Multi-node collaborative smart park digital operation and maintenance system and control method

Abstract

The invention relates to the technical field of intelligent park digital operation and maintenance, in particular to a multi-node collaborative intelligent park digital operation and maintenance system and a control method, and the method comprises the following steps: collecting multi-source operation data, and executing protocol conversion, time synchronization and local jitter removal processing at a gateway end; sending the processed collected data to a central access layer, carrying out authentication and flow limiting on the data, and writing the data into message middleware; mapping data accessed by the message-oriented middleware to a unified asset model and performing unit and semantic standardization; writing the time sequence data into a time sequence database, and writing the event and the audit information into an event and audit library; and performing missing interpolation, quality labeling and preliminary filtering on the data. According to the invention, through a hierarchical alarm and strategy scheduling mode, the accuracy of abnormity identification and the timeliness of operation and maintenance processing are improved, the frequency of manual intervention is reduced, and the stability and reliability of system operation and the overall efficiency of intelligent park digital operation and maintenance are improved.

Description

A multi-node collaborative smart park digital operation and maintenance system and control method Technical Field

[0001] This invention relates to a multi-node collaborative smart park digital operation and maintenance system and control method, belonging to the field of smart park digital operation and maintenance technology. Background Technology

[0002] While existing smart park operation and maintenance technologies have deployed a large number of sensing terminals and management platforms, most systems still rely on single-node data collection and centralized processing. There is a lack of collaborative control and status sharing mechanisms between different operation and maintenance nodes, making it difficult to form a unified operational situation awareness. Especially in scenarios involving the parallel operation of multiple subsystems such as power supply, equipment operation, environmental monitoring, and security, existing technologies generally suffer from heterogeneous data sources, poor time synchronization, and low information utilization.

[0003] Meanwhile, existing park operation and maintenance systems mostly use single alarms or simple threshold judgment methods, which cannot be combined with the operating status of multiple nodes for comprehensive evaluation. This results in a high false alarm rate, reliance on manual experience for fault location, and difficulty in taking timely linkage control measures. Therefore, there is an urgent need to improve a multi-node collaborative smart park digital operation and maintenance system and control method to solve the above-mentioned problems. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-node collaborative smart park digital operation and maintenance system and control method to solve the problems that existing park operation and maintenance systems mostly use single alarms or simple threshold judgment methods, which cannot combine the operating status of multiple nodes for comprehensive evaluation, resulting in high false alarm rates, reliance on manual experience for fault location, and difficulty in taking timely linkage control measures.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multi-node collaborative smart park digital operation and maintenance system and control method, comprising the following steps: collecting multi-source operational data and performing protocol conversion, time synchronization, and local debouncing processing at the gateway; sending the processed collected data to the central access layer, and authenticating, rate-limiting, and writing the data to the message middleware; mapping the data accessed by the message middleware to a unified asset model and performing unit and semantic standardization; writing time-series data to a time-series database, and writing events and audit information to the event and audit databases; performing missing data imputation, quality labeling, and preliminary filtering on the data; and pre-processing... The processed data is used for detection, rule matching, and anomaly filtering based on a real-time rule engine, and alarms or control actions are triggered when the triggering conditions are met. The triggered alarms are routed according to hierarchical, suppression, and aggregation strategies, and work orders are generated and sent to the operation and maintenance end for matters requiring manual handling to complete closed-loop processing. Under the condition of meeting security verification and policy permission, the central platform issues control commands to edge nodes or devices, and the edge fallback module takes over control according to predefined security policies in the event of network interruption or central failure. The system records tamper-proof audit logs of alarms, control, and work orders, and supports policy simulation, gray-scale verification, and playback based on digital twins.

[0006] Preferably, the real-time rule engine performs detection, rule matching, and anomaly screening using deterministic and statistical methods as the first-level judgment method. The method includes at least one of the following or a combination thereof: fixed threshold judgment, sliding window moving average or exponentially weighted moving average, cumulative and change point detection, Z-score-based standardized detection, minimum duration, debouncing rules, and time window-based statistical anomaly detection.

