AI intelligent operation and maintenance system and method based on digital twinning

Abstract

The invention discloses an AI intelligent operation and maintenance system and method based on digital twinning, and belongs to the technical field of intelligent operation and maintenance. The method comprises the following steps: constructing a three-dimensional map, and collecting energy data and equipment data; constructing an energy digital twinborn model and an equipment digital twinborn model, and fusing the models into the three-dimensional map; judging whether the current energy data and the current equipment data are abnormal or not; marking the position of an abnormal device or an abnormal energy node on a three-dimensional map, and performing association analysis on the digital twin model to obtain an abnormal energy node source; when the energy output of the abnormal energy node source reaches a bottleneck, constructing an energy and equipment relation graph; analyzing an association path and a closeness degree between the abnormal energy node source and the abnormal equipment through a shortest path algorithm to obtain a graph structure analysis result; the priority of the abnormal equipment is determined, adjustable parameters of the abnormal equipment with different priorities are analyzed respectively, and parameter adjustment is preferentially carried out on the abnormal equipment with the low priority.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance, and specifically to an AI intelligent operation and maintenance system and method based on digital twin. Background Technique

[0002] As a core force of modern transportation, the station square of high-speed rail bears a huge flow of people and complex functional operations. Inside the high-speed rail station square, energy systems such as power supply, heating, and ventilation are intertwined, and the number of various devices such as lighting equipment, air-conditioning units, and escalators is huge and continuously operating. The stable and efficient operation of these energy systems and devices is directly related to whether the high-speed rail station can operate normally, and plays a decisive role in creating a comfortable waiting environment and ensuring the safe and convenient travel of passengers.

[0003] Under the existing technical system, energy data and equipment data are independently collected and managed by different professional systems, which results in a lack of real-time interaction and deep integration between the two types of data, and it is impossible to insight into the internal relationship between energy and equipment from a macroscopic level. When constructing energy and equipment models by traditional methods, general and simple mathematical models are mostly used, and the complex actual working conditions and equipment characteristics of the high-speed rail station square are not fully considered. When dealing with abnormal equipment, the existing technology lacks comprehensive consideration of the importance of the equipment and the impact of the failure on the production process, which may lead to over-maintenance of low-priority equipment and waste of resources. Summary of the Invention

[0004] The purpose of the present invention is to provide an AI intelligent operation and maintenance system and method based on digital twin to solve the problems raised in the existing technology.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: In the first aspect, the present invention provides an AI intelligent operation and maintenance method based on digital twin, including the following steps: Determine the three-dimensional map range of the high-speed rail station square, construct a three-dimensional map, and collect energy data and equipment data; based on the collected data, use the multi-layer perceptron algorithm to construct an energy digital twin model and an equipment digital twin model, and integrate them into the three-dimensional map; Based on the digital twin models in the three-dimensional map, determine whether there are abnormalities in the current energy data and current equipment data; when there are abnormalities, mark the positions of abnormal equipment or abnormal energy nodes in the three-dimensional map, and perform correlation analysis on the digital twin models to obtain the source of abnormal energy nodes; When it is confirmed that the energy output of the abnormal energy node source reaches a bottleneck, based on the position of the abnormal energy node source in the three-dimensional map, use a graph database to construct an energy and equipment relationship graph; analyze the correlation path and tightness between the abnormal energy node source and the abnormal equipment through the shortest path algorithm to obtain the graph structure analysis result; Based on the graph structure analysis results, combined with the importance of the equipment and the impact degree of equipment failures on the production process, determine the priority of abnormal equipment; for abnormal equipment with different priorities, analyze their adjustable parameters respectively, generate an adjustment table for abnormal equipment parameters, and give priority to adjusting the parameters of abnormal equipment with lower equipment priority.

[0006] Combined with the first aspect, in the first implementation manner of the first aspect of the present application, the determination of the three-dimensional map range of the high-speed railway station square, the construction of the three-dimensional map, and the collection of energy data and equipment data include: Determine the three-dimensional map range of the high-speed railway station square, including the main area and the surrounding area of the square; among them, the main area includes the passenger waiting area, the ticket hall, the access and exit channels, the underground parking lot, the bus transfer area, and the taxi transfer area, and the surrounding area is specifically the area related to energy supply and equipment; mark the outlines and relative position relationships of each area; Use a laser scanner, according to the layout of the high-speed railway station square, plan scanning stations, obtain three-dimensional point cloud data at each scanning station, and use point cloud processing software to splice and fuse multi-station point cloud data to generate a three-dimensional point cloud model; use a drone, plan a flight route according to the three-dimensional map range, and set camera parameters; during the flight of the drone, monitor the flight state and image acquisition situation in real time; import the collected image data into three-dimensional reconstruction software, and generate a three-dimensional map through image stitching and three-dimensional reconstruction algorithms, combined with the three-dimensional point cloud model; use SketchUp software to model the equipment in the high-speed railway station square, and the equipment includes power supply equipment and power consumption equipment; after the modeling is completed, place it in the three-dimensional map according to the actual position and direction of the equipment in the square; The energy data includes energy nodes and energy consumption, and the equipment data includes equipment operation data, equipment control instructions, and equipment basic information.

[0007] Combined with the first aspect, in the second implementation manner of the first aspect of the present application, the construction of an energy digital twin model and an equipment digital twin model using a multi-layer perceptron algorithm based on the collected data and integrating them into the three-dimensional map includes: When constructing the energy digital twin model and the equipment digital twin model, determine the number of input layer neurons of each model respectively according to the number of features after feature engineering of the energy data and the equipment data; perform debugging among different combinations of neuron numbers through grid search and cross-validation; Use the Xavier initialization method to initialize the weights of the digital twin model, input the training set data in batches, and determine the batch size through experiments; calculate the predicted output of each batch of data through forward propagation, and calculate the loss value with the true label; use the backpropagation algorithm to calculate the gradient of the loss value with respect to the parameters of the digital twin model, update the parameters through the optimizer, and use the validation set to evaluate the performance of the digital twin model; By creating a custom data source, encapsulate the digital twin model as an entity and add it to the data source, and integrate the data source into the 3D map.

[0008] Combined with the first aspect, in the third implementation manner of the first aspect of this application, judging whether the current energy data and the current device data are abnormal based on the digital twin model in the 3D map includes: For energy data, collect historical data and combine with the energy system design standard to calculate statistics, including mean and standard deviation, and based on the statistics, set the normal range of various energy parameters in the energy data; for device data, for different types of devices, determine the allowable deviation range of device parameters according to the manufacturer's settings; Input the current energy data and the current device data into the digital twin model in the 3D map, and output the current various energy parameters and the current device parameters; compare the current various energy parameters with the normal range of various energy parameters, and compare the current device parameters with the allowable deviation range of device parameters to judge whether the current energy data and the current device data are abnormal.

