutions, and calculating the congestion degree of each solution to maintain the diversity of the solutions.

And sorting and selecting the populations by using non-dominant sorting and crowding operators to obtain a new parent population Pt.

Step 334, performing crossover and mutation operations according to the non-dominant ranking and the crowding degree selection solution to generate a new generation population.

And performing crossover and mutation operation on the parent population Pt to generate a child population Qt.

Step S335, combining the parent population Pt and the offspring population Qt to obtain a new population Rt, and repeating the steps 3 and 4 until the maximum iteration number is reached.

And step S336, selecting a non-dominant solution set as a Pareto optimal solution, namely an optimal rush repair base station layout scheme, in the final population.

In this embodiment, a certain number of solutions are randomly generated first, and a diversified starting point is provided for the multi-objective optimization algorithm, so that the possibility of finding the globally optimal solution is increased; then calculating a fitness value, wherein the fitness value reflects the coincidence degree of each solution to a preset objective function, and the fitness value is calculated to provide basis for subsequent selection, crossing and mutation operations so as to ensure that excellent genes are reserved; non-dominant sorting is carried out according to the fitness value of the solutions, the crowding degree of each solution is calculated, and the algorithm is prevented from converging to a local optimal solution too early by maintaining the diversity of the solutions; through the core operation of a genetic algorithm, a solution space is explored, and the diversity of the population is increased; gradually approaching the optimal solution through iterative search; finally, pareto optimal solutions are selected to provide a group of optimal solutions instead of a single solution, thereby providing more selection space for a decision maker.

The embodiment provides powerful support for equipment fault rush repair in the field of power automation through accurate data processing and efficient technical application, not only improves the speed and accuracy of fault response, but also optimizes the configuration and use of resources, and achieves the effects of reducing power failure time and guaranteeing the reliability and safety of power supply.

In another embodiment of the present application, step S33 is further:

step S33a, defining an optimization target: the rush repair response time is minimized, the cost benefit is maximized and the resource allocation is balanced;

the structure and fault data of the power network are encoded onto the quantum states. Let the power network be represented by graph g= (V, E), where V is the node set and E is the edge set. The characteristics of each node v and edge e are encoded as states of qubits.

The coded graph data is processed through a quantum variation circuit, and graph characteristics are extracted and updated through the operation of a quantum gate. And obtaining the optimal position and configuration scheme of the rush repair base station through quantum measurement.

Assuming that the quantum state |ψ (v) > represents the quantum encoding of node v, the quantum transformation circuit transforms it through parameterized quantum gate U (θ):

|φ(v)>＝U(θ)|ψ(v)>；

The final quantum state measurements are used to evaluate the balance of rush repair response time, cost effectiveness, and resource allocation to form an optimized objective function:

f(T，C，R)＝w₁·T+w₂·C+w₃·R；

Where w1, w2, w3 are weight coefficients for balancing the importance of different targets.

Step S33b, collecting fault information and environment change data of the power network in real time by utilizing a quantum sensor; and updating the quantum state of the power network diagram in real time based on QGNN model, and dynamically adjusting the position and the topological structure of the rush-repair base station.

The embodiment can fully utilize the parallelism of quantum computation and the high sensitivity of the quantum sensor, and realize the rapid evaluation and response of the power network state. The layout of the rush-repair base station can be dynamically adjusted according to actual conditions through real-time update of the quantum state, so that the rush-repair response time and resource allocation are optimized. Furthermore, the optimization capabilities provided by quantum computing can help find solutions that maximize cost effectiveness. The embodiment provides a high-efficiency, flexible and cost-effective solution for equipment fault rush repair in the field of power automation by combining a GIS system and a quantum technology, not only improves the speed and accuracy of fault response, but also optimizes the configuration and use of resources, is beneficial to reducing the power failure time and ensures the reliability and safety of power supply.