[0007] Preferably, in triggering alarms or control actions, a dynamic baseline and residual determination based on load perception are further set as supplementary determinations to the deterministic and statistical basis methods. The supplementary determination process is as follows: Baseline construction: Based on the sampling data of the most recent N time steps, a moving linear regression or a linear regression with L2 regularization is used to fit the monitored quantity at runtime. With one or more impact factors To obtain the dynamic regression coefficients With bias And calculate baseline estimates Residual calculation: Calculate the residuals The residuals are then incorporated into the decision flow; residual standardization and threshold determination: EWMA or a sliding window is used to estimate the local mean of the residuals. With scale and with conditions Furthermore, a duration ≥ τ serves as the triggering condition for Layer 2 anomaly detection; Evidence generation: When the aforementioned Layer 2 anomaly detection triggering condition is met, an evidence package is generated and reported along with the alarm for subsequent handling and auditing. The evidence package contains at least regression coefficients. Bias Baseline sequence , residual sequence With the decision window.

[0008] Preferably, the parameters used in baseline construction and residual determination are configurable and subject to engineering constraints, including: regression sliding window length. EWMA smoothing coefficient , The threshold coefficient k and duration τ are used; the regression is calculated using L2 regularization, and the changes in the regression coefficients are recorded for each regression.

[0009] Preferably, the baseline construction includes at least one of the following extension modes: a) generalizing univariate linear regression to multiple linear regression. a) When there is a significant nonlinear relationship between the monitored quantity and the influencing factors, a lightweight nonlinear regression model such as piecewise multinomial regression, explanatory decision tree regression, or gradient boosting tree is used instead of linear regression; c) In cases of strong noise or uneven sampling intervals, a state-space filter with input is used; After any of the above extended modes is enabled, its model structure, training, update strategy, and model version information are recorded and included in the audit and verification process.

[0010] Preferably, the load-aware residual determination is deployed in the second-layer determination process of the multi-layer detection architecture, which includes the following layers in the following order: the first layer is a low-latency safety hard threshold path for second-level response; the second layer is the load-aware residual detection for false alarm suppression and near-real-time screening; the third layer is a multivariate aggregation or machine learning model for final confirmation of complex scenarios; when the second layer determines an anomaly, the system is allowed to automatically execute control actions only after at least one of the following confirmation conditions is met: multi-sensor consistency confirmation is passed, the anomaly score or confidence level of the third-layer model exceeds a preset threshold, and the anomaly is verified by regression simulation in grayscale devices or digital twins; otherwise, only a work order is generated and manual confirmation is required before control actions can be executed.

[0011] Preferably, it further includes an adaptive recalibration and concept drift detection step: when a mutation occurs in the baseline coefficient sequence or a drift warning is triggered by a drift detection algorithm, at least one of the following processes is automatically executed: suspending the issuance of the automated threshold and switching to a conservative security policy, generating a recalibration work order for the operation and maintenance personnel and requesting manual confirmation, starting a baseline re-estimation process to refit the model on historical and recent data, or performing gray-box verification in a sandbox / digital twin environment and updating the online model after passing the verification; and all recalibration, model update, and drift trigger events are written into an immutable audit log and incorporated into the playback and verification process.

[0012] Preferably, the report of the warning and control operation is accompanied by a verifiable evidence package, which includes but is not limited to: the original time series segment at the triggering moment, the corresponding influence factor sequence, the baseline estimation sequence, the residual sequence, the regression or model coefficients, the decision threshold, the decision duration, and the trigger rule / model identifier; the evidence package is saved in a verifiable manner; and when it is determined to be a misoperation through work order verification or automatic regression check afterwards, the system automatically triggers a control rollback process and records the rollback evidence in the audit log to complete the closed-loop verification.