[0009] Combined with the first aspect, in the fourth implementation manner of the first aspect of this application, when an abnormality occurs, mark the location of the abnormal device or abnormal energy node in the 3D map, and perform correlation analysis on the digital twin model to obtain the abnormal energy node source, including: When it is determined that the energy data or the device data is abnormal, extract the coordinate information of the device or energy node involved in the abnormal data record in the 3D map and mark it in the 3D map; Extract all the energy data and device data within the time period related to the current abnormality from the digital twin model; construct a data association relationship based on the existing logical relationship description of the energy nodes and devices in the digital twin model; starting from the abnormal device or energy node, adopt a method combining forward reasoning and backward tracing according to the established data association relationship to analyze the possible propagation path of the abnormality; for each energy node that may be related to the abnormality, calculate the abnormal source probability according to its association tightness with the abnormal device and the severity of the data abnormality; verify the possible abnormal energy node sources, and when it is found during on-site inspection that a certain energy node has a power supply equipment failure and the failure matches the abnormal situation analyzed in the digital twin model, determine that the energy node is the abnormal energy node source; mark the abnormal energy node source in the 3D map.

[0010] Combined with the first aspect, in the fifth implementation manner of the first aspect of this application, when it is confirmed that the energy output of the abnormal energy node source reaches the bottleneck, based on the location of the abnormal energy node source in the 3D map, use a graph database to construct an energy and device relationship graph, including: When it is confirmed that the energy output of the abnormal energy node source reaches a bottleneck, centered on the location of the abnormal energy node source marked in the three-dimensional map, collect the energy data of the abnormal energy node source, and collect the energy data and equipment data directly or indirectly related to the abnormal energy node source; The graph database selects Neo4j, and defines the abnormal energy node source and other related energy nodes and equipment as graph database nodes; each energy node and equipment is assigned a unique identifier, and the attributes of the graph database nodes are determined; Define the energy transmission edge to describe the transmission relationship of energy between different energy nodes; define the equipment connection edge to describe the physical connection relationship between equipment; define the control relationship edge, and when there is a control relationship between equipment, define it as an edge; perform layout calculation on the graph database nodes by calling the ForceAtlas2 algorithm, and store the calculated position coordinates of the graph database nodes into the attributes of the nodes to construct the energy and equipment relationship graph.

[0011] Combined with the first aspect, in the sixth implementation manner of the first aspect of this application, the association path and tightness between the abnormal energy node source and the abnormal equipment are analyzed through the shortest path algorithm, and the graph structure analysis result is obtained, including: Extract the subgraph related to the abnormal energy node source and the abnormal equipment from the energy and equipment relationship graph; through the query statement, find the nodes corresponding to the abnormal energy node source and the abnormal equipment, and obtain all the nodes and edges with connection relationships in the corresponding nodes; set weights for the edges in the subgraph to reflect the tightness between the abnormal energy node source and the abnormal equipment; The shortest path algorithm uses the Dijkstra algorithm to calculate the shortest path from the abnormal energy node source to the abnormal equipment on the constructed subgraph; according to the calculated shortest path, analyze the association path between the abnormal energy node source and the abnormal equipment, and through the nodes and edges in the association path, obtain how the energy is transmitted from the abnormal energy node source to the abnormal equipment, and the intermediate connected energy nodes and equipment; evaluate the tightness between the abnormal energy node source and the abnormal equipment based on the length of the shortest path; obtain the graph structure analysis result, including the node sequence, path length and tightness of the shortest path.

[0012] Combined with the first aspect, in the seventh implementation manner of the first aspect of this application, based on the graph structure analysis result, combined with the importance of the equipment and the impact degree of the equipment failure on the production process, determine the priority of the abnormal equipment, including: Set evaluation indicators, including the functional criticality, usage frequency, and maintenance difficulty of the equipment, to obtain the importance of the equipment; in the functional criticality, distinguish core equipment, important equipment, and auxiliary equipment; in the usage frequency and maintenance difficulty, divide them into high, medium, and low levels; set a quantitative score range for each evaluation indicator; determine the weight for each evaluation indicator, and perform weighted calculation on the quantitative scores of the evaluation indicators to obtain the equipment importance score; Define the specific roles and locations of each equipment in the operation process of the high-speed rail station square, and evaluate the fault impact range and fault impact duration; set weights for the impact range and impact duration, and perform weighted calculation on the impact range and impact duration to obtain the equipment fault impact degree score; Comprehensively perform weighted calculation on the equipment importance score, the fault impact degree score, and the graph structure analysis result to determine the priority of the abnormal equipment.

[0013] Combined with the first aspect, in the eighth implementation manner of the first aspect of the present application, for abnormal equipment with different priorities, analyze their adjustable parameters respectively to generate an abnormal equipment parameter adjustment table, and preferentially adjust the parameters of the abnormal equipment with a lower equipment priority, including: According to the determined abnormal equipment priority score and sorting result, set the priority grouping rule; traverse all abnormal equipment, and according to the priority grouping rule, classify each equipment into the corresponding priority group to obtain a priority equipment list; Analyze the adjustable parameters of the abnormal equipment with each priority, design the framework of the abnormal equipment parameter adjustment table, sequentially extract equipment data from the priority equipment list, and fill in the adjustable parameters and equipment data of the equipment into the abnormal equipment parameter adjustment table; preferentially adjust the parameters of the abnormal equipment with a lower equipment priority.