As shown in fig. 5, according to an aspect of the present application, the step S4 is further:

step S41, acquiring real-time fault information comprising time, position, type and reason of fault occurrence through a fault rush-repair app platform;

Step S42, calculating a nearest rush-repair base station to the fault equipment and an optimal rush-repair route from the rush-repair base station to the fault equipment by utilizing a GIS system according to the real-time fault information and the position of the rush-repair base station;

Step S43, selecting a proper unmanned aerial vehicle and an optimal flight route from the unmanned aerial vehicle to the fault equipment according to the real-time fault information and the unmanned aerial vehicle position by utilizing the unmanned aerial vehicle module,

Step S44, a GIS system and an unmanned aerial vehicle module are utilized to generate an emergency repair scheduling data packet according to an emergency repair base station topological network, wherein the emergency repair scheduling data packet comprises fault information, emergency repair base station information, emergency repair route information, emergency repair task allocation information, unmanned aerial vehicle information, flight route information and unmanned aerial vehicle task allocation information;

and step S45, issuing a rush-repair scheduling data packet to the rush-repair base station and the unmanned aerial vehicle through the fault rush-repair app platform to guide a rush-repair person and the unmanned aerial vehicle to carry out rush-repair work.

The embodiment obtains real-time fault information and calculates the nearest maintenance station and the optimal maintenance route by utilizing the GIS, thereby minimizing the downtime and ensuring the effective maintenance; by using the unmanned aerial vehicle to perform optimal route selection according to the real-time fault information, the time for starting maintenance is reduced; through the combination of GIS and unmanned aerial vehicle module, can produce maintenance scheduling data package, including the detailed information about trouble, maintenance station, route and task allocation, this comprehensive data package ensures that all relevant parties are fully aware of the situation and coordinate.

In another embodiment of the present application, step S44 specifically includes:

A state space S is defined, which represents the state of the emergency repair environment, such as the fault equipment position, the fault type, the emergency repair force distribution, etc.

An action space A is defined, which represents actions that each agent (repair force) can take, such as movement, repair, etc.

A reward function R (s, a) is defined, representing the immediate reward returned by the environment after taking action a in state s.

A cost function Q (s, a) is defined that represents the long-term jackpot expectation of taking action a in state s.

Approximating the cost function by a neural network Q (s, a; θ):

Where α is the learning rate and γ is the discount factor. Each agent uses the e-greedy policy to select action a _t according to the current state s _t, obtains the next state s _t+1 and rewards r _t of the environmental feedback, and updates the neural network parameters.

The present embodiment optimizes maintenance planning using long term rewards expectations by defining states, action spaces, rewards functions and cost functions to achieve intelligent and adaptive decisions.

The embodiment combines predictive analysis, real-time data processing, unmanned plane technology and advanced machine learning technology, realizes an efficient, sensitive and intelligent emergency repair process, reduces maintenance time of maintenance and management of a power system, optimizes resource allocation and improves overall service quality.

According to one aspect of the application, the fault prediction model is a GAT-LSTM-attention model, comprising:

The GAT unit is used for distributing different weights for each node in the graph so as to extract the spatial characteristics and the topological structure of the node, the GAT unit firstly carries out linear transformation on the characteristics of the node, then calculates the attention coefficient among the nodes, and finally carries out weighted summation on the characteristics of the node according to the attention coefficient to obtain new node characteristics;

the LSTM unit is used for distributing different states for each node in the graph so as to extract time sequence characteristics and history information of the node, firstly, calculating forgetting gates, input gates, output gates and memory cells of the node according to the characteristics of the node and the state of the last moment, and then updating the states and outputs of the node according to the values of the gates to obtain new node states and outputs;

And the ATTENTION unit is used for distributing different importance to each node in the graph so as to extract global characteristics and context information of the nodes, and firstly calculating the correlation between the nodes according to the output and the state of the nodes, and then carrying out weighted summation on the output of the nodes according to the correlation to obtain new node output.

According to an aspect of the present application, the step S34 is further:

Step S341, acquiring equipment position and geographic information by using a GIS system, and calling fault prediction information;

Step S342, constructing at least two first-aid repair base station calculation models, forming a first-aid repair base station integrated calculation model, and training, wherein the first-aid repair base station calculation models comprise a GA-FL model, a PSO-GRA model, a MOO-FS model and an ANN-FS model;

And step S343, after training is completed, the position of the optimal repair base station is calculated by taking the equipment position, the geographic information and the fault prediction information as the input of the calculation model of each repair base station, weighting is carried out, and a weighting result is output.

