Railway hump air compressor station predictive operation and maintenance management system and method

Abstract

The application discloses a railway hump air compressor station predictive operation and maintenance management system and method, and belongs to the technical field of intelligent operation and maintenance of pneumatic equipment in railway marshalling stations. The system comprises a hardware sensing and collecting layer, a private network transmission layer, a platform intelligent layer and an edge execution layer. The hardware layer collects unit working conditions, energy efficiency and energy consumption and environmental security data; the platform layer is provided with a multi-mode fusion research and judgment module, which performs threshold hard alarm, trend deterioration detection and machine learning feature recognition in parallel, and accurately determines four types of faults, i.e. fire, pipeline air leakage, motor hidden abnormality and oil and gas overtemperature; a predictive maintenance closed loop module automatically matches a differentiated maintenance strategy and links a spare parts inventory; a dynamic load joint control module adaptively adjusts unit operation according to load prediction; and a digital twin module realizes 3D visualization. The application realizes the transformation from passive repair to active prediction, greatly reduces operation and maintenance cost and energy consumption, and improves air supply reliability.

Description

Technical Field

[0001] This invention relates to the cross-technical fields of intelligent operation and maintenance of pneumatic equipment in railway marshalling yards and hump yards, industrial Internet of Things, and artificial intelligence fault diagnosis and predictive maintenance, specifically to a predictive operation and maintenance management system and method for railway hump air compressor stations. Background Technology

[0002] The hump air compressor station is the core power source for automated hump operations in railway marshalling yards, continuously providing clean, dry compressed air at a constant pressure (0.75 MPa) to equipment such as electro-pneumatic switch machines and vehicle reducers. Its air supply reliability and equipment operational stability directly determine the efficiency and safety baseline of train marshalling operations.

[0003] In the process of realizing this invention, the inventors found that there are at least the following problems in the prior art: At present, most of the air compressor stations in China still adopt the three-shift or four-shift manual station duty mode, relying on regular inspections and paper records, which has the following rigid pain points: (1) High labor costs, high pressure of shift rotation in remote stations, and easy to overlook night inspections; (2) Hidden faults (such as small air leakage in the pipeline, early wear of bearings, and deterioration of lubricating oil) cannot be detected in advance, and can only be repaired passively afterward, resulting in interruption of air supply and affecting the grouping operation; (3) The energy consumption management is rough, the fixed power frequency operation of the unit cannot adapt to the tidal gas load of the hump, and the idling and leakage cause a lot of energy waste; (4) The operation and maintenance ledger is fragmented, lacks digital management of the entire life cycle of equipment, and the spare parts are blindly stored.

[0004] Existing general-purpose air compressor station monitoring systems on the market (such as Chinese patent CN202310453857.8) have basic data acquisition and alarm functions, but they are designed for civilian use and are not adapted to the special operating conditions of railway hump yards (severe load fluctuations, strong electromagnetic interference, and strict safety standards). They also cannot achieve predictive fault assessment and closed-loop maintenance based on multi-mode fusion. Therefore, there is an urgent need for a fully closed-loop intelligent predictive operation and maintenance management system and method specifically designed for railway hump yard air compressor stations. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, this invention provides a predictive operation and maintenance management system and method for railway hump air compressor stations, aiming to achieve a transformation from manual to unmanned operation and from reactive emergency repairs to proactive prediction, thereby improving the reliability and energy efficiency of air supply.

[0006] The technical solution adopted by this invention to solve its technical problem is as follows: Firstly, an embodiment of this invention provides a predictive operation and maintenance management system for railway hump yard air compressor stations, comprising a hardware sensing and acquisition layer, a dedicated network transmission layer, a platform intelligence layer, and an edge execution layer. The hardware sensing and acquisition layer includes a unit operating condition acquisition module, an energy efficiency metering acquisition module, and an environmental security sensing module, which respectively acquires the operating parameters (pressure, temperature, current, power, frequency, etc.), power consumption, air production, and station temperature and humidity, smoke detection, and access control status of the air compressor host, motor, dryer, and air tank. The dedicated network transmission layer adopts a dedicated railway network, has a built-in encryption protocol, and possesses three levels of storage: edge caching and cloud backup. The platform intelligence layer integrates a multi-mode fusion analysis module, a predictive maintenance closed-loop module, a dynamic load control module, a digital twin visualization module, and a full lifecycle asset ledger module. The multi-mode fusion analysis module operates in parallel, performing threshold over-limit hard alarms, time-series trend degradation detection, and machine learning feature recognition to categorize four types of faults: fire, pipeline leakage, hidden motor anomalies, and main unit oil and gas overheating. The predictive maintenance closed-loop module automatically matches emergency maintenance, planned maintenance, or targeted maintenance strategies based on the analysis results, generating work orders and linking spare parts inventory. The dynamic load control module adaptively adjusts unit switching and inverter speed based on load forecasts. The digital twin visualization module constructs a 1:1 3D model and maps data in real time. The edge execution layer receives platform commands to drive field equipment actions.

[0007] Secondly, the present invention provides a predictive operation and maintenance management method for a hump-type air compressor station, comprising: S1 seamless on-site deployment; S2 synchronous acquisition and three-level storage of multi-source data across the entire domain; S3 multi-mode parallel intelligent fusion analysis; S4 predictive maintenance closed-loop management; S5 dynamic load control and leakage management; and S6 remote visualization and asset ledger management. A data closed loop is formed between these steps, collaboratively achieving predictive operation and maintenance.

[0008] One of the above technical solutions has the following advantages or beneficial effects: (1) Through multi-mode fusion analysis (threshold + trend + machine learning), four types of high-risk fault hazards can be identified in advance. The fault discovery time is several hours to several days earlier than manual inspection, which greatly reduces the risk of unplanned shutdown and increases the air supply qualification rate to over 99.99%.

[0009] (2) Predictive maintenance closed-loop linkage with work orders and spare parts inventory avoids over-maintenance and maintenance gaps, extends equipment life and reduces operation and maintenance costs by more than 30%.

[0010] (3) Dynamic load control combined with air leakage control reduces the overall energy consumption of air compressor stations by 15% to 22%, resulting in significant economic benefits.

[0011] (4) Digital twins and visual management reduce the technical threshold for personnel and enable unmanned operation and remote centralized management.

[0012] (5) This invention organically combines the above-mentioned multiple technical means: the alarm level output by multi-mode analysis directly drives the maintenance strategy and load control; the pressure changes monitored during the load control process are fed back to the air leakage detection, forming a closed loop; digital twin integration of all data provides global visualization support. This synergistic effect makes the overall operation and maintenance efficiency far exceed the sum of the functions of each module. Attached Figure Description

[0013] Figure 1 This is an overall architecture diagram of a predictive operation and maintenance management system for a railway hump air compressor station, according to an exemplary embodiment. Figure 2 This is a flowchart illustrating a predictive operation and maintenance management method for a hump air compressor station according to an exemplary embodiment; Figure 3 This is a schematic diagram of the overall hierarchical hardware and software global topology architecture of a predictive operation and maintenance management system for a railway hump air compressor station, according to an exemplary embodiment. Figure 4 This is a flowchart illustrating the algorithm logic of a multi-mode fusion intelligent fault early warning model according to an exemplary embodiment; Figure 5 This is a schematic diagram illustrating the timing control of a dynamic energy efficiency optimization and control strategy for air compressor load, according to an exemplary embodiment. Figure 6 This is a flowchart illustrating a predictive maintenance closed-loop work order linkage process for high-risk faults such as fire early warning, according to an exemplary embodiment. Figure 7 This is a flowchart illustrating a predictive maintenance closed-loop work order linkage process for high-risk faults such as pipeline leakage early warning, according to an exemplary embodiment. Figure 8 This is a flowchart illustrating a predictive maintenance closed-loop work order linkage process for high-risk faults such as motor anomaly early warning, according to an exemplary embodiment. Figure 9 This is a flowchart illustrating a predictive maintenance closed-loop work order linkage process for high-risk faults such as main engine oil and gas temperature early warning, according to an exemplary embodiment. Detailed Implementation

[0014] To more clearly illustrate the technical features of the present invention, the present invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings.

