Intelligent park digital management and control method and system based on digital twinning

Abstract

This invention discloses a digital twin-based intelligent park management and control method and system. It aggregates and processes raw sensor data collected by a sensor network in real time to form multi-scenario fusion data. A mapping model between physical and virtual entities is constructed using modeling technology. Parameters are dynamically synchronized to monitor equipment status. When parameters exceed thresholds, machine learning algorithms are used to analyze the source of deviations, identify anomalies, and generate analysis results. Rules are then obtained from an intelligent decision-making system to generate an execution instruction sequence, driving the automated operation and maintenance module to adjust the virtual park model in real time. Ultimately, this achieves optimized management and control. This invention constructs a closed-loop link from perception to decision-making to execution, significantly improving the intelligence level and response efficiency of park equipment management and ensuring operational stability.

Description

Technical Field

[0001] This invention relates to the field of smart park management, and in particular discloses a smart park digital control method and system based on digital twins. Background Technology

[0002] Smart park management, a crucial field concerning urban development and resource optimization, bears the key mission of improving operational efficiency, ensuring safety, and achieving sustainable development. With the acceleration of urbanization, park management has gradually become an indispensable part of urban operations, and its level of intelligence directly impacts the vitality of the regional economy and the quality of life for residents. However, how to achieve comprehensive perception and efficient decision-making within the park through technological means remains a key area requiring breakthroughs.

[0003] Currently, although many park management solutions have introduced some digital tools, they generally suffer from data silos and system fragmentation. Information from different scenarios often cannot be effectively integrated; for example, data on security, energy consumption, and equipment maintenance operate independently, lacking a unified platform for collaborative processing. This fragmented management approach leads to inaccurate resource allocation, slow response times, and difficulty in coping with complex and dynamic environmental changes.

[0004] More importantly, the connection between physical and digital spaces often remains superficial, failing to achieve true real-time interaction and deep integration between the two. Against this backdrop, a core technical challenge for park management is how to construct a real-time mapping system between physical and digital spaces. This factor is crucial because various dynamic changes in the physical space, such as personnel flow, equipment status, and environmental parameters, need to be reflected in the digital space in real time for analysis and decision-making. If this cannot be achieved, the digital space cannot accurately represent the true state of the park, thus affecting the efficiency of subsequent intelligent judgment and management. Furthermore, the lack of such a mapping system can lead to delays or distortions in information transmission; for example, in emergencies, managers may miss the optimal intervention opportunity due to untimely data updates.

[0005] Therefore, establishing a digital mapping system that can accurately and in real-time reflect the physical space status within a complex and ever-changing park environment, and using this system to achieve efficient scheduling and management of various resources within the park, has become a key issue in the construction of smart parks. For example, in a specific scenario where equipment malfunctions suddenly in a certain part of the park, if the digital space cannot immediately obtain abnormal information about the equipment's operation and synchronize it to the management platform, the administrator will be unable to quickly locate the problem and arrange repairs, and delays may even lead to greater safety hazards.

[0006] This problem directly reflects the practical difficulties caused by the insufficient mapping between physical and digital spaces, and urgently needs to be solved by technical means. Summary of the Invention

[0007] This invention provides a digital twin-based intelligent park management and control method and system, aiming to solve at least one of the defects in the prior art.

[0008] One aspect of the present invention relates to a digital twin-based intelligent park management and control method, comprising the following steps:

[0009] S100: Collect raw sensing data on the operating status of building pipeline equipment in the park from the sensor network, and perform real-time aggregation processing on the sensing data through the data processing system to obtain aggregated multi-scene fusion data;

[0010] S200. For the aggregated multi-scenario fusion data, modeling techniques are used to construct a mapping model between physical entities and virtual entities, and the dynamic synchronization parameters included in the mapping model are determined.

[0011] S300. If the dynamic synchronization parameters in the mapping model exceed the threshold, the parameter deviation is analyzed by machine learning algorithm to determine that the source of the parameter deviation is an abnormal operation of the equipment, and the abnormality analysis result is obtained.

[0012] S400. Based on the anomaly analysis results, obtain the corresponding decision rules from the intelligent decision-making system and generate an execution instruction sequence for the anomaly. The execution instruction sequence is used to form a closed-loop link from perception modeling to analysis and decision-making.

[0013] The S500 drives the automated operation and maintenance module by executing instruction sequences, and adjusts the mapping model in the virtual park in real time to obtain an optimized digital and intelligent management and control status.

[0014] Further, step S100 includes:

[0015] S110. Obtain raw sensing data from the sensor network of building pipeline equipment in the park, perform preliminary cleaning processing on the raw sensing data, remove noise and outliers through data filtering tools, and obtain the cleaned basic data stream.

[0016] S120. Based on the cleaned basic data stream, a real-time aggregation tool is used to perform multi-dimensional integration processing on the basic data stream. If the total amount of the basic data stream exceeds the preset threshold, the load is distributed through a sharding process to determine the aggregated intermediate dataset.

[0017] S130. Using a data fusion tool, the multi-scene perception information in the intermediate dataset is correlated and mapped to obtain a fused comprehensive data view. It is then determined whether the comprehensive data view meets the multi-scene consistency requirements to obtain the aggregated multi-scene fused data.

[0018] Further, step S200 includes:

[0019] S210. Obtain the operational status information of physical entities from the aggregated multi-scenario fusion data, and use data extraction tools to classify and organize the operational status information to determine the core feature set of physical entities.

[0020] S220. Based on the core feature set, a mapping model between physical entities and virtual entities is constructed using modeling techniques. At the same time, an association mapping tool is used to match the physical features in the mapping model with the virtual template. If the matching degree is lower than a preset threshold, it is corrected by an adjustment tool to obtain a preliminary mapping relationship.

[0021] S230. Obtain the dynamic synchronization parameters in the initial mapping relationship through real-time monitoring tools.

[0022] Further, step S300 includes:

[0023] S310. Obtain real-time data of dynamic synchronization parameters from the mapping model, classify and organize the dynamic synchronization parameters using data acquisition tools, record parameters that exceed the preset threshold, and determine the set of abnormal parameters.

[0024] S320. Based on the set of abnormal parameters, a data comparison tool is used to match the abnormal parameters with the normal range in the historical records. If the deviation between the abnormal parameters and the normal range in the historical records exceeds a preset threshold, the source of the deviation is located by the deviation analysis tool, and the deviation source classification is obtained.

[0025] S330. Based on the classification of deviation sources, use status monitoring tools to obtain the status data of the equipment during operation, compare it with the pre-established normal status template, determine whether it belongs to the abnormal operation of the equipment, and obtain the abnormality analysis results.

[0026] Further, step S400 includes:

[0027] S410. Extract key deviation data from the anomaly analysis results, use data processing tools to classify the key deviation data, identify at least one source of deviation, and obtain the classified deviation dataset.

[0028] S420. Based on the classified bias dataset, obtain the pre-established decision rule base and compare it with the rule matching tool. If the matching degree between the classified bias dataset and the pre-established decision rule base exceeds the preset threshold, then determine the matching decision rule set.

[0029] S430. For the matching decision rule set, use an instruction generation tool to convert the rules into an executable instruction sequence, obtain the priority and dependency relationship, determine the execution order, and obtain an ordered execution instruction sequence.

[0030] S440. The ordered sequence of execution instructions is distributed to the perception modeling module using a data transmission tool. During distribution, the closed-loop link status is updated synchronously to determine the instruction execution feedback.