In the embodiment, a plurality of first-aid repair base station calculation models are built, and the positions of the best first-aid repair base stations can be calculated more accurately through integrating the calculation models and training; and the output of each rush-repair base station calculation model is weighted, and a weighted result is output, so that the advantages of different models are comprehensively considered, and the optimization effect of the final rush-repair base station position is improved.

Because a single topological structure can not adapt to complex and changeable fault environments and rush repair demands, a plurality of calculation models and weighting processing are introduced in the embodiment, and the flexibility and the accuracy of the position planning of the rush repair base station are improved.

According to an aspect of the present application, the process of generating the rush repair schedule packet in step S44 further includes:

Step S441, defining the targets and constraints of the rush-repair schedule, such as minimizing the rush-repair time, maximizing the rush-repair coverage rate and optimizing the rush-repair resource allocation;

s442, selecting DQN-LSTM to construct a rush repair scheduling model; historical fault data are obtained from a fault database, the data are divided into a training set and a testing set, the emergency repair scheduling model is trained and verified, model parameters are optimized, and model accuracy is improved;

Step S443, acquiring real-time fault information and a repair base station topology network by using a GIS system, taking the real-time fault information and the repair base station topology network as input of a repair scheduling model, and outputting a repair scheduling data packet comprising fault information, repair base station information, repair route information and repair task allocation information by using the trained repair scheduling model.

In the embodiment, the fault is ensured to be responded quickly by minimizing the rush repair time; maximizing the rush repair coverage rate and ensuring that all potential fault points can be treated in time; optimizing resource allocation, reasonably utilizing rush repair resources, avoiding waste, and restricting the availability of rush repair personnel and equipment, cost limitation, geography, traffic conditions and the like. And by adopting the DQN-LSTM model and combining the prediction capability of deep learning and the advantages of LSTM processing time sequence data, the DQN-LSTM model is trained through historical data, and parameters are optimized to improve the prediction accuracy and decision quality. And collecting fault information and topology network data of the rush-repair base station in real time by using a GIS system to form a rush-repair scheduling data packet containing the fault information, the rush-repair base station information, the rush-repair route and task allocation.

According to the embodiment, through advanced data processing and intelligent decision support, the efficiency and effect of power system equipment fault rush repair are remarkably improved, and the reliability of power supply and user satisfaction are improved.

According to one aspect of the present application, the step S12 further includes constructing a fault knowledge graph,

After the fault information is subjected to data mining, namely, after feature variables of the fault are extracted by utilizing clustering, classification and association rules, the fault information is subjected to knowledge graph construction, and semantic representation and knowledge reasoning of the fault data are constructed by utilizing entity extraction, relation extraction and entity linking methods.

The fault knowledge graph can reveal inherent links between different fault events, such as common causes, similar impact ranges, or frequently co-occurring fault types, which helps to understand fault patterns and potential risk factors; the fault knowledge graph can be used as a decision support tool to help a maintenance team to determine an optimal repair strategy, and by analyzing the fault knowledge graph, determining which faults are most likely to occur, and preferentially distributing resources according to the faults; by utilizing the reasoning capability of the knowledge graph, new fault modes and potential fault causes can be automatically discovered, which is very valuable for long-term maintenance strategies and improvement plans.

The embodiment remarkably improves the efficiency and effect of power system fault management by constructing a fault knowledge graph, and helps an electric company to better understand and cope with fault challenges by providing deep insight and advanced analysis capability.

In another embodiment of the present application, the method further includes step S15:

Based on the data processing of steps S11 to S14,

From the historical fault data, a fault variable set is extracted to form a characteristic variable x= { X ₁,X₂,...,X_n }, and each variable X _i is a fault entity and can be a fault code, a fault component, a fault phenomenon and the like.

And S151, learning a causal structure among variables by using a PC algorithm to obtain a Directed Acyclic Graph (DAG) G= (V, E). Wherein g= (V, E) represents a Directed Acyclic Graph (DAG) learned from causal structure, V is a node set, and represents a fault variable; e is an edge set, representing causal relationships. Based on the conditional independence test, edges are recursively deleted until there are no redundant edges. The condition independence can be measured in terms of mutual information:

Wherein I (X _i;X_j |s) represents mutual information for measuring conditional independence, S is a set of conditional variables, and X _i and X _j are two fault variables. If I (X _i;X_j |s) =0, then X _i and X _j are said to be independent given S.