[0015] Example 1 like Figure 1As shown in the figure, the predictive operation and maintenance management system for railway hump air compressor stations provided by this invention includes a hardware sensing and acquisition layer, a private network transmission layer, a platform intelligence layer, and an edge execution layer. The hardware sensing and acquisition layer includes a unit operating condition acquisition module, an energy efficiency metering acquisition module, and an environmental security sensing module, which are used to collect the operating condition parameters, power consumption and gas production of the air compressor host, motor, dryer, and air tank, as well as the temperature and humidity of the station, smoke detector and access control status, and upload the collected data to the platform intelligent layer through the dedicated network transmission layer. The dedicated network transmission layer adopts a railway independent dedicated network, with built-in encrypted transmission protocol and a three-level storage mechanism of edge caching and cloud backup; The platform's intelligent layer includes a multi-mode fusion analysis module, a predictive maintenance closed-loop module, a dynamic load control module, a digital twin visualization module, and a full lifecycle asset ledger module. The multi-mode fusion analysis module executes three parallel logics: threshold exceeding hard alarm, time-series trend degradation detection, and machine learning feature recognition. Based on the fusion results, it identifies four types of potential faults: fire, pipeline leakage, hidden motor anomalies, and main unit oil / gas overheating, generating analysis results including alarm levels and fault locations. The predictive maintenance closed-loop module automatically matches differentiated maintenance strategies based on the analysis results, generates electronic work orders, and links spare parts inventory. The dynamic load control module adaptively adjusts the number of air compressors in operation and the inverter speed based on real-time gas load forecasts. The digital twin visualization module constructs a three-dimensional virtual model of the air compressor station and maps it to real-time operating data. The full lifecycle asset ledger module records equipment maintenance history and wear status. The edge execution layer receives control commands from the platform's intelligent layer, driving the air compressor frequency converter cabinet, start / stop switch, dryer adjustment module, and audible and visual alarm terminal to perform corresponding actions.

[0016] Preferably, the threshold over-limit hard alarm in the multi-mode fusion judgment module includes: triggering an oil and gas over-temperature emergency alarm when the host exhaust temperature is ≥96℃; and triggering a fire emergency alarm when the station temperature and humidity and smoke concentration simultaneously exceed a preset threshold.

[0017] Preferably, the time-series trend deterioration detection includes: under the constraint of a unit full power ≥ 74kW, monitoring the rate at which the gas supply pressure decreases from 0.75MPa; if the rate of decrease exceeds 0.02MPa / min and the duration is greater than a set value, it is determined to be a potential gas leakage problem in the pipeline network.

[0018] Preferably, the machine learning feature recognition adopts a decision tree and neural network fusion model. The input features include three-phase current imbalance, active power deviation, and body temperature rise rate. The output is the probability value of motor bearing wear, rotor imbalance or stator winding damage, and positioning components.

[0019] Preferably, the differentiated maintenance strategy includes: matching emergency alarms with immediate shutdown and emergency maintenance; matching trend deterioration alarms with off-peak window planned maintenance; and matching machine learning-based component-level faults with targeted and precise maintenance.

[0020] Preferably, the dynamic load control module uses a time series prediction algorithm based on historical gas consumption data to divide peak, off-peak, and low-peak periods. During low-peak periods, it automatically shuts down redundant power frequency units and keeps variable frequency units running at low load. During peak periods, it increases the number of units in operation and increases the variable frequency speed to maintain a stable gas supply pressure of 0.75 MPa.

[0021] Preferably, the digital twin visualization module acquires device data in real time via OPC UA or MQTT protocol, drives the synchronous update of the 3D model, and supports automatic highlighting of fault locations via pop-up windows.

[0022] Example 2 like Figure 2 As shown in the figure, the predictive operation and maintenance management method for a hump air compressor station provided by this invention includes the following steps: Step S1: Install a hardware sensing and acquisition terminal on the existing hump air compressor station equipment in situ, which is compatible with Modbus, PLC or OPC communication protocols, without modifying the original automatic air-operating control circuit.

[0023] The hardware sensing and acquisition terminal includes a unit operating condition acquisition module, an energy efficiency metering acquisition module, and an environmental security sensing module. The unit operating condition acquisition module connects to the air compressor host, drive motor, dryer, air tank, and pipeline valves to collect data on supply pressure, exhaust temperature, frequency conversion frequency, three-phase current, active power, oil-gas separator filter element wear time, and lubricating oil aging wear time. The energy efficiency metering acquisition module integrates a smart energy meter and a pipeline flow meter to collect data on unit power consumption, instantaneous air production, and cumulative air consumption. The environmental security sensing module integrates a temperature and humidity sensor, a smoke detector, an audible and visual access control sensor, and a noise, dust, and oil mist sensor.

[0024] The in-situ installation adopts a non-stop construction method, with each data acquisition terminal installed at corresponding points on the outer wall of the unit, the power supply cabinet, the main gas supply pipeline, and the wall of the machine room. Industrial shielded cables are used for nearby wiring and grounding to resist interference, without affecting the normal programming and gas supply operation of the hump.

[0025] The adapted communication protocols include: Modbus RTU, Modbus TCP, PLC backplane bus and OPC UA multi-protocol adaptation and connection can be completed with one click through the edge acquisition gateway, without modifying the original control logic of the original automatic ventilation main control cabinet and pressure interlock protection circuit.

[0026] The requirement to not modify the original automatic ventilation control circuit specifically means that all data acquisition terminals are connected in parallel to the existing equipment signal circuit using a passive sensing method, without changing the transmission path and electrical parameters of the original control signal, thus maintaining the independent operation capability of the original system.

[0027] Step S1 also includes: deploying an edge acquisition gateway and a private network encrypted transmission switch in the back-end computer room. The edge acquisition gateway has a built-in multi-protocol parsing engine that supports automatic identification and dynamic parsing of various industrial protocols such as Modbus, Profibus, CANopen and OPC DA.

[0028] The hardware sensing and acquisition terminal adopts an industrial-grade wide-temperature design, with an operating temperature range of -40℃ to 85℃ and a protection level of no less than IP65, making it suitable for the strong electromagnetic interference environment at railway sites.

[0029] The in-situ installation process also includes: scanning the communication ports and listening to the protocols of existing equipment in the station, automatically identifying the communication protocol type, data frame format and register mapping table of each device, and generating a protocol adaptation configuration file.

[0030] The hardware sensing and acquisition terminal and the edge acquisition gateway adopt a dual-redundant communication link, including a main link RS-485 bus and a backup link wireless Mesh network. When the main link is interrupted, it automatically switches to the backup link.

[0031] Step S2: The hardware sensing and acquisition terminal collects unit operating parameters, energy efficiency metering data and environmental security data in real time and uploads them to the platform's intelligent layer through the railway's independent private network. At the same time, it performs three-level hierarchical storage in the local edge gateway, monitoring master station and cloud.

[0032] The unit operating parameters include the exhaust temperature, supply pressure, inverter output frequency, three-phase operating current, active power, reactive power, power factor, pressure difference before and after the oil-gas separator, lubricating oil temperature, and inlet-outlet temperature difference of each air compressor.

[0033] The energy efficiency metering data includes real-time electricity consumption, daily cumulative electricity consumption, monthly cumulative electricity consumption, instantaneous gas production, daily cumulative gas production, monthly cumulative gas production, and unit compressed air comprehensive energy consumption for each unit, and is calculated separately according to peak, off-peak, and low-peak periods.