[0013] Preferably, it includes multi-source field nodes, which are used for distributed deployment in the power supply and distribution, HVAC, security, fire protection, lighting, water supply and drainage, and environmental monitoring subsystems in the park to collect corresponding operation status data and environmental parameter data; a data collection and access module, which is used for protocol parsing, time synchronization, data verification, and unified formatting of the data from the multi-source field nodes, and accessing the processed data to the central platform; a central operation and maintenance platform, which is used for asset model binding, data storage, and unified management of the accessed data, and centrally monitoring the operation status of each node based on a rule engine and policy configuration; a real-time analysis and warning module, which is used for performing real-time analysis, anomaly detection, and event identification on the operation data of multiple nodes in the park, and generating warning information when preset conditions are met; a collaborative control and policy distribution module, which is used for issuing control instructions to the corresponding field nodes or edge control units according to the warning results, operation and maintenance policies, or dispatching instructions to achieve collaborative linkage control between multiple nodes; an operation and maintenance management and audit module, which is used for work order management of the warning handling, control execution, and manual intervention processes, and auditing and tracing the system operation data, control behaviors, and policy changes; wherein, the central operation and maintenance platform realizes information sharing, status coordination, and linkage control between multiple nodes in the park through a unified data model and control policy, thereby improving the overall efficiency and reliability of digital operation and maintenance in the smart park.

[0014] The present invention has at least the following beneficial effects: By constructing a multi-node collaborative digital operation and maintenance control method for smart parks, the present invention realizes the unified collection, fusion analysis, and collaborative decision-making of operational data from each operation and maintenance node, enabling the overall operational status of the park to be continuously and accurately perceived and evaluated; through a hierarchical control and linkage execution mechanism, while ensuring rapid local response, it achieves coordinated control across nodes, effectively reducing false alarms and control deviations caused by load fluctuations, inconsistent states, or single-point failures; at the same time, through hierarchical alarms and strategy scheduling, it improves the accuracy of anomaly identification and the timeliness of operation and maintenance handling, reduces the frequency of manual intervention, and enhances the stability, reliability, and overall efficiency of digital operation and maintenance of the smart park. Attached Figure Description

[0015] Figure 1 is a flowchart of a control method for a multi-node collaborative smart park digital operation and maintenance system according to the present invention; Figure 2 is a flowchart of a supplementary determination method based on determinism and statistics according to the present invention; Figure 3 is a system block diagram of a multi-node collaborative smart park digital operation and maintenance system according to the present invention. Detailed Implementation

[0016] The following will describe in detail the implementation of this application with reference to the accompanying drawings and embodiments, so that the implementation process of how this application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0017] As shown in FIGS. 1 - 3, a multi - node collaborative digital operation and maintenance system and control method for a smart park provided in this embodiment: The system includes multi - source on - site nodes, which are used for distributed deployment in the power supply and distribution, heating, ventilation and air conditioning, security, fire protection, lighting, water supply and drainage, and environmental monitoring subsystems in the park, and collect corresponding operation status data and environmental parameter data; a data collection and access module, which is used for protocol parsing, time synchronization, data verification and unified formatting of the data from the multi - source on - site nodes, and accesses the processed data to the central platform; a central operation and maintenance platform, which is used for binding asset models, storing and uniformly managing the accessed data, and centrally monitoring the operation status of each node based on a rule engine and policy configuration; a real - time analysis and alarm module, which is used for performing real - time analysis, anomaly detection and event identification on the operation data of multiple nodes in the park, and generating alarm information when preset conditions are met; a collaborative control and policy distribution module, which is used for issuing control instructions to corresponding on - site nodes or edge control units according to alarm results, operation and maintenance policies or scheduling instructions to achieve collaborative linkage control between multiple nodes; an operation and maintenance management and audit module, which is used for work - order management of the alarm handling, control execution and manual intervention processes, and auditing and tracing the system operation data, control behaviors and policy changes; wherein, the central operation and maintenance platform realizes information sharing, status coordination and linkage control between multiple nodes in the park through a unified data model and control strategy, thereby improving the overall efficiency and reliability of digital operation and maintenance of the smart park.