[0031] Further, step S500 includes:

[0032] S510. When distributing the execution instruction sequence, the status information in the data synchronization tool is updated synchronously to determine the priority of the execution instruction sequence and obtain an ordered instruction execution queue.

[0033] S520 drives the automated operation and maintenance module through an ordered instruction execution queue. The automated operation and maintenance module adjusts the mapped area of ​​the virtual park in real time, records status feedback data during the adjustment, and determines the optimized digital management and control status.

[0034] Another aspect of the present invention relates to a digital twin-based intelligent park management and control system, used to execute the aforementioned digital twin-based intelligent park management and control method, comprising:

[0035] The multi-scenario fusion data acquisition module is used to collect raw sensing data on the operating status of building pipeline equipment in the park from the sensor network, and to aggregate the sensing data in real time through the data processing system to obtain aggregated multi-scenario fusion data.

[0036] The dynamic synchronization parameter determination module is used to construct a mapping model between physical entities and virtual entities for aggregated multi-scenario fusion data, and to determine the dynamic synchronization parameters contained in the mapping model.

[0037] The anomaly analysis result acquisition module is used to analyze the parameter deviation through machine learning algorithms if the dynamic synchronization parameter in the mapping model exceeds the threshold, determine whether the source of the parameter deviation belongs to the abnormal operation of the equipment, and obtain the anomaly analysis result.

[0038] The execution instruction sequence generation module is used to obtain corresponding decision rules from the intelligent decision-making system based on the anomaly analysis results and generate an execution instruction sequence for the anomaly. The execution instruction sequence is used to form a closed-loop link from perception modeling to analysis and decision-making.

[0039] The digital management and control status acquisition module is used to drive the automated operation and maintenance module by executing command sequences to adjust the mapping model in the virtual park in real time and obtain an optimized digital management and control status.

[0040] The beneficial effects achieved by this invention are as follows:

[0041] This invention provides a digital twin-based intelligent park management and control method and system, addressing the unique business scenario of monitoring and handling anomalies in the operational status of building pipeline network equipment within the park. Specifically, it addresses how to achieve closed-loop management from sensing data to decision execution, ensuring timely detection and optimization of equipment malfunctions. This invention aggregates raw sensing data collected by sensor networks in real time to form multi-scenario fused data. It then uses modeling techniques to construct a mapping model between physical and virtual entities, dynamically synchronizing parameters to monitor equipment status. When parameters exceed thresholds, machine learning algorithms analyze the source of deviations, identify anomalies, and generate analysis results. Rules are then obtained from the intelligent decision-making system to generate execution command sequences, driving the automated operation and maintenance module to adjust the virtual park model in real time, ultimately achieving optimized management and control. The core innovation of this invention lies in constructing a closed-loop link from sensing to decision-making to execution, significantly improving the intelligence level and response efficiency of park equipment management and ensuring operational stability. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating an embodiment of a digital twin-based intelligent park management and control method according to the present invention.

[0043] Figure 2 This is a functional block diagram of an embodiment of a digital twin-based smart park digital management and control system according to the present invention.

[0044] Explanation of icon numbers:

[0045] 10. Multi-scenario fusion data acquisition module; 20. Dynamic synchronization parameter determination module; 30. Anomaly analysis result acquisition module; 40. Execution instruction sequence generation module; 50. Data-driven management and control status acquisition module. Detailed Implementation

[0046] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0047] like Figure 1 As shown, the first embodiment of the present invention proposes a smart park digital management and control method based on digital twins, including the following steps:

[0048] Step S100: Collect raw sensing data on the operating status of building pipeline equipment in the park from the sensor network, and perform real-time aggregation processing on the sensing data through the data processing system to obtain aggregated multi-scene fusion data.

[0049] The monitoring targets within the smart park are buildings (such as office buildings and dormitories), pipelines (water supply and drainage, power supply, HVAC), and terminal equipment (such as air conditioners, water pumps, electricity meters, and security cameras). Relying on a sensor network (including various types of sensors such as temperature, pressure, flow, voltage, and operational status sensors) deployed at key locations within the park, raw sensing data reflecting the operational status of physical entities is collected in real time (data types include structured data, time-series data, and unstructured data, among other heterogeneous forms). A dedicated data processing system performs real-time aggregation processing on the collected raw sensing data: first, data cleaning (removing noise, outliers, and missing values) and format standardization (unifying data format, units, and timestamps); then, redundancy removal algorithms remove duplicate data; finally, cross-source data association and fusion are performed based on park business scenarios (such as energy consumption monitoring, equipment maintenance, and security monitoring scenarios). This integrates scattered, single-dimensional data into multi-scenario fused data that covers multiple scenarios and possesses correlation and completeness, ensuring that the data accurately supports subsequent digital twin modeling and management decisions.

[0050] Step S200: For the aggregated multi-scenario fusion data, a mapping model between physical entities and virtual entities is constructed using modeling techniques, and the dynamic synchronization parameters included in the mapping model are determined.

[0051] Using the multi-scenario fusion data aggregated in step S100 as the core input, a comprehensive approach is adopted, employing various modeling techniques (such as 3D geometric modeling to restore the physical entity's form, mechanism modeling to replicate equipment operating rules, and data-driven modeling to capture parameter relationships) to construct a one-to-one mapping model between physical entities (building structures, pipelines, terminal equipment, etc.) and virtual entities in the smart park. The model must fully reproduce the spatial layout, connection relationships, and operational logic of the physical entities, while clearly defining dynamic synchronization parameters. These dynamic synchronization parameters are key indicators reflecting the real-time status of the physical entities, including equipment operating parameters (such as power, speed, and operating voltage), environmental sensing parameters (such as regional temperature, humidity, and air quality), pipeline transmission parameters (such as medium flow rate, pressure, and velocity), and energy consumption parameters (such as instantaneous power consumption and water consumption). A real-time linkage mechanism between the dynamic synchronization parameters and the physical entities must be established to ensure that the virtual model can follow the changes in the physical entity's state within milliseconds, achieving precise synchronization between the "physical state" and the "virtual model."

[0052] Step S300: If the dynamic synchronization parameters in the mapping model exceed the threshold, the parameter deviation is analyzed by machine learning algorithm to determine that the source of the parameter deviation belongs to the abnormal operation of the equipment, and the abnormality analysis result is obtained.

[0053] The real-time monitoring step S200 constructs a mapping model to dynamically synchronize parameters (such as equipment operating parameters, environmental sensing parameters, and pipeline transmission parameters), and compares them with preset normal operating thresholds (set according to equipment rating standards, park operating specifications, and historical best operating conditions). If a parameter exceeds the threshold range, a machine learning algorithm (such as anomaly detection algorithm, decision tree, neural network, support vector machine, etc.) is triggered to conduct in-depth analysis of the parameter deviation. The machine learning algorithm mines deviation features, correlates multi-source data (historical operating data, equipment mechanism data, environmental impact data, etc.), and learns parameter anomaly patterns to determine whether the source of the deviation belongs to equipment operating anomalies (including equipment failure, performance degradation, parameter drift, overload, environmental interference, etc.). Finally, it outputs anomaly analysis results including anomaly type, location, severity, and possible causes, providing accurate basis for subsequent decision-making.