[0014] In the second aspect, the present invention provides a digital twin-based AI intelligent operation and maintenance system, including: Three-dimensional map generation module: including: a three-dimensional map construction unit and a digital twin model construction unit; wherein, the three-dimensional map construction unit determines the three-dimensional map range of the high-speed rail station square, constructs a three-dimensional map, and collects energy data and equipment data; the digital twin model construction unit constructs an energy digital twin model and an equipment digital twin model based on the collected data using a multi-layer perceptron algorithm and integrates them into the three-dimensional map; Abnormal energy node source location module: including: an equipment data anomaly judgment unit and an abnormal energy node source location unit; wherein, the equipment data anomaly judgment unit judges whether there is an anomaly in the current energy data and the current equipment data based on the digital twin model in the three-dimensional map; when there is an anomaly, the abnormal energy node source location unit marks the location of the abnormal equipment or abnormal energy node in the three-dimensional map, and performs correlation analysis on the digital twin model to obtain the abnormal energy node source; Graph Structure Analysis Module: It includes: a relationship graph construction unit and a graph structure analysis unit; among them, when the relationship graph construction unit confirms that the energy output of the abnormal energy node source reaches a bottleneck, based on the location of the abnormal energy node source in the 3D map, it constructs an energy and equipment relationship graph using a graph database; the graph structure analysis unit analyzes the association path and tightness between the abnormal energy node source and the abnormal equipment through the shortest path algorithm to obtain the graph structure analysis result; Abnormal Equipment Parameter Adjustment Module: It includes: an abnormal equipment priority determination unit, a parameter adjustment table generation unit, and an abnormal equipment parameter adjustment unit; among them, the abnormal equipment priority determination unit determines the priority of the abnormal equipment based on the graph structure analysis result, combined with the importance of the equipment and the impact of the equipment failure on the production process; the parameter adjustment table generation unit analyzes the adjustable parameters for different priorities of abnormal equipment respectively to generate an abnormal equipment parameter adjustment table, and the abnormal equipment parameter adjustment unit preferentially adjusts the parameters of the abnormal equipment with a lower equipment priority.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. The present invention constructs a 3D map to achieve precise correlation annotation of energy data and equipment data with the map. Through data fusion technology, it breaks data silos and organically integrates energy and equipment data, providing a rich and accurate data basis for subsequent analysis and modeling.

[0016] 2. Based on the digital twin model, the present invention can judge in real time and accurately whether the current energy data and equipment data are abnormal. Through in-depth analysis of a large amount of historical data and real-time data, it makes abnormal judgments, discovers various abnormalities, and warns of early potential faults.

[0017] 3. The present invention comprehensively considers the importance of the equipment, the impact of the failure on the production process, and the graph structure analysis result to determine the priority of the abnormal equipment, and formulates personalized processing strategies respectively; it preferentially adjusts the parameters of the low-priority equipment to optimize the equipment performance and energy utilization efficiency without affecting the production process. Brief Description of the Drawings

[0018] Figure 1 It is a step schematic diagram of an AI intelligent operation and maintenance method based on digital twin of the present invention; Figure 2 It is a system structure diagram of an AI intelligent operation and maintenance system based on digital twin of the present invention. Detailed Embodiments

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution, such as Figure 1 shown in the step schematic diagram of an AI intelligent operation and maintenance method based on digital twin, the present invention provides an AI intelligent operation and maintenance method based on digital twin, including the following steps: Step S100: Determine the three-dimensional map range of the high-speed railway station square, construct a three-dimensional map, and collect energy data and equipment data; based on the collected data, use the multi-layer perceptron algorithm to construct an energy digital twin model and an equipment digital twin model, and integrate them into the three-dimensional map; Specifically, determining the three-dimensional map range of the high-speed railway station square includes the main area and the surrounding area of the square; among them, the main area includes the passenger waiting area, the ticket hall, the access and exit channels, the underground parking lot, the bus transfer area, and the taxi transfer area, and the surrounding area is specifically the area related to energy supply and equipment; mark the outlines and relative position relationships of each area; Use a laser scanner, plan scanning stations according to the layout of the high-speed railway station square, obtain three-dimensional point cloud data at each scanning station, and use point cloud processing software to splice and fuse the multi-station point cloud data to generate a three-dimensional point cloud model; use a drone to plan a flight route according to the three-dimensional map range and set the camera parameters; during the flight of the drone, monitor the flight status and image acquisition situation in real time; import the collected image data into three-dimensional reconstruction software, and generate a three-dimensional map through image stitching and three-dimensional reconstruction algorithms in combination with the three-dimensional point cloud model; select SketchUp software to model the equipment in the high-speed railway station square, and the equipment includes power supply equipment and power consumption equipment; after the modeling is completed, place it in the three-dimensional map according to the actual position and direction of the equipment in the square; The energy data includes energy nodes and energy consumption, and the equipment data includes equipment operation data, equipment control instructions, and equipment basic information.

[0021] Furthermore, when constructing the energy digital twin model and the equipment digital twin model, determine the number of input layer neurons of each according to the number of features after feature engineering of the energy data and the equipment data respectively; through grid search and cross-validation, debug in different combinations of the number of neurons; Initialize the weights of the digital twin model using the Xavier initialization method, input the training set data in batches, and determine the batch size through experiments; calculate the predicted output of each batch of data through forward propagation, and calculate the loss value with the true label; use the backpropagation algorithm to calculate the gradient of the loss value with respect to the parameters of the digital twin model, update the parameters through the optimizer, and evaluate the performance of the digital twin model using the validation set; By creating a custom data source, encapsulate the digital twin model as an entity and add it to the data source, and integrate the data source into the 3D map.

[0022] In a specific embodiment, take a high-speed railway station square as the object. The main area of the square reaches 150,000 square meters, including typical passenger waiting areas, ticket halls, access channels, underground parking lots, bus transfer areas, and taxi transfer areas. There are 3 substations, 10 distribution boxes, and various energy transmission lines distributed in the surrounding areas related to energy supply and equipment.

[0023] According to the complex layout of the square, a total of 60 scanning stations are planned to ensure full coverage of all areas of the square. 20 stations are set in the passenger waiting area, 12 in the ticket hall, 10 in the access channels, 15 in the underground parking lot, 5 in each of the bus and taxi transfer areas, and 8 in the surrounding energy areas. The laser scanner used is the RIEGL VZ - 400i, and the scanning accuracy can reach ±0.03 meters. Three-dimensional point cloud data is collected at each station, and the average amount of data collected at each station is about 600,000 points. Use PolyWorks point cloud processing software for stitching and fusion, and control the stitching error within ±0.08 meters, and finally generate a three-dimensional point cloud model with a resolution of 0.08 m × 0.08 m × 0.08 m.

[0024] Adopt the DJI Phantom 4 RTK drone, plan 12 flight routes according to the scope of the 3D map to ensure coverage of the entire square. Set the camera parameters as: resolution 5472×3648 pixels, focal length 12 mm, and shooting interval 4 seconds. During the flight of the drone, a total of 1000 images are collected. Import these images into the Agisoft Metashape 3D reconstruction software, through image stitching and 3D reconstruction algorithms, combined with the 3D point cloud model, the generated 3D map has a reduction degree of more than 92% for the surface texture of buildings in terms of detail presentation.

[0025] For the power supply equipment (such as 3 substations, 10 distribution boxes) and power-consuming equipment (2000 lighting fixtures, 50 air-conditioning units, 30 escalators, etc.) in the square, use SketchUp software for modeling. After the modeling is completed, through on-site measurement, according to the actual position and direction of the equipment in the square, accurately place it in the 3D map, and the position error is less than 0.15 meters.