[0018] The process is as follows: Step 1: In the subsystems of power supply and distribution, HVAC, security, fire protection, lighting, water supply and drainage, and environmental monitoring within the smart park, multiple field operation nodes are set up respectively. Each field node collects the operating status data and environmental parameter data of the corresponding equipment in real time through sensors, controllers, or smart terminals. Protocol parsing, time synchronization, data debouncing, outlier screening, and basic quality verification are completed at the node side or data acquisition gateway to ensure that the collected data has a unified time sequence and basic reliability before entering the central system. Step 2: The pre-processed data is uploaded to the central operation and maintenance platform by the data acquisition and access module. The central operation and maintenance platform performs unified authentication, flow control, and format standardization on the data, and maps and binds the data to a unified asset model to realize the integration of different subsystems, different manufacturers, and different... Unified semantic management of nodes within the same protocol provides a data foundation for subsequent centralized analysis and collaborative control; Step 3: The central operation and maintenance platform writes the standardized continuous operation data into the time-series database, and writes event information, alarm information, control commands, and manual operation records into the event database and audit database; Before or during the data entry process, missing value imputation, noise filtering, and quality labeling are performed on the data to meet the needs of real-time analysis, historical backtracking, and audit traceability; Step 4: After completing data storage and preprocessing, the system enters the real-time analysis and control judgment stage; The central operation and maintenance platform uses a real-time rule engine to perform first-level anomaly judgment on multi-node operation data using deterministic and statistical methods, including but not limited to: fixed threshold judgment, sliding window mean analysis, exponentially weighted moving average analysis, and CUSUM. Change point detection, Z-score normalization detection, and minimum duration and de-jitter judgment rules; the first-layer judgment is used to quickly screen potential anomalies to ensure that the system has basic response capabilities under second-level or near-real-time conditions, and to provide triggering conditions for subsequent fine-grained judgment; Step 5: In view of the problem that the park system has significant load fluctuations under different operating conditions and that relying solely on fixed thresholds and statistical methods is prone to false alarms, based on the first-layer judgment, the system further introduces a load-aware dynamic baseline and residual judgment mechanism as a second-layer judgment in the real-time rule engine; the second-layer judgment includes the following complete sub-processes: 1) Baseline construction: Based on the sampling data of the most recent N time steps, a sliding linear regression or a linear regression with L2 regularization is used to fit the monitored quantity at runtime. With one or more impact factors To obtain the dynamic regression coefficients With bias And calculate baseline estimates 2) Residual calculation: Calculate the residuals The residuals are then incorporated into the decision flow; 3) Residual standardization and threshold determination: EWMA or a sliding window is used to estimate the local mean of the residuals. With scale and with conditions And the duration ≥ τ is used as the triggering condition for the second-level anomaly determination; 4) Evidence generation: When the second-level anomaly determination triggering condition is met, an evidence package is generated and reported with the alarm for subsequent handling and auditing. The evidence package contains at least regression coefficients. Bias Baseline sequence , residual sequence With the decision window; Step 6: In the dynamic baseline and residual determination process, the system performs engineered configuration and constraint management of parameters such as regression window length, smoothing coefficient, threshold coefficient, and duration, and introduces a regularization mechanism into the regression model to prevent parameter (parameters include regression sliding window length) from being used for other purposes. EWMA smoothing coefficient , The threshold coefficient k and duration τ) oscillate and the model becomes unstable; at the same time, the trajectory of model parameter changes is recorded to provide a basis for subsequent trend analysis and model optimization; Step 7: When the detected object has multiple load coupling or obvious nonlinear characteristics, the system supports the following during the dynamic baseline construction process: extending univariate linear regression to multiple linear regression. To simultaneously consider multiple influencing factors; when there is a significant nonlinear relationship between the monitored quantity and the influencing factors, lightweight nonlinear regression models such as piecewise multinomial regression, explanatory decision tree regression, or gradient boosting tree are used instead of linear regression; in cases of strong noise or uneven sampling intervals, a state-space filter with input is used; model selection, switching, and version information are uniformly incorporated into the platform management and auditing process; Step 8: Based on the first-layer deterministic judgment and the second-layer load-aware residual judgment, the system forms a multi-layer anomaly confirmation architecture; the multi-layer detection architecture includes the following in order: the first layer is a low-latency safety hard threshold path for second-level response; the second layer is the load-aware residual detection for false alarm suppression and near-real-time screening; the third layer is a multivariate aggregation or machine learning model for final confirmation of complex scenarios; only when the anomaly simultaneously meets the multi-sensor consistency, model confidence, or simulation verification criteria... When a condition is met, the system will allow automatic control actions to be triggered; otherwise, only alarms or maintenance work orders will be generated to avoid operational risks caused by erroneous control triggers. Step 9: During the execution of control commands, the system continuously monitors the model residuals and parameter changes. When model failure, changes in operating characteristics, or concept drift are detected, the relevant automatic control strategies will be automatically paused, and a recalibration process will be triggered or verification will be performed in the digital twin environment. The system will be restored to online operation after successful verification to ensure long-term operational stability. Step 10: Throughout the alarm and control process, the system automatically generates and stores a complete evidence package containing original data, baselines, residuals, model parameters, judgment rules, and execution results, and saves it in an immutable manner. When subsequent verification determines that the action is erroneous, the system automatically executes control rollback and records the rollback process and results in the audit system, forming a complete and traceable digital operation and maintenance closed loop.