[0034] The environmental security data includes temperature, humidity, smoke concentration, sound and light intrusion status, noise level, dust concentration, and oil mist concentration throughout the station area, with a collection frequency of no less than once per second.

[0035] The railway's independent private network is a physically isolated industrial Ethernet network that does not occupy the railway's shared office network resources. It uses fiber optic leased lines or 5G private network slicing for access, and the data transmission latency is no more than 100 milliseconds.

[0036] The encrypted upload uses a combination of symmetric and asymmetric encryption algorithms recognized by the State Cryptography Administration to encrypt and desensitize the original data before transmission, and the key is rotated and updated periodically.

[0037] The three-tiered storage is as follows: the edge gateway retains the most recent 7 days of high-frequency raw data (sampling interval not exceeding 1 second), the monitoring master station retains the most recent 1 year of detailed data (sampling interval not exceeding 1 minute), and the cloud archive saves all historical data with a long-term storage period of not less than 10 years. Data consistency is ensured between the storage levels through incremental synchronization and breakpoint resume mechanisms.

[0038] Step S2 also includes: the edge gateway performs real-time preprocessing on the collected raw data. The preprocessing includes data filtering and noise reduction, outlier removal, missing data interpolation and completion, and dimension normalization. The preprocessed data is packaged and uploaded according to a set period.

[0039] The real-time synchronous acquisition adopts a millisecond-level polling mechanism. The edge gateway polls the global sensing terminals at a period of no more than 500 milliseconds. When a data change is detected, it actively reports it to achieve second-level refresh of important parameters.

[0040] Step S2 also includes: the edge gateway has a built-in data verification module to cross-verify the collected parameters. When the deviation of multiple sensor readings of the same physical quantity exceeds a set threshold, a sensor abnormality alarm is triggered and the data is marked as unreliable data.

[0041] The platform's intelligent layer is deployed in a private or dedicated cloud environment within the railway industry. The platform's entry point is equipped with two-factor authentication and IP whitelist access control, allowing only authorized monitoring master stations and maintenance terminals to access the platform.

[0042] In step S3, the platform's intelligent layer performs a triple judgment process, which includes threshold over-limit hard alarm, time-series trend degradation detection, and machine learning feature recognition. The three results are then fused and voted to identify four types of faults: fire, pipeline leakage, motor hidden abnormalities, and main unit oil and gas overheating. The alarm level and location information are then output.

[0043] The threshold-exceeding hard alarms include: triggering a first-level emergency alarm for oil and gas overheating when the main unit exhaust temperature is ≥96℃; triggering a first-level emergency alarm for fire when the station temperature is ≥50℃ and the smoke concentration is ≥5%obs / m; triggering a second-level alarm for insufficient gas supply when the gas supply pressure is lower than 0.65MPa for more than 10 seconds; and triggering a second-level alarm for motor overload when the single unit operating current exceeds 110% of the rated current.

[0044] The time-series trend deterioration detection includes: under the constraint of stable operation at full power ≥74kW of the unit, monitoring the rate at which the pressure of the main gas supply pipe decreases from the reference value of 0.75MPa. If the average rate of decrease exceeds 0.02MPa / min within 5 consecutive minutes, it is determined to be a potential air leakage problem in the pipeline network; monitoring the active power deviation rate of the motor under the same operating conditions. If the deviation rate shows a monotonically increasing trend within 24 consecutive hours and the final deviation exceeds ±8%, it is determined to be a potential problem of motor efficiency deterioration.

[0045] The machine learning feature recognition adopts a fusion model of random forest and deep neural network. The input feature vector includes three-phase current imbalance, active power deviation rate under the same working condition, surface temperature rise rate of the fuselage, vibration spectrum feature value and lubricating oil metal particle content. The model output is the bearing wear probability, rotor imbalance probability, stator winding damage probability and suggested maintenance components. When any probability value is greater than 0.7, the corresponding level of alarm is triggered.

[0046] The triple result fusion voting adopts the Dempster-Shafer evidence theory and / or a weighted voting mechanism, where the threshold hard alarm weight is 0.4, the trend detection weight is 0.3, the machine learning weight is 0.3, the comprehensive confidence score is ≥0.8 to trigger a level 1 emergency alarm, 0.6≤comprehensive confidence score<0.8 to trigger a level 2 warning, 0.4≤comprehensive confidence score<0.6 to trigger a level 3 prompt, and values ​​below 0.4 only log the data without triggering an alarm.

[0047] The fire fault identification adopts a dual-sensor over-limit logic of temperature, humidity and smoke detection. When the temperature threshold and the smoke concentration threshold are exceeded at the same time, a fire alarm is triggered. If a single sensor exceeds the limit, only a sensor abnormality prompt is triggered to reduce the false alarm rate.

[0048] The identification of pipeline leakage faults also includes: under the full power operation of the unit, synchronously monitoring the rate of decrease in gas supply pressure and the rate of increase in unit output flow. When the rate of decrease in pressure exceeds the threshold and the rate of increase in flow is lower than the theoretical gas replenishment rate, it is determined that the pipeline is damaged or the joint is loose and leaking air. When the rate of decrease in pressure and the rate of increase in flow rise synchronously, it is determined that the gas consumption at the user end has increased, and no leakage alarm is triggered.

[0049] The identification of hidden abnormalities in motors also includes: comparing and analyzing the real-time three-phase current waveform with historical healthy waveforms, calculating the waveform distortion rate and harmonic content changes, and determining that when the distortion rate increases by more than 15% from the baseline value, it is a potential risk of bearing wear or rotor bar breakage.

[0050] The main unit's oil and gas over-temperature identification adopts a dual logic of fixed threshold and rate detection: when the exhaust temperature is ≥96℃, an emergency alarm is immediately triggered; when the exhaust temperature rises at a rate exceeding 2℃ / min within 30 minutes and the current temperature is ≥88℃, an early warning is triggered to prevent over-temperature seizure failure in advance.

[0051] Step S3 also includes: establishing a dedicated fault feature knowledge base for the Hump air compressor station. The knowledge base stores typical feature patterns of various faults, historical case data and corresponding handling solutions. The machine learning model regularly uses newly added fault samples for incremental training and model iteration updates to achieve self-optimization of the judgment algorithm.

[0052] Step S4: Automatically match differentiated maintenance strategies based on alarm levels, generate standardized electronic work orders, link spare parts inventory ledgers for remaining quantity verification, complete fault handling, component replacement and operating condition verification, and update the equipment lifecycle ledger.

[0053] The differentiated maintenance strategy includes: Level 1 emergency alarms are matched with immediate shutdown for emergency preventative maintenance, requiring maintenance personnel to arrive on-site within 30 minutes; Level 2 early warnings are matched with off-peak window planned maintenance, scheduled during off-peak hours or early morning of the next day; Level 3 prompts are matched with daily inspections for enhanced attention, and are included in the weekly maintenance plan for unified handling.

[0054] The standardized electronic work order includes the work order number, alarm type and level, fault location equipment and components, suggested maintenance plan, recommended replacement spare parts list and models, safety precautions, planned completion time and acceptance criteria.

[0055] The linked spare parts inventory ledger for remaining quantity verification includes: the system automatically queries the current inventory quantity based on the spare parts list recommended by the work order; when the inventory quantity is lower than the safety stock threshold, it automatically generates a spare parts purchase requisition and pushes it to the purchasing department; when the inventory quantity is sufficient, it automatically locks the corresponding spare parts and updates the reservation status.

[0056] The fault handling includes: maintenance personnel receiving work orders through a mobile terminal APP, and performing fault diagnosis, component replacement, parameter calibration and functional testing item by item according to the standardized operation instructions pushed by the system. Each step of the operation is confirmed on the APP and on-site photos or videos are uploaded as evidence.