[0026] In the surrounding energy area, a total of 3 main energy nodes were monitored, including 2 substations and 1 large distribution box. Energy consumption data was collected through devices such as smart meters. During a continuous 7-day monitoring period, the total daily energy consumption fluctuated between 12,000 - 14,000 kWh, and the energy consumption during peak hours (9 - 11 am, 6 - 8 pm) accounted for 42%.

[0027] Operation data of 120 typical devices (such as lighting fixtures, air conditioning units, escalators) was collected. The average daily on-time of lighting fixtures was 13 hours, the average operating power of air conditioning units was between 6 - 9 kW, and the daily operating frequency of escalators was approximately 550 times. 1200 device control instruction data were collected, covering instructions such as device start / stop and mode switching. Regarding the basic information of the devices, the model, manufacturer, rated power, etc. of the devices were recorded, involving a total of 25 different models of devices.

[0028] After feature engineering of the energy data, 18 key features were extracted, and the number of neurons in the input layer was determined to be 18. Through grid search and cross-validation, debugging was carried out among the combinations of the number of neurons in the hidden layer [30, 60, 90], and finally the hidden layer structure was determined to be [60, 90]. After feature engineering of the device data, 20 key features were obtained, and the number of neurons in the input layer was set to 20. Similarly, through debugging, the hidden layer structure was determined to be [70, 100].

[0029] The Xavier initialization method was used to initialize the weights of the digital twin model. The training set data was input in batches, and after experiments, the batch size was determined to be 64. During the training process, the Adam optimizer was used, and the initial learning rate was set to 0.001. The predicted output was calculated for each batch of data through forward propagation, the loss value was calculated with the true labels, and the gradient was calculated using the backpropagation algorithm to update the parameters.

[0030] On the validation set, the root mean square error (RMSE) of the energy digital twin model stabilized at around 80 kWh after 60 rounds of training, and the accuracy rate of the device digital twin model in judging the operating state of the device (normal, warning, failure) reached 93%.

[0031] By creating a custom data source, the digital twin model was encapsulated as an entity and added to the data source.

[0032] Step S200: Based on the digital twin model in the 3D map, determine whether there are abnormalities in the current energy data and current device data; when there are abnormalities, mark the locations of abnormal devices or abnormal energy nodes on the 3D map, and perform correlation analysis on the digital twin model to obtain the source of abnormal energy nodes; Specifically, for energy data, historical data is collected and combined with the energy system design standards to calculate statistics, including the mean and standard deviation. Based on the statistics, the normal ranges of various energy parameters in the energy data are set; for equipment data, for different types of equipment, the allowable deviation ranges of equipment parameters are determined according to the manufacturer's settings. The current energy data and current equipment data are input into the digital twin model in the 3D map, and the current various energy parameters and current equipment parameters are output; the current various energy parameters are compared with the normal ranges of the various energy parameters, and the current equipment parameters are compared with the allowable deviation ranges of the equipment parameters to determine whether there are abnormalities in the current energy data and current equipment data.

[0033] Further, when it is determined that there are abnormalities in the energy data or equipment data, the coordinate information of the equipment or energy node involved is extracted from the abnormal data record and marked in the 3D map. From the digital twin model, all the energy data and equipment data within the time period related to the current abnormality are extracted; based on the logical relationship descriptions of the existing energy nodes and equipment in the digital twin model, data association relationships are constructed; starting from the abnormal equipment or energy node, according to the established data association relationships, a method combining forward reasoning and backward tracing is adopted to analyze the possible propagation paths of the abnormality; for each energy node that may be related to the abnormality, according to its degree of association with the abnormal equipment and the severity of the data abnormality, the abnormality source probability is calculated; the possible abnormal energy node sources are verified. When it is found during on-site inspection that there is a power supply equipment failure at a certain energy node and the failure matches the abnormal situation analyzed in the digital twin model, then it is determined that the energy node is the abnormal energy node source; the abnormal energy node source is marked in the 3D map.

[0034] In a specific embodiment, the energy data of the high-speed railway station square in the past year is collected. For the voltage data, the calculated mean is 382V and the standard deviation is 5V, and the normal range is set to 372V - 392V (mean ± 2 times the standard deviation). For the current data, the statistical mean is 200A and the standard deviation is 15A, and the normal range is set to 170A - 230A. In terms of energy consumption, the daily energy consumption data is statistically analyzed, the mean is 12,500 kWh, and the standard deviation is 1,000 kWh, and the normal range is set to 10,500 - 14,500 kWh. At a certain monitoring moment, the currently collected voltage is 365V, which is lower than the lower limit of the normal range, and it is determined that the voltage data is abnormal. The current is 240A, which exceeds the upper limit of the normal range, and the current data is abnormal. The energy consumption at this moment is 15,000 kWh, which exceeds the upper limit of the normal range, and the energy consumption data is abnormal.

[0035] Taking an escalator as an example, the manufacturer sets its normal operating speed at 0.5 m / s, with an allowable deviation range of ±0.05 m / s. For an air-conditioning unit, the manufacturer stipulates that the allowable deviation of its set refrigeration temperature is ±2°C. The normal operating current of a lighting fixture is 0.5 A, with an allowable deviation range of ±0.1 A. The current operating speed of an escalator is monitored to be 0.6 m / s, exceeding the allowable deviation range, and it is determined that the speed parameter of the escalator is abnormal. The current refrigeration temperature of an air-conditioning unit is 20°C, while the set value should be 24°C, exceeding the allowable deviation range, and the temperature parameter of the air-conditioning unit is abnormal. The current operating current of a lighting fixture is 0.7 A, exceeding the allowable deviation range, and the current parameter of the lighting fixture is abnormal.

[0036] After determining the abnormalities of the above energy data and equipment data, extract the coordinate information from the abnormal data record. For example, the coordinates of the energy node with abnormal voltage in the three-dimensional map are (100, 200, 5), and it is marked with a red flashing icon in the three-dimensional map. The coordinates of the escalator with abnormal operating speed in the three-dimensional map are (300, 400, 0), and it is also marked with a red flashing icon.