Step S152, estimating the causal effect intensity between the fault variables using a causal reasoning algorithm on the causal graph G.

An intervention operator do (X _i＝x_i) is introduced, which indicates that the variable X _i is interfered, and the value of the intervention operator is set as X _i. The joint distribution of the dry prognosis is:

Where pa (X _j) represents the parent node set of X _j, do (X _i＝x_i) represents the intervention operator, represents the intervention on the variable X _i, and sets its value to X _i,P(X1,...,Xn|do(X_i＝x_i)) represents the joint probability distribution of the dry prognosis.

Step S153, calculating the causal effect P (y|do (x=x)) of the target fault variable Y on the cause fault variable X, i.e. the conditional probability distribution of Y in case the intervention X is X.

Wherein Y represents a target fault variable, i.e., a fault result of interest; x represents a cause fault variable, namely, a cause which can cause the target fault; p (y|do (x=x)) represents a causal effect, i.e. a conditional probability distribution of Y in case of an intervention X of X.

The present embodiment makes the set of fault variables extracted from the historical fault data not just for descriptive statistics, but into useful information that can be causally analyzed by using a PC algorithm and causal reasoning algorithm.

The PC algorithm generates a Directed Acyclic Graph (DAG) by learning the causal structure between variables. This DAG not only reveals the correlation between fault variables, but also reveals possible causal relationships between them. This is critical to understanding the inherent mechanism by which a fault occurs, and by the DAG it can be identified which components the particular fault code is typically caused by the fault, or which environmental factors are typically behind the fault phenomenon.

The causal reasoning algorithm estimates the causal effect strength between fault variables, not only can the correlation between variables be known, but also the strength of such correlation can be quantified. This is of great importance for fault prediction and repair decisions. When the influence of the fault of a certain component on the system is very large, the fault of the component can be preferentially treated under the condition of limited rush repair resources.

Calculating the conditional probability distribution of the target fault variables provides a way to evaluate the possible effects of different interventions. When considering replacement of a certain component, it can be predicted by calculation how much this intervention has an effect on reducing the probability of a specific failure occurring.

The embodiment not only enhances the analysis depth of fault data, but also provides scientific basis for formulating more effective fault prevention and rush repair strategies. This is extremely beneficial for improving the reliability and efficiency of the power system, reducing blackout time, and optimizing maintenance costs.

In another embodiment of the present application, step S2 is further:

s21, constructing a fault prediction model based on a fault knowledge graph

Step S211, converting the nodes and edges in the causal graph G into concepts and relations in the ontology, and constructing a fault knowledge graph { G }.

Step S212, extracting fault entities and attributes thereof from structured, semi-structured and unstructured data sources, such as sensor data, maintenance manuals, expert experiences and the like, mapping the fault entities and attributes thereof to ontology concepts, and enriching knowledge maps.

And S213, embedding the entities and the relations in the knowledge graph into a low-dimensional vector space by using a knowledge representation learning method TransE to obtain the distributed representation of the entities and the relations.

TransE for a triplet (h, r, t) there is h+r≡t based on the assumption of translational invariance. Wherein h, r and t are respectively embedded vectors of a head entity, a relation and a tail entity in the knowledge graph. The goal of learning is to minimize the loss function:

Wherein L represents a loss function of TransE model for learning embedded representations of entities and relationships; s is a real triplet set, S' is an error triplet set generated by negative sampling, gamma is a hyper-parameter, and d is a distance metric function [ x ] ₊ =max (0, x).

Step S22, knowledge-based fault diagnosis reasoning;

Step S221, given a set of observed fault phenomena o= { O1, O2.,. The term, om }, performing rule-based forward reasoning on the knowledge graph { G }, and finding a possible root cause fault C:

(O, { G } I-C })/({ G } I-O→C });

where C represents the possible cause of failure resulting from reasoning, I-represents the rules of reasoning.

Step S222, calculating the conditional probability P (C|O) of each candidate root factor by using a probabilistic logic reasoning method such as a Markov logic network:

Wherein P (c|o) represents the conditional probability of the root fault C given the fault phenomenon O; w _i is the weight of the ith rule in the Markov logic network, n _i (C, O) is the number of times the rule is satisfied given C and O, Z is a normalization factor used to ensure that the sum of probabilities is 1.