[0054] Step S400: Based on the anomaly analysis results, obtain the corresponding decision rules from the intelligent decision system and generate an execution instruction sequence for the anomaly, wherein the execution instruction sequence is used to form a closed-loop link from perception modeling to analysis and decision-making.

[0055] Taking the anomaly analysis results obtained in step S300 (including anomaly type, location, severity, and possible causes) as input, the system calls the pre-set decision rule base in the intelligent decision-making system. This decision rule base is built based on park operation and maintenance specifications, equipment fault handling procedures, and safe operation standards, and includes the mapping relationship between different anomaly scenarios and corresponding handling strategies (e.g., equipment overload anomaly corresponds to the "load reduction + alarm" rule, and pipeline leakage anomaly corresponds to the "valve closure + dispatch for repair" rule). The system accurately matches the optimal decision rule according to the specific characteristics of the anomaly and generates an ordered sequence of execution instructions according to the execution logic priority (e.g., equipment parameter adjustment instructions, emergency start / stop instructions, operation and maintenance personnel dispatch instructions, system linkage instructions, etc.). This sequence of execution instructions not only clarifies the specific operation content and execution order, but its core function is to connect front-end data perception, virtual and real modeling, anomaly analysis, and back-end execution implementation, forming a closed-loop link of "perceived data → modeling mapping → anomaly analysis → decision instructions → execution feedback," ensuring deep collaboration between decision-making and front-end links and precise adaptation of execution actions to anomaly scenarios.

[0056] Step S500: Drive the automated operation and maintenance module by executing the instruction sequence to adjust the mapping model in the virtual park in real time, thereby obtaining an optimized digital and intelligent management and control status.

[0057] Driven by the execution instruction sequence generated in step S400, the automated operation and maintenance modules deployed in the park (such as intelligent equipment controllers, remote operation platforms, robot operation and maintenance systems, pipeline control terminals, etc.) are triggered. On the one hand, targeted operations (such as parameter adjustment, emergency start / stop, fault repair triggering, etc.) are performed on entities (equipment, pipelines, building systems, etc.) with abnormalities in the physical park. On the other hand, the mapping model in the virtual park is adjusted in real time. Based on the actual execution results and status changes of the physical entities, the dynamic synchronization parameters in the model are updated, model deviations are corrected, and model operation logic is optimized to ensure that the virtual model and the actual state of the physical park entities remain highly consistent. Through the coordinated actions of "physical execution - virtual synchronization - status verification", potential abnormalities are continuously eliminated and resource allocation is optimized. Ultimately, an optimized digital and intelligent management state is formed, in which the physical park and the virtual park are optimized in synergy, and the management efficiency and stability are significantly improved, laying the foundation for the continuous optimization of subsequent park operations.

[0058] Furthermore, the intelligent management and control method for smart parks based on digital twins provided in this embodiment includes step S100 as follows:

[0059] Step S110: Obtain raw sensing data from the sensor network of building pipeline equipment in the park, perform preliminary cleaning processing on the raw sensing data, remove noise and outliers through data filtering tools, and obtain the cleaned basic data stream.

[0060] The following formula is used to identify and process outlier data points that fall outside the normal range using statistical methods:

[0061] (1)

[0062] In formula (1), This indicates the sensor data after cleaning. Represents raw sensory data, This represents the mean of the data. Indicates the standard deviation of the data. This represents the outlier detection threshold coefficient.

[0063] The following formula is used to perform continuous processing on building pipeline network equipment data using an exponential smoothing method, ensuring the stability of the data stream:

[0064] (2)

[0065] In formula (2), This represents the output cleaned data stream. This indicates the current input data. This represents the data value at the previous moment. This represents the smoothing factor.

[0066] In a smart park's digital management system, acquiring raw sensor data from the building's pipeline network is a fundamental step. These sensors, including temperature, humidity, and flow sensors, are deployed in the building's water supply pipes, air conditioning systems, and power lines, collecting data in real time such as water pressure, air quality indicators, and energy consumption. Specifically, the acquisition process is implemented using wireless sensor network protocols such as ZigBee or LoRa. The sensors send data packets to a central gateway once per second, and the gateway then aggregates this raw sensor data onto a cloud platform. The technical goal of this approach is to ensure the real-time nature and integrity of the data, providing a reliable foundation for subsequent processing.

[0067] The raw sensor data undergoes initial cleaning by using data filtering tools to remove noise and outliers. For example, if abnormally high peak values ​​appear in the acquired temperature data stream, such as a sudden jump from 25 degrees Celsius to 100 degrees Celsius, this could be noise caused by sensor malfunction. The cleaning process employs median filtering, first sorting the data sequence and replacing outliers with the median value; or using statistical methods such as Z-score calculation, setting a threshold such as three times the standard deviation, marking and deleting any values ​​exceeding this threshold as outliers, resulting in a cleaned base data stream. This helps improve data accuracy and avoids interference with subsequent analysis. In practical applications, for an air conditioning system in a large industrial park, without cleaning, noise could lead to false alarms and affect energy management efficiency.

[0068] Step S120: Based on the cleaned basic data stream, use a real-time aggregation tool to perform multi-dimensional integration processing on the basic data stream. If the total amount of basic data stream exceeds the preset threshold, load distribution is performed through sharding to determine the aggregated intermediate dataset.

[0069] The following formula is used to define the activation conditions of the fragmentation mechanism:

[0070] (3)

[0071] In formula (3), Indicates the first Data processing volume per partition Indicates the total amount of the basic data stream. Indicates the number of fragments. This represents a preset threshold. When the total amount of basic data stream exceeds the preset threshold, the sharding mechanism is activated to distribute the total amount of data evenly among the shards for load balancing. Otherwise, no sharding is performed.

[0072] The following formula describes the computational process of generating the intermediate dataset through multiple rounds of aggregation:

[0073] (4)

[0074] In formula (4), This represents the final, determined intermediate dataset. Indicates the number of rounds in the aggregation process. Indicates the first The fusion coefficient of the wheel aggregation, Indicates the first The number of data blocks participating in aggregation in each round. Represents aggregate functions, Indicates the first Wheel of Life One data block.

[0075] Based on the cleaned basic data stream, a real-time aggregation tool is used to perform multi-dimensional integration processing. This real-time aggregation tool can be a stream processing framework based on Apache Kafka, which supports data aggregation based on multiple dimensions such as time, location, and type. For example, for electricity consumption data from different buildings, the average and sum are calculated hourly, while also integrating location dimensions such as floor grouping. If the total amount of basic data exceeds a preset threshold, such as more than 100,000 data entries per minute, load balancing is achieved through sharding. Specifically, the data stream is divided into multiple partitions, each assigned to a different processing node, and a consistent hashing algorithm is used to ensure load balancing. This method of determining the intermediate dataset after aggregation effectively handles large data volumes and avoids system bottlenecks. In park energy monitoring operations, this aggregation helps to quickly identify peak usage patterns and improve response speed.

[0076] Step S130: Use a data fusion tool to correlate and map the multi-scene perception information in the intermediate dataset to obtain a fused comprehensive data view. Determine whether the comprehensive data view meets the multi-scene consistency requirements to obtain the aggregated multi-scene fused data.

[0077] The merged, integrated data view is derived using the following formula:

[0078] (5)

[0079] In formula (5), This represents the integrated data view after merging. This indicates the total number of scenarios. Indicates the first The weighting coefficients for each scenario. Indicates the first The association mapping matrix of each scene, Indicates the first A dataset of perception information for each scene.