[0037] Extract all energy data and equipment data within 2 hours before the occurrence of the abnormality from the digital twin model. After constructing the data association relationship, analyze the abnormal propagation path through the combination of forward reasoning and backward tracing. For example, for abnormal voltage, it is found through analysis that it is caused by a fault in a transformer of a certain substation, resulting in a decrease in the output voltage. When calculating the probability of the abnormal source, a higher abnormal source probability is assigned to the energy node that is closely associated with the abnormal equipment and has severe data abnormalities. For the above-mentioned substation node, since it is associated with multiple abnormal equipment and the degree of voltage abnormality is large, the abnormal source probability is calculated to be 80%. During on-site inspection, it is found that the transformer of the substation indeed has a winding short-circuit fault, which matches the abnormal situation analyzed by the digital twin model. Finally, it is determined that the substation is the abnormal energy node source and is highlighted with a larger red flashing icon in the three-dimensional map.

[0038] Step S300: When it is confirmed that the energy output of the abnormal energy node source reaches a bottleneck, based on the location of the abnormal energy node source in the three-dimensional map, construct an energy and equipment relationship graph using a graph database; analyze the association path and tightness between the abnormal energy node source and the abnormal equipment through the shortest path algorithm to obtain the graph structure analysis result; Specifically, when it is confirmed that the energy output of the abnormal energy node source reaches a bottleneck, taking the location of the abnormal energy node source marked in the three-dimensional map as the center, collect the energy data of the abnormal energy node source, and collect the energy data and equipment data directly or indirectly related to the abnormal energy node source; The graph database Neo4j is selected, and the abnormal energy node sources and other related energy nodes and devices are defined as graph database nodes; each energy node and device is assigned a unique identifier to determine the attributes of the graph database nodes. The energy transmission edge is defined to describe the transmission relationship of energy between different energy nodes; the device connection edge is defined to describe the physical connection relationship between devices; the control relationship edge is defined as an edge when there is a control relationship between devices; the layout calculation of the graph database nodes is performed by calling the ForceAtlas2 algorithm, and the calculated position coordinates of the graph database nodes are stored in the attributes of the nodes to construct the energy and device relationship graph.

[0039] Furthermore, a subgraph related to the abnormal energy node source and abnormal device is extracted from the energy and device relationship graph; through query statements, the nodes corresponding to the abnormal energy node source and abnormal device are found, and all the nodes and edges with connection relationships in the corresponding nodes are obtained; weights are set for the edges in the subgraph to reflect the closeness between the abnormal energy node source and the abnormal device. The shortest path algorithm uses the Dijkstra algorithm to calculate the shortest path from the abnormal energy node source to the abnormal device on the constructed subgraph; according to the calculated shortest path, the associated path between the abnormal energy node source and the abnormal device is analyzed, and through the nodes and edges in the associated path, how the energy is transmitted from the abnormal energy node source to the abnormal device and the intermediate connected energy nodes and devices are obtained; the closeness between the abnormal energy node source and the abnormal device is evaluated based on the length of the shortest path; the graph structure analysis results are obtained, including the node sequence, path length, and closeness of the shortest path.

[0040] In a specific embodiment, an abnormal energy node source (substation) is defined as a node with a unique identifier of "substation_01". Its attributes include type (substation), location coordinates (coordinates in the 3D map are [150, 250, 10]), rated output power (10000 kW), current output power (9500 kW), etc. For the 3 transmission lines, they are respectively defined as nodes with unique identifiers such as "transmission_line_01", "transmission_line_02", "transmission_line_03". The attributes include transmission capacity, current transmission power, transmission loss, etc. For example, the transmission capacity of "transmission_line_01" is 3000 kW, the current transmission power is 2800 kW, and the transmission loss is 5%. The 5 distribution boxes are also defined as nodes with unique identifiers such as "distribution_box_01", etc. The attributes are rated capacity, current load power, etc. For example, the rated capacity of "distribution_box_01" is 500 kW, and the current load power is 450 kW. The 50 electrical equipment are also defined as nodes with unique identifiers such as "lighting_01", "air_conditioner_01", etc. The attributes include equipment type, rated power, current operating status, etc. For example, the equipment type of "lighting_01" is lighting fixture, the rated power is 100 W, and the current operating status is on.

[0041] Define the energy transmission edge, such as the edge from the substation to "transmission_line_01". The attributes include transmission energy type (electricity), transmission capacity (3000 kW), current transmission power (2800 kW), transmission loss (5%). The equipment connection edge, for example, the edge from "distribution_box_01" to "lighting_01". The attributes include connection line specification (such as RVV2.5), length (20 meters). If there is a control relationship, such as the control edge of a certain control system over "air_conditioner_01", the attributes are control signal type (switch signal), control priority (high).

[0042] Call the ForceAtlas2 algorithm in Neo4j to perform layout calculation on the nodes of the graph database. After 1000 iterations of calculation, the location coordinates of each node are obtained. For example, the location coordinates of "substation_01" are updated to [155, 255, 10], and the location coordinates of "distribution_box_01" are [180, 270, 0]. These coordinates are stored in the attributes of the nodes to complete the construction of the energy and equipment relationship graph.

[0043] From the constructed energy and equipment relationship diagram, extract the sub-diagram related to the abnormal energy node source (substation) and an abnormal escalator (with the unique identifier "escalator_01"). Through query statements, find the corresponding nodes and the connected nodes and edges. Set weights for the edges in the sub-diagram. For example, the weight of the energy transmission edge is calculated comprehensively based on the transmission loss and the proportion of the current transmission power. The higher the transmission loss and the larger the proportion of the current transmission power to the transmission capacity, the higher the weight. For instance, for an energy transmission edge with a transmission loss of 8% and the current transmission power accounting for 90% of the transmission capacity, the calculated weight is 0.8. The weight of the equipment connection edge is set according to the connection line length and the importance of the equipment. The longer the length and the less important the equipment, the higher the weight. For example, for a connection line with a length of 30 meters and the connected equipment being an ordinary lighting fixture, the weight is set to 0.6.

[0044] Use the Dijkstra algorithm to calculate the shortest path from the abnormal energy node source (substation) to the abnormal escalator on the sub-diagram. The calculated node sequence of the shortest path is: "substation_01" -> "transmission_line_02" -> "distribution_box_03" -> "escalator_01". The path length is the sum of the weights of each edge, which is calculated to be 2.2. By analyzing this path, it can be seen that the energy is transmitted from the substation through "transmission_line_02" to "distribution_box_03", and then "distribution_box_03" powers "escalator_01". Evaluate the closeness based on the length of the shortest path. Since the path length is short, it indicates a high degree of closeness between the abnormal energy node source and the abnormal equipment, and the energy transmission path is relatively direct with fewer intermediate links. The final graph structure analysis result is: the node sequence of the shortest path is ["substation_01", "transmission_line_02", "distribution_box_03", "escalator_01"], the path length is 2.2, and the degree of closeness is high.

