Edge cloud cooperation-based device dynamic control method and device

Abstract

The application discloses a kind of based on edge cloud cooperation's equipment dynamic control method and device, comprising: through field sensor network acquisition operating state data and environmental parameter data, and acquisition data is sent to remote computing node, by remote computing node generates preliminary global optimization control strategy and is issued to local control terminal, local control terminal analysis generates control instruction set and forms instruction execution sequence, equipment response state is monitored in the process of executing control instruction and is triggered fallback operation when abnormal.In equipment operation period, the parameter that characterizes operating state is continuously monitored, and when meeting abnormal condition, equipment switching or operation mode adjustment is executed.The application generates optimization control strategy by edge cloud cooperation and realizes automatic execution locally, improves multi-device scheduling efficiency, enhances system operation safety through abnormal identification and fallback protection, maintains energy saving and stability in running process through continuous monitoring and dynamic adjustment, realizes efficiency improvement and risk reduction.

Description

Technical Field

[0001] This invention relates to the field of energy management and automatic control technology, and in particular to a method and apparatus for dynamic control of equipment based on edge-cloud collaboration. Background Technology

[0002] Existing cooling systems typically consist of multiple types of equipment, including chillers, chilled water pumps, cooling water pumps, and cooling towers. The operation and scheduling of these devices are relatively independent, lacking a coordination mechanism based on the overall load status. Under partial load conditions, the equipment cannot form a unified adjustment logic, resulting in low overall operating efficiency. This is especially true when the ratio of chillers to pumps is unreasonable, making uneven energy distribution more likely.

[0003] In traditional operating modes, equipment start-up and shutdown, parameter adjustment, and load switching mainly rely on manual operation. Due to the long decision-making cycle and limited response speed of manual operations, it is difficult to adapt to real-time changes in equipment load and environmental parameters, causing the system's operating state to easily deviate from the reasonable range and resulting in additional energy consumption. Furthermore, manual methods rely on experience-based judgment, which can easily lead to decreased efficiency or even equipment risks when operating conditions are complex or frequently changing due to untimely operation or judgment errors.

[0004] Furthermore, in the current technological system, cooling and refrigeration circuits are typically operated as relatively independent links, lacking a linkage mechanism for overall energy efficiency. Due to the lack of a global coordination strategy across devices, the system struggles to maintain a coordinated and consistent operating state under dynamic load conditions, thus making it difficult to guarantee overall operating efficiency.

[0005] In terms of equipment stability, traditional cooling systems have limited real-time monitoring capabilities for equipment operation, lacking continuous monitoring of key operating parameters and intelligent anomaly identification. When equipment exhibits abnormal trends or deviates from its normal operating range, the system struggles to detect and take timely countermeasures, resulting in delayed fault response, impacting equipment reliability, and potentially leading to further increases in energy consumption. Summary of the Invention

[0006] The main objective of this invention is to provide a device for dynamic control of equipment based on edge-cloud collaboration, aiming to solve the technical problem of low operating efficiency, high energy consumption and delayed abnormal response of equipment groups due to the lack of a collaborative generation of optimized control strategies based on real-time field data and remote calculation results.

[0007] To achieve the above objectives, this invention provides a device dynamic control method based on edge-cloud collaboration, applied to a local control terminal, comprising: The sensor network deployed at the equipment site periodically collects real-time operating status data and environmental parameter data of the equipment group; The real-time operating status data and the environmental parameter data are sent to a remote computing node; Receive the preliminary global optimization control strategy generated by the remote computing node based on the real-time operating status data and the environmental parameter data; The global optimization control strategy is parsed to generate control instruction sets corresponding to each device, and the control instruction sets are sorted according to the preset device control logic to form an instruction execution sequence; The system executes each control instruction according to the instruction execution sequence, and monitors the device response status during the execution process. When the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. During the operation of the equipment group, parameters characterizing the operating status of the equipment are continuously monitored. When the parameters meet preset abnormal conditions, equipment switching or operating mode adjustment operations are initiated.

[0008] This invention provides a device dynamic control method based on edge-cloud collaboration, applied to remote computing nodes, comprising: Receive real-time operating status data and environmental parameter data of the equipment group periodically collected by the sensor network deployed at the equipment site from the local control terminal; Based on the real-time operating status data and the environmental parameter data, the predicted value of the system cooling load demand is calculated through a preset load prediction model. Based on the predicted system cooling load demand, a preliminary global optimization control strategy is generated using a multi-objective optimization method. This preliminary global optimization control strategy is then sent to the local control terminal, which parses the strategy to generate control instruction sets corresponding to each device. These control instruction sets are then sorted according to preset device control logic to form an instruction execution sequence. Each control instruction is executed according to the instruction execution sequence, and the device response status is monitored during execution. If the device response status indicates an anomaly, a rollback operation is triggered to return the device to a safe state. During the operation of the device group, parameters characterizing the device operating status are continuously monitored. If these parameters meet preset abnormal conditions, device switching or operating mode adjustment operations are initiated.