[0080] The following formula is used to determine whether a comprehensive data view meets the consistency requirements across multiple scenarios:

[0081] (6)

[0082] In formula (6), This represents a consistency evaluation metric across multiple scenarios. This indicates the number of feature dimensions in the comprehensive data view. Indicates the first A numerical vector with 1 feature dimension. This represents the mean vector of all scene features. The standard deviation threshold representing the consistency judgment.

[0083] The aggregated multi-scenario fusion data is obtained through the following formula:

[0084] (7)

[0085] In formula (7), This represents the aggregated multi-scenario fusion data. This represents the set of options that require aggregation for decision-making. Indicates the number of data samples. The number of dimensions representing data features. Indicates the first The first sample The correlation coefficient of each feature This represents the weight of the perceived information at the corresponding location. This represents the fusion quality assessment factor.

[0086] Data fusion tools are used to correlate and map multi-scenario perception information from intermediate datasets, resulting in a fused, comprehensive data view. For example, a fusion tool using a graph database like Neo4j (Neo4j Graph Database) can correlate electricity and temperature data, creating relationships such as "high temperature leads to increased electricity consumption." The mapping process includes feature extraction, similarity calculation, and joint queries, such as calculating the Pearson correlation coefficient to quantify the correlation strength, and then generating a view like a comprehensive dashboard displaying multi-scenario indicators. The comprehensive data view is then evaluated to determine if it meets multi-scenario consistency requirements, resulting in aggregated multi-scenario fused data. Specifically, consistency checks verify the logic between data points; for example, electricity consumption should increase accordingly with rising temperature. If this doesn't match, it's marked as inconsistent and corrected. Through this fusion, park managers can obtain a holistic view, supporting decision-making and optimizing resource allocation. The technical effect is improved system intelligence and reduced manual intervention.

[0087] Furthermore, the intelligent management and control method for smart parks based on digital twins provided in this embodiment includes step S200 as follows:

[0088] Step S210: Obtain the operational status information of physical entities from the aggregated multi-scenario fusion data, and use data extraction tools to classify and organize the operational status information to determine the core feature set of physical entities.

[0089] The following formula is used to extract comprehensive operational status information of physical entities from aggregated multi-scenario fusion data:

[0090] (8)

[0091] In formula (8), Indicates the first The operational status information of each physical entity. This indicates the total number of scenarios in the multi-scenario fusion data. Indicates the first The weighting coefficients for each scenario. Indicates the first The physical entity in the first Data values ​​in each scenario.

[0092] The core feature set of a physical entity is derived using the following formula:

[0093] (9)

[0094] In formula (9), Represents the set of core features of a physical entity. Indicates the first One candidate feature, Indicates the total number of candidate features. Indicates the first The importance ranking score of each feature This represents the threshold for feature selection.

[0095] The operational status information of physical entities is extracted from aggregated multi-scenario fusion data through a dedicated data query interface. For example, in a smart park's intelligent management and control system, this multi-scenario fusion data includes comprehensive indicators of building water supply systems, lighting equipment, and security facilities. The operational status information of physical entities, such as water pumps or lamps, covers speed, voltage, and fault signals. Data extraction tools, such as SQL (Structured Query Language)-based extraction engines, are used to classify and organize this operational status information. The specific process involves first defining classification rules, such as grouping operational status information by equipment type (water pumps into mechanical categories, lamps into electrical categories), and then using filtering algorithms to remove redundant data, ensuring that each type of information focuses on key parameters.

[0096] When determining the core feature set of a physical entity, essential attributes such as the average rotational speed, vibration amplitude, and energy consumption rate of a water pump are extracted to form a feature vector set. This processing provides an accurate foundation for subsequent modeling, avoiding the impact of data clutter on accuracy. In actual operation, for a large park's water supply network, if the operational status information is not categorized and organized, it may lead to maintenance delays. However, through this extraction and organization, problematic equipment can be quickly located, improving overall operational efficiency.

[0097] Step S220: Based on the core feature set, a mapping model between physical entities and virtual entities is constructed using modeling techniques. At the same time, an association mapping tool is used to match the physical features in the mapping model with the virtual template. If the matching degree is lower than a preset threshold, it is corrected by an adjustment tool to obtain a preliminary mapping relationship.

[0098] The mapping model between the constructed physical entities and virtual entities is derived using the following formula:

[0099] (10)

[0100] In formula (10), Represents physical entities With virtual entities The mapping model between them This represents the total number of core features. Indicates the first The weight coefficients of each feature, Indicates the first physical characteristics Corresponding virtual features The similarity function between them This indicates the search for the mapping relationship that optimizes the overall mapping effect. .

[0101] The following formula is used to measure the similarity between the physical features in the mapping model and the two vectors of the virtual template:

[0102] (11)

[0103] In formula (11), This represents the score indicating the degree of matching between physical features and the virtual template. Indicates the number of feature dimensions involved in the matching. Indicates the first A physical feature vector, Indicates the first A virtual template feature vector, This represents the function for calculating cosine similarity.

[0104] The feature vector, after being corrected by the adjustment tool, is obtained using the following formula:

[0105] (12)

[0106] In formula (12), This represents the feature vector after adjustment using the tool. Represents the original physical feature vector. This indicates that the strength parameters are being adjusted. Represents the target virtual template vector. This indicates the preset matching threshold. Indicates the current matching degree. This represents a step function that activates a correction mechanism when the matching degree falls below a threshold.

[0107] Based on the core feature set, a mapping model between physical and virtual entities is constructed using modeling techniques. These modeling techniques can be machine learning-based regression models, such as using linear regression algorithms to associate physical features with virtual representations. Specifically, the core features of a physical entity, such as an air conditioning unit, include temperature readings and power consumption, while the virtual entity is its digital twin model. The construction process first collects historical data to train the model, calculates the weights between features (e.g., the coefficient of power increase for every 1 degree Celsius increase in temperature is solved using the least squares method), and then generates a mapping function to project the physical data into the virtual space in real time.

[0108] Simultaneously, association mapping tools, such as graph-based matching systems, are employed to match physical features in the mapping model with virtual templates. The virtual templates predefine standard feature ranges, such as the ideal power curve of an air conditioner. The matching process calculates the Euclidean distance as a matching degree index. If the distance is lower than a preset threshold, such as 0.8, it is corrected using tools like gradient descent optimizers. The model parameters are iteratively updated until the matching degree reaches the target, resulting in a preliminary mapping relationship. This method ensures a close correspondence between physical and virtual entities. In park air conditioning management, it can simulate fault scenarios, support predictive maintenance, and its technical effect is to reduce actual downtime and improve system reliability.

[0109] Step S230: Obtain the dynamic synchronization parameters in the initial mapping relationship through a real-time monitoring tool.

[0110] The following formula is used to calculate the dynamic synchronization parameters of the weighted average:

[0111] (13)

[0112] In formula (13), Indicates time Dynamic synchronization parameters, Indicates the total number of monitoring nodes. Indicates the first The weight coefficient of each monitoring node, Indicates the first Each node at time... The time deviation.

[0113] Obtaining dynamic synchronization parameters from the initial mapping relationship through real-time monitoring tools can be achieved using event-driven monitoring frameworks such as Prometheus. Specifically, the real-time monitoring tool polls the mapping relationship every minute, extracting parameters such as synchronization delay time and data update frequency. For example, in the virtual model, it tracks the deviation between the real-time temperature changes of the physical air conditioner and the virtual simulation, ensuring that the parameters remain within 5 milliseconds. Parameters obtained in this way help maintain the dynamic consistency of the mapping.