[0045] Step S400: Based on the graph structure analysis result, combined with the importance of the equipment and the impact degree of the equipment failure on the production process, determine the priority of the abnormal equipment; for abnormal equipment with different priorities, analyze their adjustable parameters respectively, generate an abnormal equipment parameter adjustment table, and preferentially adjust the parameters of the abnormal equipment with a lower equipment priority.

[0046] Specifically, set evaluation indicators, including the functional criticality, usage frequency, and maintenance difficulty of the equipment, to obtain the importance of the equipment; in the functional criticality, distinguish core equipment, important equipment, and auxiliary equipment; in the usage frequency and maintenance difficulty, divide them into high, medium, and low levels; set a quantitative score range for each evaluation indicator; determine the weight for each evaluation indicator, and perform weighted calculation on the quantitative scores of the evaluation indicators to obtain the equipment importance score. Define the specific role and location of each equipment in the operation process of the high-speed railway station square, and evaluate the fault impact range and fault impact duration; set weights for the impact range and impact duration, and perform weighted calculation on the impact range and impact duration to obtain the equipment fault impact degree score. Comprehensively perform weighted calculation on the equipment importance score, the fault impact degree score, and the graph structure analysis result to determine the priority of the abnormal equipment.

[0047] Furthermore, based on the determined priority score and sorting result of the abnormal equipment, set the priority grouping rules; traverse all abnormal equipment, and according to the priority grouping rules, classify each equipment into the corresponding priority group to obtain the priority equipment list. Analyze the adjustable parameters of each priority abnormal equipment, design the framework of the abnormal equipment parameter adjustment table, sequentially extract the equipment data from the priority equipment list, and fill in the adjustable parameters and equipment data of the equipment into the abnormal equipment parameter adjustment table; preferentially adjust the parameters of the abnormal equipment with a lower equipment priority.

[0048] In a specific embodiment, the escalator belongs to important equipment, and the quantitative score is set to 7 points; the lighting fixture is auxiliary equipment, with a score of 4 points; the air-conditioning unit is core equipment, with a score of 10 points. The escalator has a high usage frequency and gets 8 points; the lighting fixture has a medium usage frequency and gets 5 points; the air-conditioning unit has a high usage frequency in summer and gets 8 points. The escalator has a high maintenance difficulty and gets 9 points; the lighting fixture has a low maintenance difficulty and gets 3 points; the air-conditioning unit has a medium maintenance difficulty and gets 6 points.

[0049] Set the weights for functional criticality, usage frequency, and maintenance difficulty to 0.5, 0.3, and 0.2 respectively. The importance score of the escalator = 7×0.5 + 8×0.3 + 9×0.2 = 7.7 points. The importance score of the lighting fixture = 4×0.5 + 5×0.3 + 3×0.2 = 4.1 points. The importance score of the air-conditioning unit = 10×0.5 + 8×0.3 + 6×0.2 = 8.6 points.

[0050] The scope of the fault impact is regional (affecting the passage of some passengers in the waiting area), with a quantified score of 7 points; the estimated fault repair duration is 4 hours, and the impact duration score is 6 points. The scope of the fault impact is local (only affecting the lighting in individual areas), with a score of 4 points; the repair duration is estimated to be 1 hour, and the score is 3 points. The scope of the fault impact is system-level (affecting the comfort of the entire waiting area), with a score of 10 points; the repair duration is estimated to be 8 hours, and the score is 9 points.

[0051] Set the weights for the impact scope and impact duration to 0.6 and 0.4 respectively. The impact degree score of the escalator fault = 7×0.6 + 6×0.4 = 6.6 points. The impact degree score of the lighting fixture fault = 4×0.6 + 3×0.4 = 3.6 points. The impact degree score of the air-conditioning unit fault = 10×0.6 + 9×0.4 = 9.6 points.

[0052] Combined with the results of the graph structure analysis (the score conversion of the tightness between the escalator and the abnormal energy node source is 0.8, the lighting fixture is 0.4, and the air-conditioning unit is 0.7), set the weight of the equipment importance score to 0.4, the weight of the fault impact degree score to 0.4, and the weight of the graph structure analysis result score to 0.2.

[0053] The comprehensive priority score of the escalator = 7.7×0.4 + 6.6×0.4 + 0.8×0.2 = 7.04 points.

[0054] The comprehensive priority score of the lighting fixture = 4.1×0.4 + 3.6×0.4 + 0.4×0.2 = 3.8 points.

[0055] The comprehensive priority score of the air-conditioning unit = 8.6×0.4 + 9.6×0.4 + 0.7×0.2 = 9.02 points.

[0056] Set the priority grouping rules: the top 30% of the scores are high-priority, 30% - 70% are medium-priority, and the bottom 30% are low-priority. After sorting, the air-conditioning unit is high-priority, the escalator is medium-priority, and the lighting fixture is low-priority. Classify these devices into the corresponding priority groups to form a priority device list.

[0057] Design the framework of the abnormal device parameter adjustment table, including columns such as device name, device identifier, priority, adjustable parameter 1, current value of parameter 1, adjustment range of parameter 1, impact of parameter 1 adjustment, adjustable parameter 2... etc.

[0058] Extract the data of the lighting fixture from the priority device list, such as "lighting_05". Adjustable parameter: brightness level, the current value is 8 levels, the adjustment range is 1 - 10 levels, and the impact of adjustment is that increasing the brightness will increase energy consumption, and reducing the brightness can save energy but may affect the lighting effect. Fill in these data into the abnormal device parameter adjustment table.

[0059] First, adjust the parameters of lighting fixtures with low priority. For example, adjust the brightness level of "lighting_05" from level 8 to level 6. After the adjustment, it is monitored that the energy consumption in this area has decreased by 10%, and the lighting effect can still meet the basic requirements.