[0057] The working condition verification includes: after maintenance is completed, the system automatically performs a load test, collects the operating parameters of the equipment under three working conditions of no load, half load and full load, compares them with the data before maintenance and the health baseline data, and the work order can be closed only after all indicators meet the standards.

[0058] The updated equipment lifecycle ledger includes: recording the date, type, replaced parts and batch number, maintenance personnel, parameter comparison before and after maintenance, next recommended maintenance date and estimated remaining lifespan for each maintenance, forming a complete electronic file from equipment commissioning to scrapping.

[0059] Step S4 also includes: the system dynamically and adaptively optimizes the filter replacement cycle, lubricating oil filling cycle, bearing maintenance cycle and cooler cleaning cycle based on the actual operating conditions of the equipment and maintenance history data, so as to avoid over-maintenance or maintenance gaps caused by fixed-cycle maintenance.

[0060] The spare parts inventory ledger includes spare parts name, model and specifications, current inventory, safety stock threshold, in-transit purchase quantity, supplier information and historical consumption rate. The system predicts future consumption trends based on historical consumption data and automatically adjusts the safety stock threshold.

[0061] Step S4 also includes: the system automatically calculates the frequency of occurrence, average repair time and spare parts consumption of each type of fault, and generates equipment reliability analysis report and maintenance quality assessment report to provide data support for equipment selection and maintenance strategy optimization.

[0062] Step S5: Based on the real-time gas load forecast, the system automatically divides the time period and dynamically adjusts the number of air compressors in operation and the variable frequency speed. At the same time, it combines pipeline pressure decay detection to locate air leakage points, thereby achieving stable gas supply and energy-saving joint control.

[0063] The gas load forecast uses a long short-term memory neural network model. The input features include historical gas consumption for the same period in the past 7 days, gas consumption in the current period, peak load programming plan, ambient temperature and humidity, and forecast gas demand curves for the next 1 to 4 hours at 15-minute intervals.

[0064] The automatic time segmentation specifically includes: 1) The cloud-based operation and maintenance platform first obtains the gas load forecast results. This gas load forecast is based on an LSTM (Long Short-Term Memory) neural network model. The input feature vector includes: historical gas consumption for the same period 7 days prior to the current time, gas consumption time-series data for the 24 hours prior to the current time, the daily peak load programming schedule, and ambient temperature and humidity. The LSTM neural network model outputs a gas demand forecast curve for the next 1 to 4 hours at 15-minute intervals, i.e., a sequence of predicted gas load values ​​for each 15-minute time window within the next 1 to 4 hours. This predicted gas load value sequence is represented as: L1, L2, ..., L n , where n is the total number of prediction time windows; 2) Based on the predicted gas load sequence and the rated gas supply capacity C of the air compressor station, the cloud-based operation and maintenance platform divides each day into peak, off-peak, and low-peak periods according to the following rules: (1) If the predicted gas load L in a certain 15-minute time window i Satisfy L i If the temperature is >0.7°C, then this time window is marked as a peak period. (2) If the predicted gas load L in a certain 15-minute time window i Satisfy L i If the temperature is less than 0.3°C, then the time window is marked as a low period. (3) If the predicted gas load L in a certain 15-minute time window i Satisfying 0.3C≤L i If the temperature is ≤0.7°C, then this time window is marked as an off-peak period; Wherein, the rated air supply capacity C is the total air supply of all operational air compressors in the air compressor station under rated operating conditions, and the unit is consistent with the predicted air load, which is cubic meters per minute; 3) The cloud-based operation and maintenance platform merges the time windows marked as peak periods and continuously distributed into peak continuous periods, merges the time windows marked as trough periods and continuously distributed into trough continuous periods, and merges the time windows marked as off-peak periods and continuously distributed into off-peak continuous periods, and records the start and end times of each continuous period.

[0065] Preferably, the division of peak hours, off-peak hours and low-peak hours also satisfies the minimum duration constraint: when the duration of any continuous period is less than 30 minutes, the continuous period is merged into the adjacent period. Specifically, if the continuous period has both preceding and following periods, it is merged into the adjacent period with the closer predicted load average.

[0066] Preferably, the manual adjustment includes: maintenance personnel adjusting the 70% threshold and the 30% threshold through the human-computer interaction interface of the cloud-based maintenance platform, with the adjustment range not exceeding ±10% of the rated gas supply capacity, that is, the adjusted peak-hour threshold range is 60% to 80% of the rated gas supply capacity, and the off-peak-hour threshold range is 20% to 40% of the rated gas supply capacity; the manual adjustment permission is only granted to the administrator role, and each adjustment operation is recorded in the traceability log.

[0067] The dynamic adjustment of the number of air compressors in operation and the variable frequency speed includes: increasing the number of units in operation during peak hours and evenly distributing the load of each unit, and increasing the variable frequency speed to maintain the pipeline pressure at 0.75MPa±0.02MPa; maintaining the optimal number of operating units during off-peak hours and using variable frequency speed control to precisely match the air demand; shutting down redundant fixed frequency units during off-peak hours and retaining only a single variable frequency unit to operate at the lowest energy consumption speed to maintain basic pressure stabilization and air supply.

[0068] The method for detecting and locating air leaks by detecting pressure decay in the pipeline network includes: monitoring the time-series decay curve of the main gas supply pressure under full-power operation of the unit, collecting leakage sound wave signals by combining the acoustic sensor array arranged in the pipeline network, and calculating the location of the air leak point through a time-difference positioning algorithm, with a positioning error of less than 1 meter.

[0069] The pressure stabilization and energy-saving control also includes: real-time monitoring of the running time, temperature rise and load rate of each air compressor, dynamically balancing the cumulative running time of each unit to avoid premature aging caused by long-term high-load operation of a single device; when the main equipment malfunctions, it automatically switches to the standby unit within seconds to ensure uninterrupted air supply.

[0070] Step S5 also includes: automatically calculating the electricity consumption, gas production, and unit compressed air energy consumption of each unit on a daily, weekly, and monthly basis, generating an energy efficiency comparison report for each unit, marking inefficient and high-consumption units, and outputting special recommendations for energy-saving renovation.

[0071] The dynamic adjustment adopts a model predictive control algorithm, which uses the gas demand forecast for the next hour as feedforward and the current pipeline pressure deviation as feedback to calculate the optimal unit combination and speed setpoint, so as to achieve advance adjustment and precise control of gas supply pressure.

[0072] The method for locating the air leak also includes: combining the step-by-step shut-off test of the pipeline network zone valves to further narrow down the air leak area based on acoustic positioning, and determining the specific pipeline section where the air leak is located by comparing the pressure attenuation rate changes before and after the shut-off of each zone.

[0073] Step S6: The digital twin 3D model displays the real-time operating status and alarm points of the entire station, and supports remote parameter adjustment, equipment start-up and shutdown, and hierarchical access control, and automatically generates operation and maintenance reports and traceability logs.

[0074] The digital twin 3D model is constructed by collecting data on the building structure, equipment layout, and pipeline routing of the air compressor station, and using 3D modeling software to build a 1:1 scale virtual model. The model includes the shape, location, connection relationship, and process parameter labels of all core equipment.

[0075] The real-time display of the entire station's operating status includes: using color changes on the 3D model to indicate the equipment's operating status—green for normal operation, yellow for warning, red for alarm, and gray for shutdown; using dynamic particle flow or arrow animation to indicate the flow direction and flow rate of compressed air and cooling water; and using numerical labels to update key parameters such as pressure, temperature, and current at each measuring point in real time.

[0076] The automatic highlighting pop-up location of alarm points includes: when the system generates an alarm, the 3D model automatically rotates and focuses on the alarm device, which is highlighted in red flashing, and at the same time, an information window pops up to display alarm details, fault type, suggested handling solution and historical records of similar faults.