[0009] Furthermore, to achieve the above objectives, the present invention provides a device dynamic control apparatus based on edge-cloud collaboration, comprising: The sensor acquisition module is used to periodically collect real-time operating status data and environmental parameter data of the equipment group through a sensor network deployed at the equipment site; The strategy generation module is used to generate a global optimization control strategy in collaboration between the local control terminal and the remote computing node based on the real-time running status data and environmental parameter data. The instruction parsing module is used to parse the global optimization control strategy on the local control terminal to generate control instruction sets corresponding to each device, and sort the control instruction sets according to the preset device control logic to form an instruction execution sequence. The instruction execution module is used to execute each control instruction according to the instruction execution sequence, and monitor the device response status during the execution process. When the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. An anomaly monitoring module is used to continuously monitor parameters characterizing the operating status of the equipment during the operation of the equipment group. When the parameters meet preset anomaly conditions, the module initiates equipment switching or operating mode adjustment operations.

[0010] Beneficial Effects: This invention relates to the field of energy management and automatic control technology, and discloses a device and method for dynamic equipment control based on edge-cloud collaboration. The method includes: periodically collecting operating status data and environmental parameter data through a sensor network deployed on-site, sending the collected data to a remote computing node, whereby the remote computing node generates a preliminary global optimization control strategy and sends it to a local control terminal. The local control terminal parses and generates a control instruction set and forms an instruction execution sequence. During the execution of control instructions, the device response status is monitored, and a rollback operation is triggered in case of anomalies. During the operation of the equipment group, parameters characterizing the device operating status are continuously monitored, and device switching or operating mode adjustment operations are performed when abnormal conditions are met. This invention improves the real-time performance and accuracy of control strategy generation through edge-cloud collaborative computing mechanisms; achieves dynamic linkage and energy efficiency optimization among multiple devices by automatically parsing the global optimization control strategy locally and forming a directly executable instruction sequence; enhances system operating safety and stability through anomaly detection and rollback operations; and maintains high efficiency and reliability throughout the overall operation process by continuously monitoring device status and performing switching or mode adjustments. Attached Figure Description

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a schematic diagram of an application environment for a device dynamic control method based on edge-cloud collaboration in one embodiment of the present invention; Figure 2 This is a flowchart illustrating an embodiment of the device dynamic control method based on edge-cloud collaboration of the present invention applied to a local control terminal; Figure 3 This is a flowchart illustrating an embodiment of the device dynamic control method based on edge-cloud collaboration of the present invention applied to a remote computing node; Figure 4 This is a schematic diagram of the functional modules of a preferred embodiment of the device dynamic control device based on edge-cloud collaboration of the present invention. Detailed Implementation

[0012] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0013] The device dynamic control method based on edge-cloud collaboration provided in this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can control the field-deployed sensor network to periodically collect operational status data and environmental parameter data through the client, and send the collected data to a remote computing node. The remote computing node generates a preliminary global optimization control strategy and sends it to the local control terminal. The local control terminal parses and generates a control instruction set and forms an instruction execution sequence. During the execution of control instructions, it monitors the device response status and triggers a rollback operation when an anomaly occurs. During the operation of the device group, it continuously monitors the parameters characterizing the device's operational status and performs device switching or operating mode adjustment operations when abnormal conditions are met. This invention improves the real-time performance and accuracy of control strategy generation through edge-cloud collaborative computing mechanisms; achieves dynamic linkage and energy efficiency optimization among multiple devices by automatically parsing the global optimization control strategy locally and forming a directly executable instruction sequence; enhances system operational security and stability through anomaly detection and rollback operations; and maintains high efficiency and reliability in the overall operation process by continuously monitoring device status and switching or adjusting modes. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster composed of multiple servers. The present invention will now be described in detail through specific embodiments.

[0014] Please see Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the edge-cloud collaborative device dynamic control method provided by the present invention. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0015] like Figure 2 As shown, the device dynamic control method based on edge-cloud collaboration proposed in this invention, applied to the local control terminal, includes the following steps: S10 periodically collects real-time operating status data and environmental parameter data of the equipment group through a sensor network deployed at the equipment site; In this embodiment, a sensor network deployed at the equipment site is used to collect operational status data and environmental parameter data. The sensor network typically consists of multiple types of sensors, communication modules, and local data nodes. Operational status data originates from measurement points on the equipment surface, such as current, temperature, pressure, and flow rate. These measurement points use electromagnetic induction to measure current, thermistors to measure temperature, pressure sensors to measure pressure, and turbine or ultrasonic methods to measure flow rate, thus quantifying the equipment's operational status. Environmental parameter data describes the external thermal and humidity environment of the equipment and can be collected by temperature and humidity sensors. The resistance changes of the sensitive elements generate readable values. Periodic acquisition relies on a timing unit to generate fixed time intervals. An internal counting mechanism triggers the sensor to perform sampling, generating a timestamped record after each sampling. The collected data is transmitted to the local node via a fieldbus or Ethernet link. Before transmission, the data is encapsulated in a structured format, including sensor identifiers, numerical fields, and checksums to ensure correct transmission and parsing. Periodic acquisition ensures data continuity and comparability, facilitating subsequent status interpretation and trend analysis.