[0060] As Figure 2 shown, the present invention provides an AI intelligent operation and maintenance system based on digital twin, including: 3D map generation module: including: 3D map construction unit and digital twin model construction unit; wherein, the 3D map construction unit determines the 3D map range of the high-speed railway station square, constructs the 3D map, and collects energy data and equipment data; the digital twin model construction unit constructs an energy digital twin model and an equipment digital twin model based on the collected data using a multi-layer perceptron algorithm and integrates them into the 3D map; Abnormal energy node source location module: including: equipment data abnormality judgment unit and abnormal energy node source location unit; wherein, the equipment data abnormality judgment unit judges whether there is an abnormality in the current energy data and the current equipment data based on the digital twin model in the 3D map; the abnormal energy node source location unit marks the location of the abnormal equipment or abnormal energy node in the 3D map when there is an abnormality, and performs correlation analysis on the digital twin model to obtain the abnormal energy node source; Graph structure analysis module: including: relationship graph construction unit and graph structure analysis unit; wherein, the relationship graph construction unit constructs an energy and equipment relationship graph using a graph database based on the location of the abnormal energy node source in the 3D map when it is confirmed that the energy output of the abnormal energy node source reaches a bottleneck; the graph structure analysis unit analyzes the association path and tightness between the abnormal energy node source and the abnormal equipment through the shortest path algorithm to obtain the graph structure analysis result; Abnormal equipment parameter adjustment module: including: abnormal equipment priority determination unit, parameter adjustment table generation unit and abnormal equipment parameter adjustment unit; wherein, the abnormal equipment priority determination unit determines the priority of the abnormal equipment based on the graph structure analysis result, combined with the importance of the equipment and the impact of the equipment failure on the production process; the parameter adjustment table generation unit analyzes the adjustable parameters of different-priority abnormal equipment respectively to generate an abnormal equipment parameter adjustment table, and the abnormal equipment parameter adjustment unit preferentially adjusts the parameters of the abnormal equipment with low equipment priority.

[0061] It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in all respects, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes that fall within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. An AI intelligent operation and maintenance method based on digital twins, characterized in that: The following steps are involved: Determine the 3D map range of the high-speed railway station square, build a 3D map, and collect energy data and equipment data; based on the collected data, use the multi-layer perceptron algorithm to build energy digital twin models and equipment digital twin models and integrate them into the 3D map; Based on the digital twin model in the three-dimensional map, determine whether there are abnormalities in the current energy data and current equipment data; When an abnormality occurs, the location of the abnormal equipment or abnormal energy node is marked on the three-dimensional map, and the digital twin model is correlated and analyzed to obtain the abnormal energy node source; When it is confirmed that the energy output of the abnormal energy node source has reached a bottleneck, the energy and equipment relationship diagram is constructed using the graph database based on the location of the abnormal energy node source in the three-dimensional map; The shortest path algorithm is used to analyze the association path and closeness between abnormal energy node sources and abnormal equipment, and the graph structure analysis results are obtained; Based on the results of graph structure analysis, combined with the importance of the equipment and the impact of equipment failure on the production process, the priority of abnormal equipment is determined; for abnormal equipment of different priorities, their adjustable parameters are analyzed separately, and an abnormal equipment parameter adjustment table is generated, and parameter adjustment is performed first for abnormal equipment with low equipment priority.

2. According to the AI ​​intelligent operation and maintenance method based on digital twins according to claim 1, it is characterized in that: The determining of the three-dimensional map range of the high-speed railway station square, constructing the three-dimensional map, and collecting energy data and equipment data include: Determine the three-dimensional map range of the high-speed railway station square, including the main area and surrounding areas of the square; wherein the main area includes the passenger waiting area, ticket hall, entrance and exit passages, underground parking lot, bus transfer area and taxi transfer area, and the surrounding area is specifically the area related to energy supply and equipment; mark the outline and relative position relationship of each area; Using a laser scanner, according to the layout of the high-speed railway station square, planning scanning sites, obtaining three-dimensional point cloud data at each scanning site, using point cloud processing software, stitching and fusion of multi-site point cloud data, and generating a three-dimensional point cloud model; using a drone, planning a flight route according to the three-dimensional map range, setting camera parameters; during the flight of the drone, real-time monitoring of the flight status and image acquisition; importing the collected image data into the three-dimensional reconstruction software, and generating a three-dimensional map by combining the three-dimensional point cloud model with the image stitching and three-dimensional reconstruction algorithm; using SketchUp software to model the equipment in the high-speed railway station square, the equipment includes power supply equipment and power-consuming equipment; after the modeling is completed, the equipment is placed in the three-dimensional map according to its actual position and direction in the square; The energy data includes energy nodes and energy consumption, and the equipment data includes equipment operation data, equipment control instructions and equipment basic information.

3. According to the AI ​​intelligent operation and maintenance method based on digital twins according to claim 1, it is characterized in that: Based on the collected data, the energy digital twin model and the equipment digital twin model are constructed using a multi-layer perceptron algorithm and integrated into the three-dimensional map, including: When building energy digital twin models and equipment digital twin models, the number of neurons in the input layer of each model is determined based on the number of features of energy data and equipment data after feature engineering. Through grid search and cross-validation, debugging is performed in different combinations of the number of neurons. The weights of the digital twin model are initialized using the Xavier initialization method, and the training set data is input in batches. The batch size is determined through experiments. Each batch of data is predicted and output through forward propagation, and the loss value is calculated with the true label. The gradient of the loss value to the parameters of the digital twin model is calculated using the backpropagation algorithm, and the parameters are updated through the optimizer. The performance of the digital twin model is evaluated using the validation set. By creating a custom data source, the digital twin model is encapsulated as an entity and added to the data source, and the data source is integrated into the three-dimensional map.

4. According to the AI ​​intelligent operation and maintenance method based on digital twins according to claim 1, it is characterized in that: The digital twin model in the three-dimensional map is used to determine whether the current energy data and the current equipment data are abnormal, including: For energy data, historical data is collected and combined with energy system design standards to calculate statistics, including mean and standard deviation. Based on the statistics, the normal range of various energy parameters in the energy data is set; for equipment data, the allowable deviation range of equipment parameters is determined according to the manufacturer's settings for different types of equipment; Input the current energy data and current equipment data into the digital twin model in the three-dimensional map, and output the current energy parameters and current equipment parameters. Compare the current energy parameters with the normal range of the energy parameters, and compare the current equipment parameters with the allowable deviation range of the equipment parameters to determine whether there are any abnormalities in the current energy data and the current equipment data.

5. The AI ​​intelligent operation and maintenance method based on digital twin according to claim 1 is characterized in that: When an abnormality occurs, the location of the abnormal equipment or abnormal energy node is marked on the 3D map, and the digital twin model is correlated and analyzed to obtain the abnormal energy node source, including: When it is determined that energy data or equipment data is abnormal, the coordinate information of the equipment or energy node involved in the three-dimensional map is extracted from the abnormal data record and marked in the three-dimensional map; From the digital twin model, extract all energy data and equipment data in the time period related to the current anomaly; construct data association relationships based on the logical relationship descriptions of existing energy nodes and equipment in the digital twin model; starting from the abnormal equipment or energy node, according to the established data association relationship, use a combination of forward reasoning and reverse tracing to analyze the possible propagation path of the anomaly; for each energy node that may be related to the anomaly, calculate the probability of the anomaly source based on the closeness of its association with the abnormal equipment and the severity of the data anomaly; verify the possible abnormal energy node sources, and when a power supply equipment failure is found in an energy node during the field inspection, and the failure matches the abnormal situation analyzed in the digital twin model, the energy node is determined to be an abnormal energy node source; mark the abnormal energy node source in the three-dimensional map.