[0077] The remote parameter adjustment includes: users with administrator privileges can click on the device icon in the 3D model interface to enter the parameter setting panel and remotely modify operating parameters such as pressure setpoint, temperature alarm threshold, variable frequency speed limit, and unit start-stop priority. The parameter modification records are fully archived.

[0078] The equipment supports two start / stop modes: remote manual control and automatic strategy control. In the remote manual control mode, maintenance personnel can perform single-unit start / stop operations by clicking on the equipment icon on the 3D model. In the automatic strategy control mode, the system automatically decides on start / stop based on load forecasting and energy efficiency optimization algorithms, and the operator only needs to confirm the strategy execution.

[0079] The hierarchical access control is divided into three levels: administrator, operation and maintenance operator, and read-only inspector. Administrators have full access to the system, including parameter modification, user management, and policy configuration. Operation and maintenance operators can process work orders, start and stop equipment, and fine-tune parameters. Read-only inspectors can only view real-time data, historical reports, and alarm information, and have no right to perform any control operations.

[0080] The operation and maintenance reports include: daily operation and maintenance briefing (including equipment runtime, alarm statistics, and energy consumption overview), weekly safety analysis report (including fault trends, hidden danger distribution, and maintenance completion rate), and monthly energy efficiency analysis ledger (including energy efficiency comparison of each unit, energy saving results, and carbon emission accounting). All reports can be exported to PDF or Excel format with one click according to the time range.

[0081] The traceability logs include device operation logs and user operation logs. The operation logs record all device parameter changes, status switching, and alarm events, while the operation logs record all user login, logout, parameter modification, device control, and work order processing behaviors. The log data is tamper-proof and must be stored for at least 10 years.

[0082] Step S6 also includes: the digital twin model supports historical data playback function, and users can select any historical time period to play back the dynamic process of equipment operation status, parameter change curves and alarm events in the three-dimensional model during that period, which can be used for fault tracing analysis and operation and maintenance training.

[0083] Example 3 The present invention provides a predictive operation and maintenance management system for railway hump air compressor stations, comprising four parts: a hardware sensing and acquisition unit, a dedicated network secure transmission unit, a cloud-based intelligent operation and maintenance management platform unit, and a field edge linkage execution unit. It is fully compatible with the existing automatic air supply system of railway hump stations and can be seamlessly integrated with it.

[0084] The hardware sensing and acquisition unit is fully deployed within the Camel Peak air compressor station, comprising three integrated sensing terminals: a unit operating condition acquisition module, an energy efficiency metering acquisition module, and a station building security and environmental sensing module. The unit operating condition acquisition module interfaces with all core equipment, including the air compressor main unit, drive motor, dryer, air tank, and pipeline valves, to collect real-time operating condition parameters such as supply pressure, exhaust temperature, main unit output frequency converter frequency, three-phase operating current, output active power, oil-gas separator filter wear time, and lubricating oil aging wear time. The energy efficiency metering acquisition module integrates a high-precision smart energy meter and a high-precision pipeline flow meter to accurately collect real-time power consumption, instantaneous air production, and cumulative air consumption, calculating core energy efficiency indicators for the unit's comprehensive compressed air energy consumption. The station building security and environmental sensing module integrates an integrated temperature and humidity sensor, a smoke detection sensor, an audible and visual intrusion access control sensor, and an integrated noise, dust, and oil mist sensor terminal to collect multi-dimensional data on the station's environmental security. The dedicated network security transmission unit adopts closed-loop transmission of the railway's independent dedicated industrial network, which does not occupy shared office network resources. It has a built-in data encryption and desensitization transmission protocol, and is equipped with a three-level hierarchical local data caching + cloud long-term backup storage mechanism. The core operation and maintenance data of the monitoring master station and operation and maintenance center are stored for no less than 10 years, ensuring the security and compliance of the entire link of operation and maintenance data transmission, storage and retrieval.

[0085] The cloud-based intelligent operation and maintenance management platform unit is equipped with a lightweight B / S access architecture, supports multi-terminal network login without discrimination, and has built-in seven compliance design adaptation modules, a multi-mode integrated intelligent early warning and judgment module, a predictive maintenance closed-loop management module, a dynamic load energy efficiency optimization and joint control module, a remote unmanned dispatch and management module, a full life cycle asset ledger management module, and a digital twin panoramic visualization interaction module; it also features multi-role hierarchical permission isolation management, adapting to the hierarchical permission division of administrators, front-line operation and maintenance personnel, and read-only inspectors, to prevent unauthorized access and misoperation. The multi-mode integrated intelligent early warning and judgment module incorporates a triple parallel judgment logic: threshold over-limit hard alarm, time-series trend degradation dynamic detection, and machine learning feature autonomous judgment. This logic jointly constructs a dedicated fault feature knowledge base for the Hump air compressor station. It accurately targets and identifies four types of high-risk core fault hazards: First, dual-sensor joint threshold judgment for fire hazards, triggering a level-one emergency alarm by synchronously exceeding limits in temperature, humidity, and smoke sensor data. Second, time-series pressure decay judgment for hidden air leaks in the pipeline network under full-power operating conditions, identifying all types of air leaks, including pipeline damage, loose joints, cracked switch machine hoses, and internal leakage in the air cylinder. Third, current and power time-series distortion trend judgment for hidden abnormalities such as motor bearing wear, rotor imbalance, and stator winding damage. Fourth, a fixed high-temperature threshold to identify high-risk faults such as main unit oil and gas overheating and seizure, with tiered alarm SMS messages, mobile APP, and maintenance terminal notifications. The predictive maintenance closed-loop management module is adapted to the coordinated execution of three types of differentiated and precise maintenance strategies: threshold-based emergency over-limit alarms are matched with immediate shutdown for emergency preventative maintenance; time-series slow deterioration trend alarms are matched with off-peak window planned scheduled maintenance; machine learning-based fault feature-based precise tracing is matched with component-level targeted precision maintenance; and the synchronous linkage system automatically generates standardized electronic maintenance work orders, dynamically and adaptively optimizes the full-cycle maintenance intervals for filter replacement, lubrication oil filling, and bearing maintenance, automatically calculates the remaining amount in the spare parts inventory ledger, and automatically pushes spare parts procurement and replenishment reminders at critical thresholds, thus eliminating the dual problems of excessive maintenance and redundant losses, as well as equipment failure due to insufficient maintenance. The dynamic load energy efficiency optimization and control module collects real-time data on the instantaneous gas load fluctuations of the entire station. Relying on the intelligent load prediction algorithm, it dynamically adapts and adjusts the number of air compressors in operation and the real-time speed of the unit's frequency converter to accurately match the real-time gas supply and demand balance during peak hours. Simultaneously, it uses dual technologies of acoustic leak detection and pressure attenuation reverse verification to accurately locate hidden air leaks in the pipeline network and reduce ineffective gas source losses. It automatically compiles energy consumption and gas production ledgers by time period and individual unit, automatically generates energy efficiency analysis reports, intelligently identifies inefficient operating units and energy consumption bottlenecks, and outputs customized energy-saving optimization and maintenance strategies.The digital twin panoramic visualization interaction module creates a 1:1 full-domain 3D virtual digital model of the hump air compressor station, dynamically mapping the real-time operating conditions of all core equipment, precise positioning of alarm points, real-time energy consumption curves, and full-domain environmental security status; it is equipped with functions including centralized display on the operation and maintenance data large screen, automatic generation of daily / weekly / monthly operation and maintenance account statements, and one-click fault traceback, enabling on-site operation and maintenance personnel to get started quickly without thresholds, and reducing post training and operation and maintenance costs.

[0086] The on-site edge linkage execution unit connects to the air compressor frequency conversion control cabinet, unit start-stop linkage switch, intelligent dryer regulation module, and on-site acousto-optic alarm terminal, receives intelligent regulation instructions, emergency shutdown instructions, and hierarchical alarm linkage instructions issued by the platform, and completes adaptive adjustment of equipment working conditions, emergency fault linkage disposal, and on-site security acousto-optic alarm at the second level.