[0019] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or system comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the product or system that includes that element.

Claims

1. A control method for a multi-node collaborative smart park digital operation and maintenance system, characterized in that, Includes the following steps: Collect multi-source operational data and perform protocol conversion, time synchronization, and local debouncing at the gateway; send the processed collected data to the central access layer, where data authentication, rate limiting, and writing to the message middleware are performed; map the data accessed by the message middleware to the unified asset model and perform unit and semantic standardization; write time-series data to the time-series database, and write events and audit information to the event and audit databases; perform missing data imputation, quality labeling, and preliminary filtering on the data; The preprocessed data is subjected to detection, rule matching, and anomaly filtering based on a real-time rule engine, and alarms or control actions are triggered when the triggering conditions are met. The triggered alarms are routed according to hierarchical, suppression, and aggregation strategies, and work orders are generated and sent to the operation and maintenance end for matters requiring manual handling to complete closed-loop processing. Under the condition of meeting security verification and policy permission, the central platform issues control commands to edge nodes or devices, and the edge fallback module takes over control according to predefined security policies in the event of network interruption or central failure. The system records tamper-proof audit logs of alarms, control and work orders, and supports policy simulation, gray-scale verification and playback based on digital twins.

2. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 1, characterized in that: The real-time rule engine employs deterministic and statistical methods as its first-level judgment methods for execution detection, rule matching, and anomaly screening. These methods include at least one of the following or a combination thereof: fixed threshold judgment, sliding window moving average or exponentially weighted moving average, cumulative and change point detection, Z-score-based standardized detection, minimum duration, debouncing rules, and time window-based statistical anomaly detection.

3. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 1, characterized in that: In triggering alarms or control actions, a dynamic baseline and residual determination based on load perception are further set as supplementary determinations to the aforementioned deterministic and statistical basis methods. The process of the supplementary determination is as follows: Baseline construction: Based on the sampling data of the most recent N time steps, a moving linear regression or a linear regression with L2 regularization is used to fit the monitored quantity at runtime. With one or more impact factors Obtain the dynamic regression coefficients With bias And calculate baseline estimates Residual calculation: Calculate the residuals The residuals are then incorporated into the decision flow; residual standardization and threshold determination: EWMA or a sliding window is used to estimate the local mean of the residuals. With scale and with conditions Furthermore, a duration ≥ τ serves as the triggering condition for Layer 2 anomaly detection; Evidence generation: When the aforementioned Layer 2 anomaly detection triggering condition is met, an evidence package is generated and reported along with the alarm for subsequent handling and auditing. The evidence package contains at least regression coefficients. Bias Baseline sequence residual sequence With the decision window.

4. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 3, characterized in that: The parameters used in baseline construction and residual determination are configurable and subject to engineering constraints. These parameters include: regression sliding window length. EWMA smoothing coefficient , The threshold coefficient k and duration τ are used; the regression is calculated using L2 regularization, and the changes in the regression coefficients are recorded for each regression.

5. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 3, characterized in that: The baseline construction includes at least one of the following extension patterns: a) generalizing univariate linear regression to multiple linear regression. a) When there is a significant nonlinear relationship between the monitored quantity and the influencing factors, a lightweight nonlinear regression model such as piecewise multinomial regression, explanatory decision tree regression, or gradient boosting tree is used instead of linear regression; c) In cases of strong noise or uneven sampling intervals, a state-space filter with input is used; After any of the above extended modes is enabled, its model structure, training, update strategy, and model version information are recorded and included in the audit and verification process.

6. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 3, characterized in that: The load-aware residual determination is deployed in the second-layer determination process of the multi-layer detection architecture, which includes the following layers in the following order: the first layer is a low-latency safety hard threshold path for second-level response; the second layer is the load-aware residual detection for false alarm suppression and near-real-time screening; the third layer is a multivariate aggregation or machine learning model for final confirmation of complex scenarios; when the second layer determines an anomaly, the system is allowed to automatically execute control actions only after at least one of the following confirmation conditions is met: multi-sensor consistency confirmation is passed, the anomaly score or confidence level of the third-layer model exceeds a preset threshold, and the anomaly is verified by regression simulation in grayscale devices or digital twins; otherwise, only a work order is generated and manual confirmation is required before control actions can be executed.

7. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 3, characterized in that: It also includes an adaptive recalibration and concept drift detection step: when a mutation occurs in the baseline coefficient sequence or a drift warning is triggered by a drift detection algorithm, at least one of the following processes is automatically executed: pausing the issuance of the automated threshold and switching to a conservative security policy, generating a recalibration work order for the operation and maintenance personnel and requiring manual confirmation, starting a baseline re-estimation process to refit the model on historical and recent data, or performing gray-box verification in a sandbox / digital twin environment and updating the online model after passing the verification; and all recalibration, model update, and drift trigger events are written into an immutable audit log and incorporated into the playback and verification process.

8. The control method for a multi-node collaborative smart park digital operation and maintenance system according to claim 3, characterized in that: The reporting of the warning and control operations is accompanied by a verifiable evidence package, which includes but is not limited to: the original time series segment at the trigger moment, the corresponding impact factor sequence, the baseline estimation sequence, the residual sequence, the regression or model coefficients, the decision threshold, the decision duration, and the trigger rule / model identifier; the evidence package is saved in a verifiable manner; and when it is determined to be a misoperation through work order verification or automatic regression inspection afterwards, the system automatically triggers a control rollback process and records the rollback evidence in the audit log to complete the closed-loop verification.

9. A multi-node collaborative smart park digital operation and maintenance system based on any one of claims 1-8, characterized in that: It includes multi-source field nodes, which are used for distributed deployment in the power supply and distribution, heating, ventilation and air conditioning, security, fire protection, lighting, water supply and drainage, and environmental monitoring subsystems in the park to collect the corresponding operation status data and environmental parameter data; a data collection and access module, which is used to perform protocol parsing, time synchronization, data verification, and unified formatting on the data from the multi-source field nodes, and access the processed data to the central platform; a central operation and maintenance platform, which is used to bind the asset model, store and uniformly manage the accessed data, and centrally monitor the operation status of each node based on a rule engine and policy configuration; a real-time analysis and warning module, which is used to perform real-time analysis, anomaly detection, and event identification on the operation data of multiple nodes in the park, and generate warning information when preset conditions are met; a collaborative control and policy distribution module, which is used to issue control instructions to the corresponding field nodes or edge control units according to the warning results, operation and maintenance policies, or dispatching instructions to achieve collaborative linkage control between multiple nodes; an operation and maintenance management and audit module, which is used to manage the warning handling, control execution, and manual intervention processes in a work order manner, and audit and trace the system operation data, control behavior, and policy changes; among which, the central operation and maintenance platform realizes information sharing, status coordination, and linkage control between multiple nodes in the park through a unified data model and control policy, thereby improving the overall efficiency and reliability of the digital operation and maintenance of the smart park.

Images