[0016] In implementation, a centralized architecture can be adopted, where a single node broadcasts trigger signals to all sensors and aggregates the returned data; alternatively, a distributed architecture can be used, where multiple acquisition units independently complete sampling and upload data to their local nodes. The sampling period can be set according to the fluctuation level of equipment operation; for example, the period can be shortened during high-load periods to improve the speed of status perception, while the period can be lengthened during low-load periods to reduce communication and processing pressure. To improve data stability, multiple sampling points can be performed during the sampling process and the average value can be taken, or transient interference can be suppressed through amplitude-limiting filtering. The communication method can be selected from RS485, CAN, or industrial Ethernet according to the equipment layout to adapt to different distances and anti-interference requirements. For high humidity, high heat, or vibration environments, sealed sensors or output circuits with noise-resistant designs can be used to improve long-term reliability.

[0017] This embodiment utilizes a data acquisition structure to generate a continuous and reliable state data stream, enabling timely quantification of changes in equipment operation and the external environment. This reduces reliance on human intervention and improves the accuracy of subsequent analysis and control processes. Through periodic triggering and structured encapsulation mechanisms, data integrity and timing consistency are enhanced, providing a stable foundation for subsequent processing.

[0018] S20, the real-time operating status data and the environmental parameter data are sent to the remote computing node; In this embodiment, real-time operating status data and environmental parameter data are centrally managed by the local control terminal after local acquisition. The local control terminal first merges data from different sensor channels according to a predetermined data structure, adding a timestamp, device identifier, acquisition cycle identifier, and acquisition integrity marker to each data record, thus organizing multi-source data within the same strategy cycle into a group of transmittable data units. To facilitate parsing by remote computing nodes, the local control terminal can map operating parameter fields such as current, temperature, pressure, and flow rate, along with environmental parameter fields such as temperature and humidity, to a unified field index table based on a pre-configured data model, generating an information payload with field labels and data type descriptions.

[0019] Based on this, the local control unit constructs data packets according to the communication link type, encapsulating the aforementioned information payload into a packet format that conforms to the requirements of Industrial Ethernet, Fieldbus, or dedicated transmission protocols. The packet may include a header, a data body, and a checksum. The header records the remote computing node address, packet sequence number, data batch number, and priority flag, ensuring that the remote end can reassemble the data sequentially and distinguish between different acquisition cycles. For scenarios with large data volumes, the local control unit can employ fragmented transmission, splitting the data from the same acquisition cycle into multiple segments, and configuring an offset and a total segment count flag for each segment, so that the remote computing node can reconstruct the complete dataset.

[0020] During transmission, the local control unit can maintain a transmission queue, recording the current status of each data unit to be transmitted, including whether it has been transmitted but not acknowledged, is awaiting retry, or has been transmitted successfully. By setting an acknowledgment mechanism at the application or transport layer, the local control unit triggers retransmission logic and re-enqueues the corresponding message if it does not receive acknowledgment from the remote end within a predetermined time window. For long-distance or cross-network transmission scenarios, the local control unit can also compress and encrypt the data before transmission, reducing bandwidth consumption and protecting operational status and environmental parameter data from unauthorized access. In scenarios with short policy cycles or rapidly changing operating conditions, the local control unit can adopt a double-buffered or ring-buffered structure to decouple the acquisition and transmission threads, allowing the acquisition and processing of the next cycle's data to be completed in parallel during the transmission of the previous cycle, thus ensuring that the transmission action can stably keep up with the system control cycle.

[0021] This embodiment, by structuring, encapsulating, and reliably transmitting real-time operating status data and environmental parameter data at the local control terminal, can stably provide complete, continuous, and time-stamped data input to remote computing nodes even under conditions of network jitter, bandwidth limitations, or link interruption risks. This enables remote computing nodes to obtain a data view that is highly consistent with the actual on-site operating conditions, reducing the risk of strategy deviations caused by data loss, out-of-order delivery, or delays, and improving the accuracy of subsequent optimization calculations and the overall stability of the control process.

[0022] S30, Receive the preliminary global optimization control strategy generated by the remote computing node based on the real-time operating status data and the environmental parameter data; In this embodiment, after establishing a communication session with the remote computing node, the local control terminal sets up a dedicated receiving channel and buffer for the initial global optimization control policy. When the local control terminal detects a policy message sent by the remote computing node, it first parses the message header according to a predefined protocol format, reading the policy version number, policy effective time, applicable device range, and policy calculation timestamp. It then uses the message length field and checksum field to determine message integrity and transmission errors. If a checksum failure or missing field is detected, the policy message can be discarded, and a retransmission from the remote terminal can be requested if necessary, to avoid introducing incomplete or corrupted policy data on the control side.