[0087] The system described in this embodiment adopts a newly built layered architecture, full-domain perception, intelligent judgment, closed-loop maintenance, energy-saving joint control, and visual management integrated exclusive technical solution, which is divided into four core layers: hardware full-domain non-inductive collection, private network secure transmission, platform intelligent judgment and management, and edge on-site linkage execution, supporting an exclusive standardized intelligent operation and maintenance management method. At the hardware end, multiple types of integrated low-power sensor collection terminals are installed in situ without modifying the original core gas supply control loop, to non-inductively collect full-dimensional data of unit working conditions, energy efficiency and consumption, and environmental security; the transmission end adopts the railway independent private dedicated network for encrypted closed-loop transmission, with three-level layered long-term data storage, to meet the railway data security compliance requirements; the platform end is equipped with a multi-mode integrated judgment and early warning model to accurately lock four types of high-risk faults, and link full-chain functions of predictive maintenance work orders, dynamic management of spare parts inventory, dynamic load energy-saving joint control, hierarchical authority asset accounts, and digital twin visualization; the edge end links equipment start-stop, frequency conversion speed regulation, on-site acousto-optic alarm, and automatic switching of standby units at the second level, to ensure uninterrupted air supply throughout the process. The overall modular design and standardized interfaces enable rapid batch adaptation and deployment at various hump air compressor stations across the entire railway network.

[0088] As Figure 3 shown, the system described in this embodiment adopts a four-layer architecture as a whole: operation and maintenance decision management layer → intelligent operation and maintenance platform layer → secure transmission network layer → on-site collection and execution layer. The network adopts an independent railway private dedicated network with physical isolation and encrypted transmission to ensure industrial control security. The present invention is compatible with existing automatic air supply systems, PLC, Modbus industrial bus protocol, and OPC (OLE for Process Control) protocol, without the need to modify original equipment. It realizes a full closed-loop data flow of collection → transmission → analysis → early warning → regulation → operation and maintenance → visualization. It supports all services including unattended operation, remote monitoring, intelligent fault early warning, predictive maintenance, and energy efficiency optimization.

[0089] The system described in this embodiment employs a multi-mode fusion intelligent fault early warning model algorithm for intelligent early warning of air compressor station equipment. The algorithm flow is as follows: Figure 4 As shown, the process involves: multi-source sensor data access → data preprocessing → three-way parallel analysis (threshold alarm + trend detection + machine learning) → fault feature matching → fault location and classification → early warning push + maintenance suggestions → model self-learning iteration.

[0090] 1. Data Access: Real-time collection of multi-dimensional operational and environmental data, including air compressor pressure, exhaust temperature, motor three-phase current, power, frequency, machine room temperature and humidity, smoke detection, and pipeline pressure.

[0091] 2. Data preprocessing: The raw data is filtered and denoised, outliers are removed, missing data is filled in, and units are normalized to eliminate on-site interference and ensure the reliability of the algorithm input.

[0092] 3. Three-way parallel independent analysis: Threshold mode: Set a fixed threshold, such as main unit oil and gas temperature ≥96℃, smoke / temperature and humidity exceeding the threshold, directly trigger over-limit alarm; Trend pattern: Perform trend fitting on time series parameters to identify degradation processes such as continuous decrease in pipeline pressure and slow distortion of motor current and power under full power. Machine learning mode: Call the trained neural network and decision tree model, compare with the fault feature library, identify hidden fault modes and locate faulty components.

[0093] 4. Integrated decision-making: The results of the three-way analysis are integrated and voted on to avoid false alarms and omissions caused by a single mode, thereby improving the accuracy of early warning.

[0094] 5. Fault Classification and Location: Accurately distinguish four major types of faults: computer room fire, pipeline air leakage, motor body abnormality, and host oil and gas overheating, and locate the specific equipment / component.

[0095] 6. Risk Classification: Divide into Level 1 Emergency Alarm, Level 2 Key Warning, and Level 3 General Alert, and implement differentiated management.

[0096] 7. Multi-channel push notifications: Alarms are simultaneously pushed through the operation and maintenance platform pop-up window, mobile APP, SMS, email, and on-site sound and light alarms.

[0097] 8. Intelligent maintenance strategy matching: Threshold exceeded: Immediate shutdown and emergency repairs; Deteriorating trend: Schedule planned maintenance during designated work windows; Multi-mode fusion analysis model identification: accurately locate components and perform targeted repairs.

[0098] 9. Work order closed loop and model iteration: Automatically generate maintenance work orders and archive operation and maintenance records; store real fault samples in the feature library, continuously train and optimize machine learning models, and realize algorithm self-upgrade.

[0099] The system described in this embodiment employs a time-series control strategy for energy efficiency optimization and joint control, such as... Figure 5 As shown: Gas load fluctuation data acquisition → Load forecasting and analysis → Unit switching + variable frequency speed dynamic adjustment → Pipeline air leakage energy consumption control → Load balancing distribution + automatic standby switching → Time-segmented energy efficiency assessment + optimization strategy output. The time-series segmented control rules are as follows: 1. Peak gas load periods: With busy peak air demand and a surge in gas consumption, the system automatically increases the number of air compressors in operation, evenly distributes the load across units, and adjusts the variable frequency speed to ensure a stable 0.75MPa pressure in the pipeline network, meeting peak gas demand.

[0100] 2. During off-peak gas load periods: The workload is stable and gas consumption is steady. The system maintains the optimal number of operating units, uses variable frequency speed control to precisely match the air demand, avoids redundant start-stop cycles, maintains stable gas supply, and controls energy consumption.

[0101] 3. Low gas load periods: There are fewer programming and decoding operations at night, resulting in a significant drop in gas consumption. The system automatically shuts down redundant fixed-frequency units, retaining only a single variable-frequency unit for low-load, stable-voltage operation, eliminating idling energy consumption and ineffective losses.

[0102] 4. Real-time linkage control of pipeline leakage: The entire process adopts pressure attenuation analysis + acoustic detection to monitor the time sequence changes of pipeline pressure in real time; under the condition of full power of the unit ≥74KW, if the gas supply pressure continuously drops from 0.75MPa, it is determined to be a leak, which is included in the energy efficiency optimization factor, triggering leak detection and maintenance to reduce gas waste.

[0103] 5. Main / Standby Interlocking and Load Balancing: Real-time monitoring of the operating time, temperature rise, and load rate of each air compressor. Dynamic load balancing to prevent single units from aging under prolonged high load; in case of main equipment failure, standby is activated within seconds, ensuring uninterrupted air supply.

[0104] 6. Energy efficiency closed-loop assessment: Daily / weekly / monthly statistics on electricity consumption, gas production, and unit compressed air energy consumption are compiled, compared with historical data and industry benchmarks, inefficient units are identified, and energy-saving operation strategies and equipment modification suggestions are automatically generated.

[0105] The following section provides a complete, clear, and detailed implementation description of the technical solution of this invention, using a real-world engineering scenario of an uphill hump air compressor station in a certain location as an example.

[0106] The first step was to implement the on-site, seamless adaptation and networking deployment. At an existing power frequency + variable frequency combined air compressor station in a designated location, multiple units were selected for operation. Without dismantling or altering the original automatic air intake control cabinet and pressure interlock protection core circuit, the existing components were installed on-site, attached to the unit's outer wall, power supply cabinet, main gas supply pipeline, and machine room walls. Correspondingly, integrated unit operating condition data acquisition terminals, smart energy efficiency meters, high-precision pipeline flow meters, and composite security sensor terminals for temperature, humidity, smoke detection, and access control were installed. Industrial shielded cables were used for on-site wiring, with grounding to resist interference and adapt to the strong electromagnetic environment of the railway. An edge acquisition gateway and a dedicated encrypted transmission switch were deployed in the back-end machine room, enabling one-click Modbus and PLC dual-protocol adaptation and connection. The entire deployment process was completed without downtime and did not affect the normal gas supply operation of the on-site hump.