Step S223, selecting the root cause with the highest probability as a diagnosis result to generate an explanatory diagnosis report.

The present example improves the efficiency and accuracy of fault prediction and diagnosis by integrating multi-source data and applying advanced data analysis techniques, while also providing explanatory information that helps in decision support.

In another embodiment of the present application, step S3 may further be:

Step S31, defining a state space S, an action space a, a transition probability P and a reward function R of the markov decision process. Wherein, the state is expressed as a combination of (fault equipment position, fault type, rush-repair force distribution), and the action is expressed as a scheduling decision of each rush-repair force.

Step S32, designing a reward function, and considering the factors such as fault recovery time T _recovery, rush repair cost C _repair, equipment importance I _device and the like:

R(s_t,a_t)＝-αT_recovery-βC_repair+γI_device

Wherein R (s _t,a_t) represents the immediate prize obtained by taking action a _t in state s _t, and alpha, beta, gamma are weight coefficients.

Step S33, learning an optimal rush repair scheduling strategy pi by using the DQN:

Wherein Q (s, a) is an optimal action cost function representing a long-term jackpot desire to take action a in state s; pi(s) is the optimal action in state s, i.e., the action with the largest Q value.

And step S34, updating the model on line, and continuously optimizing the strategy according to the new fault data and the rush repair feedback.

The embodiment realizes the intellectualization and automation of the rush-repair scheduling by introducing an advanced algorithm and model, further improves the response speed and decision quality of the rush-repair work, reduces the operation and maintenance cost and enhances the stability and reliability of the system.

In a further embodiment, firstly, fault positioning and diagnosis are carried out, fault distance is automatically calculated and fault reasons are evaluated through information such as wave recording, fixed value and measured value, faults of a power system are rapidly and accurately diagnosed, power failure time and train delay are reduced, and service availability of users is improved; then automatically generating an operation method, automatically generating the operation method, the steps and the safety precautions by the system based on the failure cause, simplifying the creation process of the operation ticket, reducing the risk of human errors and being beneficial to ensuring the safety and the accuracy of the operation; then, an automatic operation ticket is generated, the fault position is automatically mapped to a GIS system, the operation ticket is generated and issued to a specific workshop, so that the time for creating and distributing the operation ticket is greatly saved, the operation efficiency is improved, and meanwhile, the workload of an operator is reduced; finally, automatic navigation is carried out, and through an automatic navigation function, an executive can quickly and accurately reach a fault place without relying on manual navigation or searching, so that the response speed is improved, the working time of an operator is reduced, and the normal operation of a power system is facilitated to be quickly recovered.

According to the embodiment, through an automatic fault diagnosis and operation method, the requirement of manual intervention is reduced, the occupation of human resources is reduced, and in addition, the waiting and searching time of personnel on site is also reduced through automatic navigation and operation ticket generation; meanwhile, fault positioning and automatic generation of operation steps greatly improve the speed and efficiency of fault processing, reduce power failure time, reduce train delay and improve the usability of a power system; the automatic process reduces the chance of human error, improves the accuracy and safety of operation, and is helpful for avoiding potential accidents and losses; the cooperation among different functional modules is the key of the application, and the integration of a fault positioning and GIS system and the cooperation among automatic navigation and operation ticket generation make the whole process more efficient.

The fault positioning and rapid generating operation method greatly improves the rush-repair speed of the power system and is beneficial to rapid service recovery; the shortening of the fault processing time can reduce train delay and improve the satisfaction of passengers; the automation flow reduces the manpower resource cost and the maintenance cost; the rapid and accurate fault handling is beneficial to improving the usability of the power system and ensuring the normal operation.

In a further embodiment, as shown in fig. 6, data interaction is performed with a railway power supply Scheduling Control (SCADA) system, fault data of the SCADA system are obtained in real time, and timeliness of the data is guaranteed; then developing fault diagnosis algorithms which can calculate fault distance according to wave recording, fixed value and measured value, evaluate fault reasons and also need to develop algorithms for automatically generating operation methods, steps and safety precautions; different components and technologies are integrated into one system, so that the system can run smoothly and can be integrated with the existing railway power system; and finally, testing and verifying the system in an actual environment, checking the accuracy and the performance of the system, ensuring that the system can meet the expected requirement, evaluating the performance of the system regularly, comparing the actual operation result with the expected result, searching for an improvement opportunity and optimizing.