[0023] After the policy message passes integrity verification, the local control terminal writes the message data into the policy receiving buffer and compares it with the currently effective policy based on the policy version number or timestamp. When a newly arrived policy version number is higher than the current version or the calculation timestamp is updated, the local control terminal marks the policy as a candidate for effective policy and parses the policy content according to the internal data structure, separating information such as the target operating settings parameters, constraints, and applicable time periods for each device. If the policy message contains a policy validity period or triggering conditions, the local control terminal can associate them with local time and system status parameters to construct policy effectiveness determination logic, ensuring that the policy is only adopted within the applicable time window or operating condition range.

[0024] When handling multi-policy handover, the local control terminal can maintain a policy version table to record the most recently received preliminary global optimization control policies, including the generation time, source identifier, and applicable device set for each version. When a new policy arrives, if there is a significant mismatch with the current operating status, such as the policy requiring the enabled device to be in a maintenance shutdown state, the local control terminal can temporarily store the policy and mark it as pending verification. Further corrections will be made in subsequent processing steps based on real-time operating status data. For scenarios with high policy reception frequency, the local control terminal can also configure a policy update frequency threshold. When multiple policies are received consecutively within a short period, only the latest one is retained as a candidate policy to reduce the impact of frequent switching on device operational stability.

[0025] This embodiment performs integrity verification, version management, and effectiveness determination on the preliminary global optimization control strategy from the remote computing node at the local control terminal. This avoids erroneous control behaviors caused by transmission errors, version expiration, or content mismatch. It enables the local terminal to have basic filtering and buffering capabilities when adopting the preliminary global optimization control strategy. This reduces the impact of policy changes on the operation of field equipment while maintaining the timeliness of policy updates, thereby improving the reliability and controllability of the overall control process.

[0026] S40, the global optimization control strategy is parsed to generate control instruction sets corresponding to each device, and the control instruction sets are sorted according to the preset device control logic to form an instruction execution sequence; In this embodiment, after the global optimization control strategy is received, the local control terminal performs structured parsing of the strategy content. The device identification information in the strategy is used to determine the physical mapping relationship of the target device, and the underlying communication address of the device is obtained through an internally maintained address mapping table. The operating setting parameters in the strategy are typically in the form of load ratio, valve opening degree, frequency target value, or start / stop status. During the parsing process, the parameter conversion module converts these abstract parameters into an instruction format that the device can directly execute. The parsed instructions are organized by device to form a control instruction set.

[0027] After the control instruction set is generated, it is ordered according to the action dependencies between devices. The local control terminal maintains preset device control logic, which is derived from system debugging experience and device safety boundary rules. For example, valve action must precede water pump action, and water pump action must precede unit action. During the ordering process, the dependencies between instructions are retrieved, and the pre- and post-instructions are determined for each instruction in the ordering results to ensure that the sequence meets the safe operating order. A status check is inserted between adjacent instructions during ordering to confirm whether the device state after the previous instruction meets the conditions for executing the next instruction, thus forming a complete instruction execution sequence.

[0028] The parsing process can be completed through an embedded parsing engine or executed by a software module on the local control terminal. The mapping between the device's unique identifier and the underlying communication address can be stored in a locally built-in address table or obtained in real time from the device management system. Parameter conversion can be performed using a lookup table mapping to map policy parameters to control bytes in a fixed format; alternatively, control parameters can be generated based on the device's rated capacity using a calculation formula. When there are a large number of devices, the parsing module can process policy fragments from different devices in parallel to improve efficiency.

[0029] Control logic sequencing can be implemented using a rule-based decision tree structure or a topological sorting method, transforming device dependencies into a directed graph structure, and then generating the final sequence based on the graph's sorting result. The status check phase can use local sensor data or extract the current status from device feedback to determine whether to execute the next instruction. The instruction execution sequence can be a one-time linear sequence or include conditional jumps to terminate early or enter a rollback process when the device status does not meet expectations.

[0030] This embodiment parses and sorts the strategy, enabling the device to execute control actions while adhering to safety logic and dependency order. This transforms the abstract strategy into a sequence of execution instructions that can be directly applied to field devices, improving the accuracy and controllability of instruction issuance.

[0031] S50, execute each control instruction according to the instruction execution sequence, and monitor the device response status during the execution process. When the device response status indicates an abnormality, trigger a rollback operation to return the device to a safe state. In this embodiment, after the instruction execution sequence arrives at the local control terminal, the control instructions in the sequence are sent to the corresponding devices one by one in a predetermined order. Each control instruction contains data content adapted to the underlying communication protocol of the target device and is written to the device's command cache through the communication interface module. After the control instructions are sent, the execution process is monitored through a timing mechanism. During this monitoring process, the device's response status is continuously read, including the running status word in the device feedback register, the execution completion flag, and changes in actual output parameters. The response status is used to reflect whether the device's execution behavior after receiving the instruction is consistent with the control settings.