6. The AI ​​intelligent operation and maintenance method based on digital twin according to claim 1 is characterized in that: When it is confirmed that the energy output of the abnormal energy node source reaches a bottleneck, based on the location of the abnormal energy node source in the three-dimensional map, a graph database is used to construct an energy and equipment relationship graph, including: When it is confirmed that the energy output of the abnormal energy node source has reached a bottleneck, the energy data of the abnormal energy node source is collected with the abnormal energy node source position marked in the three-dimensional map as the center, and the energy data and equipment data directly or indirectly related to the abnormal energy node source are collected; The graph database selects Neo4j, defines abnormal energy node sources and other related energy nodes and devices as graph database nodes; each energy node and device is given a unique identifier to determine the properties of the graph database node; Define energy transmission edges to describe the transmission relationship between different energy nodes; define device connection edges to describe the physical connection relationship between devices; define control relationship edges, and define them as edges when there is a control relationship between devices; call the ForceAtlas2 algorithm to calculate the layout of the graph database nodes, store the calculated graph database node position coordinates in the node attributes, and build an energy and device relationship graph.

7. The AI ​​intelligent operation and maintenance method based on digital twin according to claim 1 is characterized in that: The shortest path algorithm is used to analyze the association path and closeness between the abnormal energy node source and the abnormal device to obtain the graph structure analysis results, including: Extract the subgraph related to the abnormal energy node source and abnormal equipment from the energy and equipment relationship graph; find the nodes corresponding to the abnormal energy node source and abnormal equipment through the query statement, and obtain all the nodes and edges with connection relationships in the corresponding nodes; set weights for the edges in the subgraph to reflect the closeness between the abnormal energy node source and the abnormal equipment; The shortest path algorithm adopts the Dijkstra algorithm to calculate the shortest path from the abnormal energy node source to the abnormal device on the constructed subgraph; based on the calculated shortest path, the associated path between the abnormal energy node source and the abnormal device is analyzed, and through the nodes and edges in the associated path, it is obtained how the energy is transmitted from the abnormal energy node source to the abnormal device, as well as the energy nodes and devices connected in the middle; the closeness between the abnormal energy node source and the abnormal device is evaluated based on the length of the shortest path; and the graph structure analysis results are obtained, including the node sequence, path length and closeness of the shortest path.

8. The AI ​​intelligent operation and maintenance method based on digital twin according to claim 1 is characterized in that: The above-mentioned priority of abnormal equipment is determined based on the results of the graph structure analysis, combined with the importance of the equipment and the impact of the equipment failure on the production process, including: Set evaluation indicators, including the functional criticality, frequency of use, and difficulty of maintenance of the equipment, to obtain the importance of the equipment; in terms of functional criticality, distinguish between core equipment, important equipment, and auxiliary equipment; in terms of frequency of use and difficulty of maintenance, divide them into high, medium, and low levels; set a quantitative score range for each evaluation indicator; determine the weight for each evaluation indicator, perform weighted calculation on the quantitative scores of the evaluation indicators, and obtain the equipment importance score; Define the specific role and location of each device in the operation process of the high-speed railway station square, and evaluate the scope and duration of the fault impact; set weights for the impact scope and impact duration, and perform weighted calculations on the impact scope and impact duration to obtain the equipment fault impact score; The equipment importance score, fault impact score, and graph structure analysis results are comprehensively considered to perform weighted calculation and determine the priority of abnormal equipment.

9. The AI ​​intelligent operation and maintenance method based on digital twin according to claim 1 is characterized in that: The method of analyzing the adjustable parameters of abnormal devices of different priorities respectively, generating an abnormal device parameter adjustment table, and preferentially adjusting the parameters of abnormal devices with low device priorities includes: According to the determined abnormal device priority scores and sorting results, priority grouping rules are set; all abnormal devices are traversed, and each device is classified into a corresponding priority group according to the priority grouping rules to obtain a priority device list; Analyze the adjustable parameters of abnormal devices of each priority level, design the abnormal device parameter adjustment table framework, extract device data from the priority device list in sequence, fill the device's adjustable parameters and device data into the abnormal device parameter adjustment table; give priority to adjusting parameters of abnormal devices with low device priority.

10. An AI intelligent operation and maintenance system based on digital twins, using an AI intelligent operation and maintenance method based on digital twins according to any one of claims 1 to 9, characterized in that: include: 3D map generation module: including: 3D map construction unit and digital twin model construction unit; wherein, the 3D map construction unit determines the 3D map range of the high-speed railway station square, constructs the 3D map, and collects energy data and equipment data; the digital twin model construction unit uses the multi-layer perceptron algorithm to construct the energy digital twin model and the equipment digital twin model based on the collected data, and integrates them into the 3D map; Abnormal energy node source positioning module: including: equipment data abnormality judgment unit and abnormal energy node source positioning unit; wherein, the equipment data abnormality judgment unit judges whether there are abnormalities in the current energy data and the current equipment data based on the digital twin model in the three-dimensional map; when there is an abnormality, the abnormal energy node source positioning unit marks the location of the abnormal equipment or abnormal energy node on the three-dimensional map, and performs correlation analysis on the digital twin model to obtain the abnormal energy node source; Graph structure analysis module: including: a relationship graph construction unit and a graph structure analysis unit; wherein, when the relationship graph construction unit confirms that the energy output of the abnormal energy node source has reached a bottleneck, based on the location of the abnormal energy node source in the three-dimensional map, the graph database is used to construct an energy and equipment relationship graph; the graph structure analysis unit uses the shortest path algorithm to analyze the association path and closeness between the abnormal energy node source and the abnormal equipment to obtain the graph structure analysis result; Abnormal equipment parameter adjustment module: including: abnormal equipment priority determination unit, parameter adjustment table generation unit and abnormal equipment parameter adjustment unit; wherein, the abnormal equipment priority determination unit determines the priority of abnormal equipment based on the graph structure analysis result, combined with the importance of the equipment and the impact of equipment failure on the production process; the parameter adjustment table generation unit analyzes the adjustable parameters of abnormal equipment of different priorities respectively, generates an abnormal equipment parameter adjustment table, and the abnormal equipment parameter adjustment unit preferentially adjusts the parameters of abnormal equipment with low equipment priority.

Images