[0114] Furthermore, the digital twin-based smart park intelligent management and control method provided in this embodiment includes step S300 as follows:

[0115] Step S310: Obtain real-time data of dynamic synchronization parameters from the mapping model, classify and organize the dynamic synchronization parameters using a data acquisition tool, record parameters that exceed the preset threshold, and determine the abnormal parameter set.

[0116] The set of abnormal parameters exceeding the preset threshold range was determined using a formula:

[0117] (14)

[0118] In formula (14), Represents a set of abnormal parameters. Indicates the first The measured values ​​of each dynamic synchronization parameter, This indicates the preset upper limit threshold. This indicates the preset lower threshold value. This indicates the number of all parameters.

[0119] Real-time data acquisition of dynamic synchronization parameters from the mapping model can be achieved through an integrated data streaming platform such as Apache Kafka. Specifically, this data streaming platform allows real-time subscription to parameter updates within the model. For example, in a smart park, dynamic synchronization parameters include the temperature synchronization delay and power matching degree between the virtual air conditioning model and the physical air conditioning system. These dynamic synchronization parameters are collected once per second via Kafka's topic channels, ensuring smooth data transmission. Next, data acquisition tools are used to categorize and organize these dynamic synchronization parameters. For instance, tools built using the Python-based Pandas library can group dynamic synchronization parameters by type, such as categorizing delay parameters into time-based categories and matching degrees into numerical categories. These parameters are then processed into structured tables through script filtering and aggregation operations.

[0120] Parameters exceeding a preset threshold are recorded. For example, if the delay threshold is set to 10 milliseconds, if a parameter exceeds this threshold, it will be automatically logged to the database. This will eventually determine the set of abnormal parameters. This set may contain delay anomalies from multiple air conditioning units, forming an anomaly list that provides a basis for subsequent analysis.

[0121] Step S320: Based on the set of abnormal parameters, use a data comparison tool to match the abnormal parameters with the normal range in the historical records. If the deviation between the abnormal parameters and the normal range in the historical records exceeds a preset threshold, use a deviation analysis tool to locate the source of the deviation and obtain the deviation source classification.

[0122] The classification of deviation sources is derived using the following formula:

[0123] (15)

[0124] In formula (15), Indicates the first The sources of deviation for each outlier parameter are the classification results. Index indicating possible sources of deviation. Indicates the number of features used for classification. Indicates the first The weights of each feature Indicates the first Sources of class bias in the first Indicator values ​​on each feature.

[0125] Based on the set of abnormal parameters, a data comparison tool is used to match the abnormal parameters with the normal range in historical records. Specifically, the data comparison tool can be a SQL-based query engine, such as a comparison function in PostgreSQL. First, it extracts the normal range from the historical database. For example, the normal deviation range for air conditioner temperature synchronization is 0.5 degrees to 1 degree. Then, it calculates the difference between the current abnormal parameter, such as a deviation of 2 degrees, and this range. If the deviation exceeds a preset threshold, such as 1.5 degrees, the next step is triggered.

[0126] By using deviation analysis tools to locate the source of deviation and classify the source of deviation, such as using root cause analysis algorithms like the fishbone diagram method, the deviation analysis tool will trace the data link and analyze whether it is caused by sensor failure or network latency, classifying it as a hardware or software deviation source. This process can clearly identify the root cause of the problem.

[0127] Step S330: Based on the classification of deviation sources, use a status monitoring tool to obtain the status data of the equipment during operation, compare it with the pre-established normal status template, determine whether it belongs to the abnormal operation of the equipment, and obtain the abnormality analysis results.

[0128] The anomaly analysis results are obtained using the following formula:

[0129] (16)

[0130] In formula (16), Indicates the first Anomaly detection results for each device Indicates the first The first device The current values ​​of each running parameter, Indicates the first Normal operating baseline values ​​for each parameter, Indicates the first Anomaly detection thresholds for each parameter, This indicates the total number of running parameters.

[0131] To categorize the sources of deviations, status monitoring tools are used to acquire operational status data of the equipment. This data is then compared with a pre-established normal status template to determine whether the equipment is malfunctioning, resulting in an anomaly analysis. Specifically, status monitoring tools, such as those based on the Zabbix open-source framework, pull real-time operational status data from the equipment. For example, the operational status of an air conditioner includes compressor speed and refrigerant pressure. This operational status data is collected via an API interface and compared with a template. The template predefines a normal speed range, such as 800 to 1200 revolutions per minute. If the comparison shows an abnormally low speed, it is determined that the equipment is malfunctioning, and the result is output as an anomaly report, including a detailed description of the source, such as a hardware fault. In park management, this analysis helps to respond quickly, prevent small problems from escalating into major failures, and improve maintenance efficiency.

[0132] Furthermore, the digital twin-based smart park intelligent management and control method provided in this embodiment includes step S400 as follows:

[0133] Step S410: Extract key deviation data from the anomaly analysis results, classify the key deviation data using data processing tools, identify at least one source of deviation, and obtain the classified deviation dataset.

[0134] The following formula is used to extract and quantify key deviation data from anomaly analysis results:

[0135] (17)

[0136] In formula (17), Indicates the first Weighted deviation values ​​of key deviation data This indicates the total number of data points in the anomaly analysis results. Indicates the first The weighting coefficients for each data point Indicates the first One actual measured value, Indicates the first One reference standard value.

[0137] The following formula is used for automatic identification of at least one source of deviation:

[0138] (18)

[0139] In formula (18), Indicates the first The results of identifying the sources of deviation, F l Indicates the first Feature vectors from various sources Indicates the first The identification threshold for each source is set. When the maximum value of the feature vector exceeds the set threshold, it is marked as 1 to indicate that a source of deviation has been detected; otherwise, it is marked as 0.

[0140] Key deviation data is extracted from anomaly analysis results using a dedicated data extraction engine. For example, in a smart park's intelligent management system, the anomaly analysis results might include temperature deviation records from air conditioning equipment. The extraction engine scans the result files, filtering out key data with deviations exceeding 2 degrees Celsius, such as synchronization delay data for specific air conditioning units, thus forming a preliminary dataset. Specifically, this extraction process involves parsing the structured format of the results and using Elasticsearch-like query tools to filter high-impact deviations, ensuring that the extracted data focuses on potential risk points and provides an accurate foundation for subsequent classification.

[0141] When classifying key deviation data using data processing tools, machine learning classification algorithms, such as those based on the Scikit-learn library, are used to group the deviation data according to their source type. For example, deviations caused by network latency are classified into the communication category, while deviations caused by sensor failures are classified into the hardware category. At least one source of deviation is identified, for example, labeled as "hardware-sensor failure". The final result is a classified deviation dataset, which is organized in tabular form for easy further analysis.

[0142] Step S420: Based on the classified deviation dataset, obtain the pre-established decision rule base and compare it with the rule matching tool. If the matching degree between the classified deviation dataset and the pre-established decision rule base exceeds the preset threshold, then determine the matching decision rule set.

[0143] The matching degree between the classified bias dataset and the pre-established decision rule base is calculated using the following formula:

[0144] (19)

[0145] In formula (19), This indicates the overall matching degree between the post-classification bias dataset and the decision rule base. This represents the biased dataset after classification. Represents the first in the dataset One biased sample, Indicates sample eigenvectors, This represents a pre-established base of decision rules. Indicates the first rule in the rule base Rule 1 This represents the size of the intersection between sample features and rules. This represents the size of the union of sample features and rules.