[0032] An indication of an anomaly is identified if the response status fails to update, times out, or deviates from the set value. Anomaly detection relies on a preset allowable deviation range and the device's safe operating boundaries; deviations exceeding these boundaries trigger a rollback operation. The rollback operation depends on records of successfully executed instructions, which store the changes each instruction caused to the device state in execution order. The rollback operation reverses these records to generate reverse action instructions for each historical instruction, undoing the already effective state changes and gradually restoring the device to its stable state before the execution sequence began. After completing this recovery process, the device enters a controlled safe state range, preventing wider device risks caused by abnormal execution.

[0033] The execution process can be carried out by sending control commands one by one via synchronous communication, and proceeding to the next command only after the status of each command has been confirmed; alternatively, an asynchronous approach can be used, employing an independent thread to monitor the response status and send abnormal signals back to the control module in real time. The device response status can be obtained by reading device registers or by parsing data frames reported by the device. The deviation judgment range can be dynamically adjusted according to the device type and operating mode to reduce false triggering.

[0034] The rollback operation can restore the device parameters to their initial opening degree, frequency, or start / stop state based on a preset stop command template; alternatively, it can generate a targeted recovery command based on the recorded previous state parameters, making the transition smoother. The rollback execution order can be processed entirely in reverse order, or the reverse order can be locally adjusted according to the dependencies between devices to avoid structural interference during the recovery process.

[0035] This embodiment monitors the execution response of the target device line by line, and uses reverse recovery logic to fall back to a safe state when an anomaly occurs, so that the device operates in a controlled range, reducing the risk of anomaly propagation and improving the reliability and controllability of the system.

[0036] S60, during the operation of the equipment group, continuously monitor the parameters characterizing the operating status of the equipment, and when the parameters meet the preset abnormal conditions, initiate equipment switching or operating mode adjustment operations.

[0037] In this embodiment, the operating status of the equipment group during long-term operation can be continuously monitored through key parameters such as electrical quantities, mechanical quantities, and fluid quantities. Parameters characterizing the operating status of the equipment include operating current, output frequency, and inlet / outlet pressure difference, which directly reflect the equipment load intensity, drive characteristics, and fluid channel resistance changes. During monitoring, parameters are collected through a periodic sampling mechanism and input into the local calculation module. The calculation module performs data cleaning, time alignment, and fluctuation analysis on these parameters to ensure that the operating data used for judgment remains stable and readable.

[0038] To ensure the reliability of anomaly detection, preset anomaly conditions are used during monitoring. These conditions can consist of thresholds for the safety domain, operating domain, and fault domain. Some conditions also include continuous interval judgment logic to avoid false triggering caused by short-term disturbances. When parameters continuously deviate from the allowable operating range or exhibit a trend inconsistent with the normal behavior model of the equipment, an anomaly is determined to meet the preset conditions.

[0039] In the event of an anomaly, equipment switching or operating mode adjustment operations are triggered. Equipment switching replaces equipment that can no longer operate stably, maintaining system output stability by putting redundant equipment into operation. Operating mode adjustment modifies the operating mode of the equipment, such as switching from variable frequency operation to mains frequency operation, to maintain basic output capability when a fault approaches. These actions rely on real-time monitoring results and execute corresponding drive logic through the control module, ensuring that the affected equipment group can maintain safe and continuous service capability even under abnormal conditions.

[0040] The monitoring module can collect operating current, output frequency, and differential voltage at fixed sampling periods and use a sliding window to smooth the parameters to reduce noise interference in the judgment. Anomaly detection can use a fixed threshold or establish an operating envelope based on historical operating data, enabling adaptive judgment. Some devices can also perform anomaly correlation analysis by comparing changes in synchronization parameters of devices in the same group to identify systematic deviations.

[0041] Equipment switching operations can be completed through a combination of stop and start commands, allowing the target equipment to smoothly exit operation while simultaneously bringing redundant equipment into the same load range. Operating mode adjustments can be achieved by changing frequency settings, valve openings, or drive signals, ensuring stable output even when equipment performance is limited. All of the above actions can be executed directly by the local control terminal, or authorized by a higher-level control node before being implemented by the local module.

[0042] This embodiment continuously monitors key operating parameters and performs equipment switching or operating mode adjustment based on abnormal conditions. It can intervene before the equipment condition deteriorates, enabling the equipment group to maintain stable operation under dynamic load conditions, effectively reducing the probability of fault propagation, and improving the overall safety and availability of operation.

[0043] In one embodiment, step S10 includes: S101 sends acquisition commands to the sensor network deployed at the equipment site via fieldbus according to the preset data polling cycle; S102, in response to the acquisition command, the outdoor environment temperature data and humidity data are acquired through the temperature sensor and humidity sensor in the sensor network, and the temperature data and humidity data are encapsulated into environmental parameter data; S103, in response to the acquisition command, the operating parameter data of each cold source device in the device group is acquired through the current sensor, temperature sensor, pressure sensor and flow sensor in the sensor network. The operating parameter data includes current data, temperature data, pressure data and flow data. S104, the operating parameter data is mapped to real-time operating status data.