[0107] The second step involves encrypted, synchronous collection and archiving of multi-source time-series data across the entire region. Edge gateways poll the entire region's sensor terminals at millisecond levels, synchronously collecting comprehensive data on each air compressor's exhaust temperature, real-time air pressure, three-phase current, active power, cumulative filter runtime, lubricant aging parameters, power consumption per unit across the entire station, hourly air production, temperature and humidity throughout the computer room, smoke detection concentration, and access control intrusion status. All original maintenance data, intermediate analysis data, and alarm results are encrypted and uploaded unidirectionally to the cloud-based intelligent maintenance platform via the railway station's independent dedicated industrial network. This is accompanied by simultaneous three-level layered caching and backup at the monitoring master station, edge gateway, and cloud center. Key maintenance data is archived and stored for over 10 years, conforming to railway safety traceability and archive management standards.

[0108] The third step involves multi-mode integrated parallel intelligent analysis, tiered early warning, and response. The platform's backend simultaneously activates a triple parallel analysis algorithm: For data center fire risks, it employs a dual-limit-crossing logic combining temperature and humidity thresholds with smoke concentration thresholds. Once both parameters exceed the limits simultaneously, a Level 1 red emergency alarm is immediately triggered, with simultaneous push notifications via SMS, the maintenance app, and on-site audio-visual terminals. For hidden air leaks in the pipeline network, it locks in the pre-constraint of stable operation at full power ≥74KW for the units, and monitors in real-time the continuous temporal decay fluctuations of the main gas supply pressure from 0.75MPa, automatically identifying hidden air leaks such as pipeline damage, loose joints, and cracked switch machine hoses. For hidden damage to the motor, it continuously tracks the three-phase operating current distortion, low active power under the same frequency operating condition, and the slow rise in the temperature of the motor body, and relies on the trained decision tree + neural network fault feature library to accurately locate the potential problems of bearing wear, rotor imbalance, and minor faults in the stator winding, triggering a level two preventive alarm; for the main engine oil and gas safety risks, it fixes a 96℃ high temperature critical threshold, and immediately alarms when the limit is exceeded, and links the edge side to prepare emergency shutdown and lockout commands to prevent the main engine from seizing and being damaged.

[0109] The fourth step is to implement a closed-loop management system for tiered predictive maintenance. For level-one fires, pipeline leaks, motor malfunctions, and emergency alarms for excessive oil and gas temperatures in the main unit, the system immediately displays a highlighted pop-up notification and automatically generates an electronic work order for emergency shutdown and troubleshooting. The work order process is as follows: Figures 6 to 9 As shown, mandatory maintenance personnel must arrive on-site within a limited time to handle the situation, and the unit can only be unlocked and restarted after high-risk safety hazards have been eliminated. For alarms regarding pipeline leaks and motor deterioration trends, the system automatically matches off-peak and off-peak operation windows during nighttime peak hours, and plans to distribute maintenance work orders for fixed-point leak detection, bearing temperature measurement, and winding insulation testing. It also synchronizes with the spare parts ledger, ensuring that sealing joints, lubricating grease, and spare bearings are prepared in advance. For component-level hidden faults identified through machine learning-based precise tracing, the system directly pushes targeted expert maintenance suggestions for accurate and targeted repairs, avoiding blind disassembly and significantly shortening fault handling time. Simultaneously, the system dynamically and adaptively optimizes the maintenance and replacement cycles for all categories of filters, lubricating oil, and seals, preventing both excessive maintenance and wasted consumables, as well as equipment damage from insufficient maintenance. It also automatically updates the electronic ledger for the entire equipment lifecycle.

[0110] The fifth step is intelligent optimization and control of energy efficiency through dynamic load linkage. The platform provides real-time statistics on the peak and tidal gas load time-series curves of the entire station over the past 1 hour and 24 hours, and uses load prediction algorithms to predict the peak and valley ranges of gas consumption in the next period. During peak gas consumption periods, the platform automatically and evenly distributes the operating load of multiple units to avoid premature aging of a single unit under long-term full load. During off-peak gas consumption periods, redundant fixed-frequency units are automatically shut down, while variable-frequency units are kept in low-load and stable-pressure operation to reduce ineffective idling energy consumption. The system also integrates with the pipeline acoustic leak detection + pressure attenuation dual verification technology to accurately locate minute air leaks and promptly arrange leak sealing operations to reduce gas source loss. The system automatically generates monthly energy efficiency comparison reports for each unit, marks inefficient and high-consumption old equipment, and outputs specific recommendations for energy-saving renovations, ensuring that the entire station achieves its long-term stable energy-saving and consumption-reducing operation and maintenance goals.

[0111] Step 6: Remote unattended dispatching + digital twin visualized operation and maintenance. Operation and maintenance personnel do not need to be stationed on-site. Through any compliant networked computer on the office intranet or a dedicated operation and maintenance mobile app, they can remotely view the operating parameters, alarm logs, energy consumption curves, and environmental security status of all units in the entire station in real time. They can remotely complete unit start-up and shutdown, pressure parameter fine-tuning, and manual switching of standby units within their compliant permissions. The platform has a built-in 1:1 3D digital twin virtual model of the air compressor station, automatically highlighting fault locations with pop-up windows. Operation and maintenance data is centrally summarized and displayed on a large screen, automatically generating daily operation and maintenance reports, weekly safety reports, and monthly energy consumption analysis logs, which are archived with one click. Simultaneously, three levels of operation permissions are defined for administrators, operation and maintenance workers, and remote inspectors, recording all remote operation logs throughout the process to prevent unauthorized or erroneous operations, fully complying with the railway safety production hierarchical management and control standards.

[0112] This invention precisely addresses six core technical pain points of existing hump yard air compressor stations: high labor costs for manual operation and maintenance, delayed early warning of hidden faults, easy interruption of air supply due to sudden faults, extensive energy consumption management without optimization methods, lack of closed-loop maintenance records making traceability difficult, and inability to adapt to unmanned digital transformation. At the same time, it solves four compatibility shortcomings: poor compatibility between general industrial control maintenance equipment and existing railway automatic air supply systems, substandard data security for public network transmission, lack of railway-specific fault analysis models, and non-compliance with railway safety production authority control regulations. It comprehensively meets the stringent on-site operating conditions of railway marshalling yards.