[0146] Based on the categorized deviation dataset, a pre-established decision rule base is obtained. The process of comparing these rules using a rule matching tool can be understood as an operation of a rule engine. For example, the decision rule base stores preset rules for the maintenance of park equipment, such as "if the hardware deviation exceeds the threshold, the power supply should be checked first." Rule matching tools such as the Drools engine calculate the matching degree. Assuming the preset threshold is 80%, if the air conditioner deviation in the dataset has a matching degree of 85%, then the matching decision rule set is determined, including maintenance priority rules, which helps in efficient decision-making.

[0147] Step S430: For the matching decision rule set, use an instruction generation tool to convert the rules into an executable instruction sequence, obtain the priority and dependency relationship, determine the execution order, and obtain an ordered execution instruction sequence.

[0148] The executable instruction sequence after rule transformation is obtained through the following formula:

[0149] (20)

[0150] In formula (20), Indicates the first The executable instructions converted from each rule This indicates that the instruction generates a conversion function. Indicates the first A matching decision rule, Indicates the first The priority weight of each rule. Represents the dependency coefficient. Indicates the first The weighting factor of each rule attribute, Indicates the first The rule in the first The rule value on each attribute, This indicates the total number of rule attributes. Indicates the first The rule in the first The reciprocal of the attribute weight factor for each attribute. Indicates the first The rule in the first The reciprocal of the dependency coefficient on each attribute.

[0151] An ordered sequence of execution instructions is derived using the following formula:

[0152] (twenty one)

[0153] In formula (21), This represents the final ordered set of executed instructions. Indicates the first position in the sequence Instructions for each position, Indicates the total length of the instruction sequence. Represents the sorting function. Indicates the number to be sorted One instruction, Indicates the first The sorting key parameter corresponding to each instruction = (1, 2, 3, ..., q).

[0154] When using an instruction generation tool to convert a set of matching decision rules into a sequence of executable instructions, it is necessary to obtain the priority and dependency relationships. For example, the instruction generation tool will parse the rule set and generate an executable instruction sequence such as "check the sensor first, then restart the device". The execution order is determined based on the dependency graph to ensure that high-priority instructions are executed first, resulting in an ordered sequence of execution instructions. In a park lighting system, this can avoid energy waste caused by blind operation.

[0155] Step S440: Distribute the ordered sequence of execution instructions to the perception modeling module using a data transmission tool, and update the closed-loop link status synchronously during distribution to determine the instruction execution feedback.

[0156] The closed-loop link state update value is obtained using the following formula:

[0157] (twenty two)

[0158] In formula (22), Indicates the first The closed-loop link state update value of the next iteration. Represents the state decay factor. Indicates the first The expected output of this instruction execution. Indicates the first The actual feedback from the execution of this instruction. This represents a small constant used to prevent division by zero. This indicates the number of iterations in the previous state update.

[0159] By executing an ordered sequence of instructions and distributing them to the perception modeling module using data transmission tools, the MQTT (Message Queuing Telemetry Transport) protocol is employed to ensure that instructions are sent to the air conditioning's perception module in real time. Simultaneously, the closed-loop link status is updated during distribution, for example, updating the status to "instruction execution." A feedback mechanism confirms the instruction execution feedback, such as confirming that the air conditioning temperature has returned to normal. This forms a complete response loop, improving the stability of the park's equipment. When extending to security facilities, camera response deviation data extracted from abnormal results, after classification and rule matching, generates an ordered sequence of execution instructions that can be distributed to the modeling module to update the link status, ensuring the continuity of security monitoring.

[0160] Furthermore, the intelligent management and control method for smart parks based on digital twins provided in this embodiment includes step S500:

[0161] Step S510: When distributing the execution instruction sequence, synchronously update the status information in the data synchronization tool, determine the priority of the execution instruction sequence, and obtain an ordered instruction execution queue.

[0162] The priority of an instruction sequence is calculated using the following formula, which comprehensively considers factors such as execution time, weight, and resource requirements:

[0163] (twenty three)

[0164] In formula (23), Indicates the first The priority value of an instruction sequence. Indicates the first The estimated execution time of a sequence of instructions. Indicates the first The weighting coefficients of each instruction sequence. Indicates the first Resource requirements for a sequence of instructions. Indicates the first The deadline for each instruction sequence. , , The weight parameters of each factor are represented. The control logic of formula (23) adopts a "multi-dimensional (time, business importance, resource-time constraints) priority dynamic calculation mechanism", which comprehensively considers the balance between the execution time of the instruction, business weight, resource requirements and deadline. The influence of each factor is flexibly adjusted through configurable weight parameters, and finally outputs the priority of the instruction sequence, providing a quantitative basis for subsequent instruction scheduling (such as execution order and resource allocation).

[0165] The following formula describes the synchronous update of state information during the distribution of execution instruction sequences:

[0166] (twenty four)

[0167] In formula (24), This indicates the status information of the updated data synchronization tool. This indicates the current state information. Indicates the first An identifier for a sequence of instructions. Indicates the first Timestamps of a sequence of instructions Indicates the first The execution result of a sequence of instructions. Represents a hash function. This represents the XOR operation. This indicates a data connection operation. The control logic of formula (24) ensures that the state of the data synchronization tool can accurately integrate the identifier, time and execution result of the instruction after each instruction sequence is distributed, thereby realizing real-time and reliable iteration of the state and providing accurate state basis for subsequent instruction distribution and system management.

[0168] The process of generating an ordered instruction execution queue based on priority determination results is described by the following formula:

[0169] (25)

[0170] In formula (25), Represents an ordered instruction execution queue. arrive Indicates unsorted A sequence of instructions, Represents the sorting function. Comparison functions for sorting, Indicates instruction sequence The priority calculation function. The control logic of formula (25) is "priority-based instruction sequence sorting mechanism". By defining the priority calculation rules of the instruction sequence, multiple instruction sequences to be executed are sorted to obtain an execution queue ordered by priority, thereby determining the execution order of the instruction sequence and realizing priority control of the instruction execution process.

[0171] During the distribution of the execution instruction sequence, the status information in the data synchronization tool needs to be updated synchronously. This data synchronization tool can be understood as a dedicated software component used to ensure data consistency between multiple systems. For example, in a smart park energy management system, it's similar to a message queue system like Apache Kafka, capable of capturing and propagating status changes in real time. Specifically, when distributing the execution instruction sequence, the system first checks each instruction in the sequence, such as the instruction to adjust the brightness of the park's lighting equipment. Then, it updates the status to "distributing" through the data synchronization tool. This involves a simple state machine mechanism, where the state transitions from "pending" to "execution." Log recording ensures synchronization across all nodes, preventing operational errors caused by data inconsistency.

[0172] When determining the priority of the execution instruction sequence, a weight-based sorting algorithm is used. First, the urgency of each instruction is evaluated. For example, the instruction on abnormal water pressure in the park's water supply system will be given high priority because it directly affects the safety of users' water use. The system will calculate a priority score, such as the urgency multiplied by the scope of impact, and sort the results to form an ordered queue. This ensures that critical instructions such as repairing water pump failures are executed first, rather than secondary instructions such as adjusting landscape lights.