[0044] In this embodiment, a basic data channel is established around the operating environment and working status of the cold source equipment by periodically collecting real-time operating status data and environmental parameter data of the equipment group through a sensor network deployed at the equipment site. The equipment site refers to the machine room or equipment area where cold source equipment such as chillers, chilled water pumps, cooling water pumps, and cooling towers are concentrated. The sensor network covers this physical space, connecting the dispersed sensor nodes to the data acquisition unit via wired or wireless means. The sensor network can consist of multi-point acquisition modules, remote I / O units, and local smart meters. Each sensor node is managed through a unified address code, ensuring that the collected data corresponds one-to-one with specific equipment and specific measurement points. Periodic acquisition is achieved through a time-driven mechanism, controlling the sampling frequency through a unified time reference to keep the environmental parameter data and operating parameter data aligned on the time axis, providing a continuous time-series data foundation for subsequent analysis.

[0045] Based on a preset data polling cycle, acquisition commands are sent to the sensor network deployed at the equipment site via the fieldbus, establishing a scheduling mechanism for the sensor network. The data polling cycle is a configurable time parameter, such as 5 seconds, 10 seconds, or 30 seconds, used to specify the time interval at which the control system requests data from the sensor network. The fieldbus can use communication links widely used in industrial fields, such as Modbus, PROFIBUS, CAN bus, BACnet, or Ethernet-based industrial protocols. The control unit, acting as the master station, sends acquisition commands in the order of the polling table. The acquisition command includes the device identifier of the polling object, the sensor channel identifier, and the sampling mode flag. Each acquisition node in the sensor network accesses its local sensor interface according to the received acquisition command, completes a measurement, and packages and transmits the result back. By combining the polling cycle with the fieldbus, a balance can be achieved between the control system load and the data update frequency, ensuring data timeliness while avoiding bus communication congestion.

[0046] In response to acquisition commands, the system collects outdoor temperature and humidity data via temperature and humidity sensors in the sensor network. This data is then encapsulated into environmental parameter data to characterize the external thermal and humidity environment of the cooling source system. Temperature sensors, such as resistance temperature detectors (RTDs), thermocouples, or digital temperature chips, are installed at the cooling tower inlet or on the exterior wall of the computer room to reflect changes in air temperature. Humidity sensors, such as capacitive or resistive humidity modules, are arranged in conjunction with the temperature sensors to obtain relative humidity. The environmental parameter data is organized in a structured format, storing a record containing a timestamp, spatial location identifier, temperature data, humidity data, and unit information. If necessary, meteorological source identifiers or sensor status markers can be added for subsequent quality verification. Encapsulating environmental information separately as environmental parameter data facilitates comparison and correlation between external climate factors and energy consumption levels and equipment load conditions in subsequent analyses.

[0047] In response to acquisition commands, the system collects operating parameter data from each cooling source device in the equipment group through current sensors, temperature sensors, pressure sensors, and flow sensors in the sensor network. This operating parameter data includes current, temperature, pressure, and flow data, used to characterize the equipment's energy consumption, heat exchange status, and fluid transport status. Current sensors, which can be current transformers, Hall effect elements, or smart power meters, are installed in the motor input circuit or distribution cabinet circuit, outputting current data to reflect the equipment load level. Temperature sensors related to chiller units, chilled water pipelines, and cooling water pipelines are installed at the inlet and outlet of key pipe sections, outputting temperature data to characterize the cold energy transfer. Pressure sensors are installed on the high and low pressure sides of the pipeline or at key nodes, outputting pressure data or differential pressure data to reflect hydraulic balance and pump head changes. Flow sensors, which can be electromagnetic flow meters, vortex flow meters, or ultrasonic flow meters, output flow data to reflect the actual supply and return water volume. The operating parameter data is combined by the acquisition module to form a record set containing equipment identification, current data, temperature data, pressure data, and flow data. Fields such as timestamps, operating mode flags, and data quality markers can be added to provide complete input for subsequent status calculations and energy efficiency analysis.

[0048] This process maps operational parameter data to real-time operational status data, transforming low-level numerical information into a representation of equipment status. The mapping process includes basic processing such as data cleaning, outlier removal, interpolation, and unit conversion, as well as derivation calculations based on equipment nameplate parameters, design conditions, and operating boundaries. For example, load rate can be calculated from current data and rated current, cooling output can be estimated from inlet and outlet temperature data and flow rate data, and pressure data can determine whether a water pump is idling, overloaded, or operating normally. The mapping results can be represented in a multi-dimensional manner, including numerical status quantities and enumerated status labels, such as load rate percentage, estimated energy efficiency coefficient, operating level markers, and operating health level. Real-time operational status data is output in a unified format, allowing subsequent control logic, optimization algorithms, and anomaly diagnosis modules to make decisions at a higher level of abstraction, rather than directly processing raw sensor data, thus improving the clarity and scalability of the overall processing chain.

[0049] This embodiment establishes a periodic data acquisition mechanism based on sensor networks at the equipment site, introduces data polling cycle management, environmental parameter data encapsulation, and mapping of operating parameter data to real-time operating status data. This enables the synchronous acquisition and unified expression of environmental information and equipment operating information. While ensuring the timeliness and completeness of the acquired data, it provides structured and computable inputs for the control decisions, energy efficiency optimization, and status diagnosis of the cold source equipment group, thereby improving the accuracy and robustness of subsequent control processes.