The embodiment is used for maintaining and managing the railway power system, mainly aims at the fault rush-repair of the railway power system, can be widely applied to the railway industry, and improves the reliability and the rush-repair efficiency of the power system.

The embodiment maps the fault position automatically to a Geographic Information System (GIS) system, and can be applied to the GIS field, which is helpful for managing geographic information and spatial data and providing more accurate position information. In addition, the automatic navigation technology can be used in other fields, such as logistics, traffic management and the like, so as to help people navigate to a destination in a complex environment; the automatic fault diagnosis and rush-repair flow can be applied to maintenance and rush-repair of substations, transmission lines and distribution systems in the power industry, and the reliability of a power grid is improved.

The application provides a high-efficiency and reliable fault rush-repair solution for the field of power automation by integrating an advanced data processing technology and a GIS system. The system not only can improve the stability and the safety of the power system, but also can improve the user experience, and brings economic benefit and social value for power companies.

According to another aspect of the application, a fault rush-repair system based on a GIS system device comprises:

At least one processor; and

A memory communicatively coupled to at least one of the processors; wherein,

The memory stores instructions executable by the processor, the instructions being for execution by the processor to implement the implementation method for fault rush repair based on the GIS system device according to any one of the above technical schemes.

The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to the specific details of the above embodiments, and various equivalent changes can be made to the technical solution of the present invention within the scope of the technical concept of the present invention, and all the equivalent changes belong to the protection scope of the present invention.

Claims (9)

1. The method for realizing the fault rush-repair of the equipment based on the GIS system is characterized by comprising the following steps:

S1, acquiring and preprocessing existing fault data, and constructing a fault database;

s2, constructing an equipment fault prediction model, training the equipment fault prediction model by taking fault data as input data, and outputting fault prediction information, wherein the fault prediction information comprises fault positions and occurrence probabilities of each future period;

step S3, constructing and calling a GIS module based on the fault prediction information, planning the position of the optimal repair base station, and forming a repair base station topology network;

and S4, acquiring real-time fault information, calculating the nearest repair base station and the optimal repair route by adopting a GIS module, forming a repair scheduling data packet, and transmitting the repair scheduling data packet to the repair base station.

2. The method for implementing fault rush-repair based on the GIS system equipment according to claim 1, wherein the step S1 is further:

Step S11, collecting fault information including fault time, fault type, fault position and fault reason through a distribution network automation system;

Step S12, data cleaning is carried out on fault information, repeated, wrong and invalid data are removed, and valid data are reserved;

s13, carrying out data analysis on fault information, and extracting fault characteristic variables including fault frequency, fault duration, fault influence range and fault occurrence environment factors;

And step S14, storing the fault information and the fault characteristic variable in a fault database, and providing data support for subsequent fault prediction and rush repair.

3. The method for implementing fault rush-repair based on the GIS system equipment according to claim 2, wherein the step S2 is further:

s21, selecting a preset machine learning algorithm, and constructing an equipment failure prediction model;

S22, obtaining fault data from a fault database, dividing the data into a training set and a testing set, training and verifying a device fault prediction model, optimizing model parameters, and improving model accuracy;

And S23, predicting equipment faults within a certain period of time in the future by using the trained equipment fault prediction model, and outputting fault prediction information comprising the fault position and occurrence probability of each future period of time.

4. The method for implementing fault rush-repair based on the GIS system equipment according to claim 3, wherein the step S3 is further:

step S31, determining the position of equipment with possible faults, the probability and the influence degree of the faults according to the fault prediction information;

Step S32, geographic information of equipment positions and surrounding environments, including roads, traffic and buildings, is acquired by utilizing a GIS system;

step S33, calculating the optimal position of the first-aid repair base station by utilizing a GIS system according to the equipment position and the geographic information, so that the first-aid repair base station can cover all equipment which possibly fails and is nearest to the failed equipment;

And step S34, forming a topology network of the rush-repair base stations, namely a connection relation among the rush-repair base stations, by utilizing a GIS system according to the positions of the rush-repair base stations so as to facilitate information transmission and coordination among the rush-repair base stations.