[0173] Step S520: Drive the automated operation and maintenance module through an ordered instruction execution queue, and adjust the mapped area of ​​the virtual park in real time through the automated operation and maintenance module. Record status feedback data during the adjustment to determine the optimized digital management and control status.

[0174] The optimized digital management and control status is derived using the following formula:

[0175] (26)

[0176] In formula (26), This indicates an optimized digital management and control status. Represents the set of digital intelligent management and control states. Indicates the total number of mapped regions. Indicates the first Weighting coefficients for each region Indicates the first The state evaluation function for each region Indicates the first Real-time adjustment parameters for each region Indicates the first Automated operation and maintenance of each region Indicates the first The timestamps of each region. The control logic of formula (26) is to achieve multi-regional and multi-factor collaborative optimization of the digital intelligent management system through the logic of "regional multi-dimensional evaluation → global weighted aggregation → optimal state selection", so as to ensure that the system achieves the global optimal management effect in the dimensions of resource adjustment, operation and maintenance.

[0177] The status feedback data is obtained using the following formula:

[0178] (27)

[0179] In formula (27), Indicates time The state feedback data function, , , Indicates the weighting coefficients for different feedback types. This indicates the change in the virtual park's mapped area. Indicates the amount of change in performance indicators. This represents the change in load status. The control logic of formula (27) integrates the changes in three core states: "virtual campus mapping, performance, and load," and assigns weights according to business needs, ultimately outputting a comprehensive feedback signal to drive the real-time control adjustment of the system.

[0180] An automated operations and maintenance (O&M) module, driven by an ordered instruction execution queue, is an integrated platform responsible for automating maintenance tasks for park equipment. For example, it can connect to a tool framework like Ansible. The ordered instruction execution queue pushes instructions one by one, driving the module to launch automated scripts and make real-time adjustments to the mapped area of ​​the virtual park. This mapped area is a digital twin model simulating the layout of the real park, such as an energy distribution map. During adjustments, the O&M module modifies model parameters according to queue instructions, such as optimizing the flow settings of water supply pipes. Status feedback data is recorded during the adjustment process, involving a real-time monitoring mechanism. The system collects adjusted data, such as water pressure returning to the normal value of 85 kPa, and stores it in a feedback database. This determines the optimized intelligent management status, a comprehensive indicator that combines AI (Artificial Intelligence) analysis of feedback data to evaluate the overall optimization effect. For example, if feedback shows a 15% reduction in energy consumption, the status is updated to "Optimization Complete," which helps park managers respond quickly to subsequent needs.

[0181] When extended to the park's waste management system, after the execution of the instruction sequence to distribute and update the status information, the priority is determined. For example, the instruction to handle the garbage compressor failure is ranked first. A queue is formed to drive the operation and maintenance module to adjust the waste flow model of the virtual mapping area, record feedback such as improved compression efficiency, determine the optimization of the control status, and ensure the continuity of environmental sanitation.

[0182] Please see Figure 2 This embodiment provides a smart park digital management and control system based on digital twins, used to execute the aforementioned smart park digital management and control method based on digital twins. It includes a multi-scenario fusion data acquisition module 10, a dynamic synchronization parameter determination module 20, an anomaly analysis result acquisition module 30, an execution instruction sequence generation module 40, and a digital management and control status acquisition module 50. The multi-scenario fusion data acquisition module 10 collects raw sensing data of the operating status of building pipeline equipment within the park from a sensor network, and performs real-time aggregation processing on the sensing data through a data processing system to obtain aggregated multi-scenario fusion data. The dynamic synchronization parameter determination module 20 uses modeling technology to construct physical entities and virtual entities based on the aggregated multi-scenario fusion data. The system includes a mapping model for quasi-entities, which determines the dynamic synchronization parameters contained in the mapping model; an anomaly analysis result acquisition module 30, which analyzes the parameter deviation using a machine learning algorithm if the dynamic synchronization parameters in the mapping model exceed the threshold, determines that the source of the parameter deviation belongs to equipment operation anomaly, and obtains the anomaly analysis result; an execution instruction sequence generation module 40, which obtains the corresponding decision rules from the intelligent decision system based on the anomaly analysis result and generates an execution instruction sequence for the anomaly, wherein the execution instruction sequence is used to form a closed-loop link from perception modeling to analysis and decision-making; and a digital intelligent management and control status acquisition module 50, which drives the automated operation and maintenance module through the execution instruction sequence to adjust the mapping model in the virtual park in real time and obtain an optimized digital intelligent management and control status.

[0183] The digital twin-based smart park intelligent management and control method and system provided in this embodiment, compared with existing technologies, aggregates and processes raw sensing data collected by sensor networks in real time to form multi-scenario fusion data, and uses modeling technology to construct a mapping model between physical and virtual entities, dynamically synchronizing parameters to monitor equipment status. When parameters exceed thresholds, machine learning algorithms are used to analyze the source of deviations, identify anomalies and generate analysis results. Rules are then obtained from the intelligent decision-making system to generate execution instruction sequences, driving the automated operation and maintenance module to adjust the virtual park model in real time, ultimately achieving optimized management and control. The core innovation of this embodiment lies in constructing a closed-loop link from perception to decision-making to execution, significantly improving the intelligence level and response efficiency of park equipment management, and ensuring operational stability.