[0113] Compared with the prior art, the present invention has the following characteristics: First, the operation and maintenance model has been upgraded and optimized to fully replace the three / four shifts of manual on-site duty, realize the normalized operation and maintenance with fewer or no staff, directly reduce the on-site duty manpower cost by more than 30%, and avoid various human operation and maintenance risks such as manual inspection errors and absence of night duty. Second, the fault prevention and control is proactive. Relying on multi-mode integrated analysis, hidden deterioration hazards are captured in advance, and pre-emptive and precise maintenance is achieved, which greatly reduces the frequency of unplanned shutdowns and increases the air pressure supply qualification rate to over 99.99%, effectively reducing the risk of train safety accidents caused by abnormal air supply. Third, precise and refined energy consumption management and control, dynamic linkage adjustment of unit load and number of operating units as needed, and simultaneous and precise treatment of hidden air leakage in the pipeline network, resulting in a 15%-22% reduction in the overall compressed air energy consumption of the entire station, with significant long-term energy-saving economic benefits; Fourth, the entire operation and maintenance process is closed-loop and traceable, with electronic work orders, full life cycle asset ledgers, and spare parts inventory linked for management and control, standardizing maintenance and digitizing assets to prevent waste of consumables and premature equipment aging; Fifth, it has extremely strong on-site adaptability, requires no modification to the existing native air-cooling system, is compatible with multiple industrial protocols, has encrypted transmission over a private network that complies with railway safety standards, and is quick to deploy, easy to maintain, and has strong resistance to on-site electromagnetic interference. Sixth, visualized management improves quality and efficiency. The 3D digital twin panoramic view intuitively displays the status of the entire site, enables one-click fault tracing, and automatically generates operation and maintenance reports, significantly reducing the technical threshold and training costs for operation and maintenance personnel. Seventh, it takes into account multiple core values ​​such as safe and stable supply, cost reduction and efficiency improvement, compliance management and digital transformation, and has broad prospects for promotion and application.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A predictive operation and maintenance management system for railway hump yard air compressor stations, characterized in that, It includes a hardware sensing and acquisition layer, a private network transmission layer, a platform intelligence layer, and an edge execution layer; The hardware sensing and acquisition layer includes a unit operating condition acquisition module, an energy efficiency metering acquisition module, and an environmental security sensing module, which are used to collect the operating condition parameters, power consumption and gas production of the air compressor host, motor, dryer, and air tank, as well as the temperature and humidity of the station, smoke detector and access control status, and upload the collected data to the platform intelligent layer through the dedicated network transmission layer. The dedicated network transmission layer adopts a railway independent dedicated network, with built-in encrypted transmission protocol and a three-level storage mechanism of edge caching and cloud backup; The platform's intelligent layer includes a multi-mode fusion analysis module, a predictive maintenance closed-loop module, a dynamic load control module, a digital twin visualization module, and a full lifecycle asset ledger module. The multi-mode fusion analysis module executes three parallel logics: threshold exceeding hard alarm, time-series trend degradation detection, and machine learning feature recognition. Based on the fusion results, it identifies four types of potential faults: fire, pipeline leakage, hidden motor anomalies, and main unit oil / gas overheating, generating analysis results including alarm levels and fault locations. The predictive maintenance closed-loop module automatically matches differentiated maintenance strategies based on the analysis results, generates electronic work orders, and links spare parts inventory. The dynamic load control module adaptively adjusts the number of air compressors in operation and the inverter speed based on real-time gas load forecasts. The digital twin visualization module constructs a three-dimensional virtual model of the air compressor station and maps it to real-time operating data. The full lifecycle asset ledger module records equipment maintenance history and wear status. The edge execution layer receives control commands from the platform's intelligent layer, driving the air compressor frequency converter cabinet, start / stop switch, dryer adjustment module, and audible and visual alarm terminal to perform corresponding actions.

2. The predictive operation and maintenance management system for railway hump yard air compressor stations according to claim 1, characterized in that, The threshold over-limit hard alarm in the multi-mode fusion analysis module includes: triggering an oil and gas over-temperature emergency alarm when the host exhaust temperature is ≥96℃; and triggering a fire emergency alarm when the station temperature and humidity and smoke concentration simultaneously exceed the preset threshold.

3. The predictive operation and maintenance management system for railway hump yard air compressor stations according to claim 1 or 2, characterized in that, The time-series trend deterioration detection includes: under the constraint of a unit full power ≥ 74kW, monitoring the rate at which the gas supply pressure drops from 0.75MPa. If the rate of drop exceeds 0.02MPa / min and the duration is greater than the set value, it is determined to be a potential air leakage problem in the pipeline network. The machine learning feature recognition adopts a decision tree and neural network fusion model. The input features include three-phase current imbalance, active power deviation, and body temperature rise rate. The output is the probability value of motor bearing wear, rotor imbalance or stator winding damage, and positioning components.

4. A predictive operation and maintenance management method for a hump air compressor station, characterized in that, Includes the following steps: Step S1: Install a hardware sensing and acquisition terminal on the existing hump air compressor station equipment in situ, which is compatible with Modbus, PLC or OPC communication protocols, without modifying the original automatic air control circuit. Step S2: Real-time synchronous collection of unit operating parameters, energy efficiency metering data and environmental security data through the hardware sensing and acquisition terminal, and encrypted upload to the platform intelligent layer through the railway independent private network, while performing three-level hierarchical storage in the local edge gateway, monitoring master station and cloud. Step S3: The platform's intelligent layer performs three-fold analysis in parallel: threshold over-limit hard alarm, time-series trend degradation detection, and machine learning feature recognition. The three results are then fused and voted to identify four types of faults: fire, pipeline leakage, motor hidden abnormality, and main unit oil and gas over-temperature. The alarm level and location information are then output. Step S4: Automatically match differentiated maintenance strategies according to alarm level, generate standardized electronic work orders, link spare parts inventory ledger for balance verification, complete fault handling, component replacement and operating condition review, and update the equipment life cycle ledger. Step S5: Based on the real-time gas load forecast results, the time period is automatically divided and the number of air compressors switched on and off and the variable frequency speed are dynamically adjusted. At the same time, the leak point is located by detecting the pressure decay of the pipeline network, so as to realize the joint control of stable gas supply and energy saving. Step S6: The digital twin 3D model displays the real-time operating status and alarm points of the entire station, and supports remote parameter adjustment, equipment start-up and shutdown, and hierarchical access control, and automatically generates operation and maintenance reports and traceability logs.

5. The predictive operation and maintenance management method for hump air compressor stations according to claim 4, characterized in that, The requirement to not modify the original automatic ventilation control circuit specifically means that all data acquisition terminals are connected in parallel to the existing equipment signal circuit using a passive sensing method, without changing the transmission path and electrical parameters of the original control signal, thus maintaining the independent operation capability of the original system.

6. The predictive operation and maintenance management method for hump air compressor stations according to claim 4, characterized in that, The three-tiered storage is as follows: the edge gateway retains the most recent 7 days of high-frequency raw data, with a sampling interval of no more than 1 second; the monitoring master station retains the most recent 1 year of detailed data, with a sampling interval of no more than 1 minute; the cloud archive saves all historical data and sets a long-term storage period of no less than 10 years; and the data consistency between the storage levels is ensured through incremental synchronization and breakpoint resume mechanisms.

7. The predictive operation and maintenance management method for hump air compressor stations according to claim 4, characterized in that, The triple result fusion voting adopts DS evidence theory and / or weighted voting mechanism, where the threshold hard alarm weight is 0.4, the trend detection weight is 0.3, the machine learning weight is 0.3, the comprehensive confidence score is ≥0.8 to trigger a first-level emergency alarm, 0.6≤comprehensive confidence score<0.8 to trigger a second-level warning, 0.4≤comprehensive confidence score<0.6 to trigger a third-level prompt, and below 0.4, only logs are recorded without triggering an alarm.

8. The predictive operation and maintenance management method for hump air compressor stations according to claim 4, characterized in that, The differentiated maintenance strategy includes: Level 1 emergency alarms are matched with immediate shutdown for emergency preventative maintenance, requiring maintenance personnel to arrive on-site within 30 minutes; Level 2 early warnings are matched with off-peak window planned maintenance, scheduled during off-peak hours or early morning of the next day; Level 3 alerts are matched with daily inspections for enhanced attention, and are included in the weekly maintenance plan for unified handling.

9. The predictive operation and maintenance management method for hump air compressor stations according to claim 4, characterized in that, The pressure stabilization and energy-saving control also includes: real-time monitoring of the running time, temperature rise and load rate of each air compressor, dynamically balancing the cumulative running time of each unit to avoid premature aging caused by long-term high-load operation of a single device; when the main equipment malfunctions, it automatically switches to the standby unit within seconds to ensure uninterrupted air supply.

10. The predictive operation and maintenance management method for hump air compressor stations according to any one of claims 4 to 9, characterized in that, The digital twin 3D model is constructed by collecting data on the building structure, equipment layout, and pipeline routing of the air compressor station, and using 3D modeling software to build a 1:1 scale virtual model. The model includes the shape, location, connection relationship, and process parameter labels of all core equipment.

Images