5. The method for implementing fault rush-repair based on the GIS system equipment according to claim 4, wherein the step S4 is further as follows:

step S41, acquiring real-time fault information comprising time, position, type and reason of fault occurrence through a fault rush-repair app platform;

Step S42, calculating a nearest rush-repair base station to the fault equipment and an optimal rush-repair route from the rush-repair base station to the fault equipment by utilizing a GIS system according to the real-time fault information and the position of the rush-repair base station;

Step S43, selecting a proper unmanned aerial vehicle and an optimal flight route from the unmanned aerial vehicle to the fault equipment according to the real-time fault information and the unmanned aerial vehicle position by utilizing the unmanned aerial vehicle module,

Step S44, a GIS system and an unmanned aerial vehicle module are utilized to generate an emergency repair scheduling data packet according to an emergency repair base station topological network, wherein the emergency repair scheduling data packet comprises fault information, emergency repair base station information, emergency repair route information, emergency repair task allocation information, unmanned aerial vehicle information, flight route information and unmanned aerial vehicle task allocation information;

and step S45, issuing a rush-repair scheduling data packet to the rush-repair base station and the unmanned aerial vehicle through the fault rush-repair app platform to guide a rush-repair person and the unmanned aerial vehicle to carry out rush-repair work.

6. The method for implementing fault rush-repair based on GIS system equipment according to claim 5, wherein the fault prediction model is a GAT-LSTM-attention model, comprising:

The GAT unit is used for distributing different weights for each node in the graph so as to extract the spatial characteristics and the topological structure of the node, the GAT unit firstly carries out linear transformation on the characteristics of the node, then calculates the attention coefficient among the nodes, and finally carries out weighted summation on the characteristics of the node according to the attention coefficient to obtain new node characteristics;

the LSTM unit is used for distributing different states for each node in the graph so as to extract time sequence characteristics and history information of the node, firstly, calculating forgetting gates, input gates, output gates and memory cells of the node according to the characteristics of the node and the state of the last moment, and then updating the states and outputs of the node according to the values of the gates to obtain new node states and outputs;

And the ATTENTION unit is used for distributing different importance to each node in the graph so as to extract global characteristics and context information of the nodes, and firstly calculating the correlation between the nodes according to the output and the state of the nodes, and then carrying out weighted summation on the output of the nodes according to the correlation to obtain new node output.

7. The method for implementing fault rush-repair based on the GIS system equipment according to claim 5, wherein the step S34 is further:

Step S341, acquiring equipment position and geographic information by using a GIS system, and calling fault prediction information;

Step S342, constructing at least two first-aid repair base station calculation models, forming a first-aid repair base station integrated calculation model, and training, wherein the first-aid repair base station calculation models comprise a GA-FL model, a PSO-GRA model, a MOO-FS model and an ANN-FS model;

And step S343, after training is completed, the position of the optimal repair base station is calculated by taking the equipment position, the geographic information and the fault prediction information as the input of the calculation model of each repair base station, weighting is carried out, and a weighting result is output.

8. The method for implementing the emergency repair based on the equipment failure of the GIS system according to claim 5, wherein the process of generating the emergency repair scheduling data packet in step S44 is further as follows:

Step S441, defining the targets and constraints of the rush-repair schedule, such as minimizing the rush-repair time, maximizing the rush-repair coverage rate and optimizing the rush-repair resource allocation;

s442, selecting DQN-LSTM to construct a rush repair scheduling model; historical fault data are obtained from a fault database, the data are divided into a training set and a testing set, the emergency repair scheduling model is trained and verified, model parameters are optimized, and model accuracy is improved;

Step S443, acquiring real-time fault information and a repair base station topology network by using a GIS system, taking the real-time fault information and the repair base station topology network as input of a repair scheduling model, and outputting a repair scheduling data packet comprising fault information, repair base station information, repair route information and repair task allocation information by using the trained repair scheduling model.

9. The method for implementing fault rush-repair based on GIS system equipment according to claim 5, wherein said step S12 further comprises constructing a fault knowledge graph,

After the fault information is subjected to data mining, namely, after feature variables of the fault are extracted by utilizing clustering, classification and association rules, the fault information is subjected to knowledge graph construction, and semantic representation and knowledge reasoning of the fault data are constructed by utilizing entity extraction, relation extraction and entity linking methods.