[0184] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A smart park digital management and control method based on digital twins, characterized in that, Includes the following steps: S100. Collect raw sensing data on the operating status of building pipeline equipment in the park from the sensor network, and perform real-time aggregation processing on the sensing data through the data processing system to obtain aggregated multi-scene fusion data. S200. For the aggregated multi-scenario fusion data, a modeling technique is used to construct a mapping model between physical entities and virtual entities, and the dynamic synchronization parameters contained in the mapping model are determined. S300. If the dynamic synchronization parameter in the mapping model exceeds the threshold, the parameter deviation is analyzed by machine learning algorithm to determine that the source of the parameter deviation belongs to the abnormal operation of the equipment, and the abnormality analysis result is obtained. S400. Based on the anomaly analysis results, obtain the corresponding decision rules from the intelligent decision system and generate an execution instruction sequence for the anomaly, wherein the execution instruction sequence is used to form a closed-loop link from perception modeling to analysis and decision-making. S500: Drive the automated operation and maintenance module through the execution instruction sequence to adjust the mapping model in the virtual park in real time to obtain an optimized digital and intelligent management and control status. Step S100 includes: S110. Obtain raw sensing data from the sensor network of building pipeline equipment in the park, perform preliminary cleaning processing on the raw sensing data, remove noise and outliers through data filtering tools, and obtain the cleaned basic data stream. S120. Based on the cleaned basic data stream, a real-time aggregation tool is used to perform multi-dimensional integration processing on the basic data stream. If the total amount of the basic data stream exceeds a preset threshold, load distribution is performed through sharding to determine the aggregated intermediate dataset. The following formula is used to define the activation conditions of the fragmentation mechanism: in, Indicates the first Data processing volume per partition Indicates the total amount of the basic data stream. Indicates the number of fragments. This indicates a preset threshold. When the total amount of basic data stream exceeds the preset threshold, the sharding mechanism is activated to distribute the total amount of data evenly among the shards for load balancing. Otherwise, no sharding is performed. The following formula describes the computational process of generating the intermediate dataset through multiple rounds of aggregation: in, This represents the final, determined intermediate dataset. Indicates the number of rounds in the aggregation process. Indicates the first The fusion coefficient of the wheel aggregation, Indicates the first The number of data blocks participating in aggregation in each round. Represents aggregate functions, Indicates the first Wheel of Life One data block; S130. Using a data fusion tool, the multi-scene perception information in the intermediate dataset is associated and mapped to obtain a fused comprehensive data view. It is then determined whether the comprehensive data view meets the multi-scene consistency requirements to obtain aggregated multi-scene fused data. The integrated data view is derived using the following formula: in, This represents the integrated data view after merging. This indicates the total number of scenarios. Indicates the first The weighting coefficients for each scenario. Indicates the first The association mapping matrix of each scene, Indicates the first A dataset of perception information for each scene; The following formula is used to determine whether a comprehensive data view meets the consistency requirements across multiple scenarios: in, This represents a consistency evaluation metric across multiple scenarios. This indicates the number of feature dimensions in the comprehensive data view. Indicates the first A numerical vector with 1 feature dimension. This represents the mean vector of all scene features. The standard deviation threshold representing the consistency judgment; The aggregated multi-scenario fusion data is obtained through the following formula: in, This represents the aggregated multi-scenario fusion data. This represents the set of options that require aggregation for decision-making. Indicates the number of data samples. The number of dimensions representing data features. Indicates the first The first sample The correlation coefficient of each feature This represents the weight of the perceived information at the corresponding location. Indicates the integration of quality assessment factors; Step S200 includes: S210. Obtain the operational status information of physical entities from the aggregated multi-scene fusion data, and use data extraction tools to classify and organize the operational status information to determine the core feature set of physical entities. S220. Based on the core feature set, a mapping model between physical entities and virtual entities is constructed using modeling technology. At the same time, an association mapping tool is used to match the physical features in the mapping model with the virtual template. If the matching degree is lower than a preset threshold, it is corrected by an adjustment tool to obtain a preliminary mapping relationship. S230. Obtain the dynamic synchronization parameters in the initial mapping relationship through a real-time monitoring tool: Step S300 includes: S310. Obtain real-time data of dynamic synchronization parameters from the mapping model, classify and organize the dynamic synchronization parameters using a data acquisition tool, record parameters that exceed a preset threshold, and determine the abnormal parameter set. S320. Based on the set of abnormal parameters, a data comparison tool is used to match the abnormal parameters with the normal range in the historical records. If the deviation between the abnormal parameters and the normal range in the historical records exceeds a preset threshold, the source of the deviation is located by a deviation analysis tool, and the deviation source classification is obtained. S330. Based on the classification of the sources of deviation, use a status monitoring tool to obtain the status data of the equipment during operation, compare it with the pre-established normal status template, determine whether it belongs to the abnormal operation of the equipment, and obtain the abnormal analysis result.

2. The intelligent park digital management and control method based on digital twins according to claim 1, characterized in that, Step S400 includes: S410. Extract key deviation data from the anomaly analysis results, classify the key deviation data using a data processing tool, identify at least one source of deviation, and obtain a classified deviation dataset. S420. Based on the classified bias dataset, obtain the pre-established decision rule base and compare it with the rule matching tool. If the matching degree between the classified bias dataset and the pre-established decision rule base exceeds the preset threshold, then determine the matching decision rule set. S430. For the matching decision rule set, use an instruction generation tool to convert the rules into an executable instruction sequence, obtain the priority and dependency relationship, determine the execution order, and obtain an ordered execution instruction sequence. S440. The ordered sequence of execution instructions is distributed to the perception modeling module using a data transmission tool. During distribution, the closed-loop link status is updated synchronously to determine the instruction execution feedback.

3. The intelligent park digital management and control method based on digital twins according to claim 1, characterized in that, Step S500 includes: S510. When distributing the execution instruction sequence, the status information in the data synchronization tool is updated synchronously to determine the priority of the execution instruction sequence and obtain an ordered instruction execution queue. S520. The automated operation and maintenance module is driven by an ordered instruction execution queue. The automated operation and maintenance module adjusts the mapping area of ​​the virtual park in real time, records status feedback data during the adjustment, and determines the optimized digital management and control status.

4. The intelligent park digital management and control method based on digital twins according to claim 3, characterized in that, In step S510, the priority of the instruction sequence is calculated using the following formula by comprehensively considering factors such as execution time, weight, and resource requirements: in, Indicates the first The priority value of an instruction sequence. Indicates the first The estimated execution time of a sequence of instructions. Indicates the first The weighting coefficients of each instruction sequence. Indicates the first Resource requirements for a sequence of instructions. Indicates the first The deadline for each instruction sequence. , , This represents the weight parameters of each factor.

5. The intelligent park digital management and control method based on digital twins according to claim 4, characterized in that, In step S510, the following formula is used to describe the synchronous update of state information during the distribution of the execution instruction sequence: in, This indicates the status information of the updated data synchronization tool. This indicates the current state information. Indicates the first Identifier of a sequence of instructions Indicates the first Timestamps of a sequence of instructions Indicates the first The execution result of a sequence of instructions, Represents a hash function. This represents the XOR operation. Indicates a data connection operation; The process of generating an ordered instruction execution queue based on priority determination results is described by the following formula: in, Represents an ordered instruction execution queue. arrive Indicates unsorted A sequence of instructions, Represents the sorting function. Comparison functions representing sorting, Indicates instruction sequence The priority calculation function.

6. The intelligent park digital management and control method based on digital twins according to claim 5, characterized in that, In step S520, the optimized digital management and control status is obtained through the following formula: in, This indicates an optimized digital management and control status. Represents the set of digital intelligent management and control states. Indicates the total number of mapped regions. Indicates the first Weighting coefficients for each region Indicates the first The state evaluation function for each region Indicates the first Real-time adjustment parameters for each region Indicates the first Automated operation and maintenance of each region Indicates the first Timestamps for each region; Status feedback data is obtained using the following formula: in, Indicates time The state feedback data function, , , Indicates the weighting coefficients for different feedback types. This indicates the change in the virtual park's mapped area. Indicates the amount of change in performance indicators. This indicates the amount of change in load status.

7. A digital twin-based intelligent park management and control system, used to execute the digital twin-based intelligent park management and control method as described in any one of claims 1 to 6, characterized in that, include: The multi-scene fusion data acquisition module (10) is used to collect raw sensing data of the operating status of building pipeline equipment in the park from the sensor network, and to perform real-time aggregation processing on the sensing data through the data processing system to obtain aggregated multi-scene fusion data. The dynamic synchronization parameter determination module (20) is used to construct a mapping model between physical entities and virtual entities using modeling technology for the aggregated multi-scenario fusion data, and to determine the dynamic synchronization parameters contained in the mapping model. The anomaly analysis result acquisition module (30) is used to analyze the parameter deviation through a machine learning algorithm if the dynamic synchronization parameter in the mapping model exceeds the threshold, determine that the source of the parameter deviation belongs to the abnormal operation of the equipment, and obtain the anomaly analysis result. The execution instruction sequence generation module (40) is used to obtain the corresponding decision rules from the intelligent decision system based on the anomaly analysis results and generate an execution instruction sequence for the anomaly, wherein the execution instruction sequence is used to form a closed-loop link from perception modeling to analysis decision; The digital management status acquisition module (50) is used to drive the automated operation and maintenance module through the execution instruction sequence to adjust the mapping model in the virtual park in real time and obtain the optimized digital management status.

Images