[0050] In one embodiment, in step S30 above, the preliminary global optimization control strategy generated by the remote computing node based on the real-time operating status data and the environmental parameter data includes: S301, the remote computing node calculates the predicted value of the system cooling load demand based on the received real-time operating status data and environmental parameter data through a preset load prediction model. S302, Based on the predicted cooling load demand of the system, a preliminary global optimization control strategy is generated using a multi-objective optimization method.

[0051] In this embodiment, after receiving real-time operating status data and environmental parameter data, the remote computing node loads this data into the input interface of the load forecasting model in chronological order. The load forecasting model is pre-trained based on historical operating data and meteorological data to estimate the predicted system cooling load demand over a future forecast period. The model structure can employ time series regression, long short-term memory networks, gated recurrent networks, or gradient boosting-based ensemble models. It outputs a cooling demand sequence for several future time slots by inputting features such as equipment load rate, supply and return water temperature, flow rate, outdoor dry-bulb temperature, humidity, and weekday type. The forecast process can also incorporate current operating modes, start / stop status, and holiday indicators to avoid significant deviations during sudden changes in operating conditions. The predicted system cooling load demand not only includes the total cooling demand but can also be refined to different temperature zones, different floors, or different user zones, providing a basis for subsequent optimization.

[0052] After obtaining the predicted system cooling load demand, the remote computing node constructs an optimization calculation scenario based on this prediction result, generating a preliminary global optimization control strategy through a multi-objective optimization method. The multi-objective optimization method can simultaneously consider energy consumption indicators, operating cost indicators, and comfort deviation indicators when setting the objective function. Examples include minimizing the sum of the unit and pump power, minimizing the number of unit start-ups and shutdowns, and minimizing the cumulative deviation of supply and return water temperatures from design values. Constraints can include equipment start-up and shutdown constraints, minimum number of operating units, upper and lower limits for outlet water temperature, upper and lower limits for flow rate, and ramp rate constraints. The optimization solution process can employ linear programming, mixed integer programming, model predictive control, Lagrange relaxation, or heuristic global optimization algorithms. Through iterative calculation, it obtains the unit start-up and shutdown combinations, set outlet water temperature, target flow allocation, and operating modes of related auxiliary equipment for each time slice within the prediction time domain. The preliminary global optimization control strategy can be packaged into a time-series decision table, containing fields such as time axis, equipment number, start-up and shutdown status, and setpoint, and then sent back to the local control terminal via a downlink communication channel.

[0053] This embodiment establishes a computational chain between the local control terminal and the remote computing node to collaboratively generate a global optimization control strategy. It combines the load forecasting and multi-objective optimization capabilities of wide-area historical data and high-computing-power environments with the real-time perception and rapid correction capabilities close to the site. This allows the remote side to calculate a preliminary global optimization control strategy that takes into account energy consumption, comfort, and equipment constraints based on the predicted value of the system's cooling load demand, thus avoiding uneven energy efficiency distribution caused by single-point experience-based adjustments.

[0054] In one embodiment, after step S30 above, the method further includes: S303, Based on the latest real-time operating status data, the preliminary global optimization control strategy is revised to generate the final global optimization control strategy.

[0055] In this embodiment, after the remote computing node completes the initial global optimization control strategy calculation and distributes it, the local control terminal does not directly apply the strategy to the device group. Instead, it modifies the strategy based on the latest real-time operating status data to generate the final global optimization control strategy. To this end, the local control terminal first compares the currently collected real-time operating status data with the data time window used by the remote computing node for prediction and optimization, calculates the time difference between the two, and detects whether key operating variables have undergone significant changes within this time difference range, including changes in cooling demand, changes in device start / stop status, decrease in the output capacity of a single device, and abnormal sensor shielding. If a sudden change is detected on the load side or device side during the time comparison, the local control terminal uses this change as a correction condition to make local adjustments to the initial global optimization control strategy.

[0056] During the correction process, the local control unit parses the unit start-up and shutdown plans, set outlet water temperatures, target flow allocations, and associated operating configurations with various auxiliary equipment corresponding to the recent time slots from the initial global optimization control strategy. It then compares these strategy parameters with the current real-time operating status data item by item to identify command combinations that may cause layout instability or exceed equipment capacity limits. For time slots with deviations, the local control unit can smooth the execution rhythm by reducing the frequency of start-ups and shutdowns, delaying the start-up and shutdown times, or limiting repeated start-ups and shutdowns within a short period, avoiding frequent start-ups and shutdowns due to prediction errors. For setting outlet water temperature and flow targets, the local control unit can fine-tune the set values ​​based on the actual heat exchange capacity of the chiller unit, the cooling water temperature level, and the range of valve openings on site, ensuring that the adjustment is within the acceptable ramp-up rate range for the equipment.

[0057] In scenarios where equipment health status varies, the local control terminal can combine local maintenance strategies and equipment health assessment results to remove or downgrade equipment marked as temporarily disabled, operating at reduced capacity, or under abnormal observation from the initial global optimization control strategy. For example, when a unit is in maintenance standby mode or has multiple protective shutdown records in previous cycles, the local control terminal can transfer some or all of the corresponding start / stop commands and load allocation to similar equipment with better health status, while maintaining the total cooling output to meet the system's predicted cooling load demand level. For equipment that is nearing the upper limit of its lifespan set in the operation and maintenance strategy or has a high cumulative operating hours, its load share in the strategy can be reduced, extending the overall lifespan of the equipment while meeting current cooling demand.

[0058] To ensure that the revised strategy still structurally meets the requirements of the subsequent instruction parsing and execution modules, the local control terminal, after completing the local revision, will reorganize the strategy data. This involves reorganizing the adjusted start / stop states, set outlet water temperature, target flow allocation, and auxiliary equipment operating modes into a time series table, maintaining consistency between the time axis, equipment identifiers, and setpoint fields with the initial strategy. After this reorganization, the local control terminal will use the revised strategy as the final globally optimized control strategy for subsequent control instruction generation and instruction execution sequence construction, achieving a closed-loop connection from remote calculation results to on-site executable control actions.

[0059] This embodiment modifies the initial global optimization control strategy based on the latest real-time operating status data at the local control terminal. This introduces a compensation mechanism tailored to real-time field conditions without altering the global optimization capabilities of the remote computing node. On one hand, by utilizing time comparison and key variable offset detection, deviations caused by prediction delays, communication latency, and sudden load changes are transformed into quantifiable correction conditions. This allows the initial strategy to automatically adapt to the current operating conditions before being deployed to the equipment, reducing energy efficiency losses and control mismatch risks caused by strategy lag. On the other hand, by combining local maintenance strategies and equipment health assessment results to filter the strategy and reallocate loads, it can meet overall requirements while considering equipment lifespan and operational stability, avoiding continuously placing excessive loads on equipment in poor condition. Through this correction process, the final global optimization control strategy retains the global advantages obtained by the remote computing node based on load prediction and multi-objective optimization calculations, while also incorporating real-time correction capabilities close to the field, enabling the control system to achieve a better balance between energy efficiency, reliability, and equipment lifespan under dynamic operating conditions.

[0060] In one embodiment, step S40 above includes: S401, the local control terminal decodes the global optimization control strategy and extracts the target device unique identifier and target operating setting parameters contained in the global optimization control strategy; S402, the local control terminal maps the unique identifier of the target device to the underlying communication address of the physical device, and converts the target running setting parameters into specific control instructions to generate a control instruction set corresponding to each device; S403, the preset device control logic is invoked through the local control terminal. The preset device control logic includes a device cascading dependency relationship in which the action of the cooling circuit valve takes precedence over the action of the cooling water pump, and the action of the cooling water pump takes precedence over the action of the chiller unit. S404, the local control terminal sorts the control instructions in the control instruction set according to the device cascading dependency relationship, and inserts a status check step between adjacent control instructions to form an instruction execution sequence.

[0061] In this embodiment, after receiving the global optimization control policy at the local control terminal, a decoding process is first performed to restore the global optimization control policy from its transmission format to a locally parsable data structure. Global optimization control policies typically employ compact encoding methods during transmission, such as array structures indexed by time slices or configuration tables grouped by device. Internal fields include device identifiers, operating modes, and target operating settings. The decoding process needs to identify the encoding rules of these fields, breaking down information such as time indexes, device categories, and operating scenarios into explicit fields, generating structured data objects for subsequent mapping and sorting. To ensure compatibility between different policy versions, the decoding module can introduce a policy version number and a field mapping table, selecting the corresponding parsing template based on the version information to avoid parsing errors caused by changes in field position or naming.

[0062] After decoding, the local control unit extracts the target device's unique identifier and target operating settings from the global optimization control strategy. The target device's unique identifier is used to uniquely identify physical equipment such as cooling loop valves, cooling water pumps, and chillers at the logical level. It can be in the form of equipment codes, asset numbers, or combination keys (room number + loop number + equipment serial number) to achieve interface with the on-site asset management system. The target operating settings correspond to the control objectives that the equipment needs to achieve within the strategy's time slice, such as the chiller's outlet water temperature setpoint, the target frequency of the chilled water pump, the target opening degree of the cooling loop valve, and the start / stop status of the cooling water pump. The local control unit iterates through the decoded strategy structure, pairing the unique identifier corresponding to each time slice and each device with the target operating settings to form a logical-level set of control entries.

[0063] Subsequently, the local control unit needs to map the unique identifier of the target device to the underlying communication address of the physical device, and convert the target operating settings parameters into specific control commands. The underlying communication address is used to accurately locate the terminal device or control unit in the fieldbus or control network, such as the PLC register address, Modbus register number, CAN node ID plus offset, Ethernet remote I / O port number, etc. The mapping process typically relies on a device resource mapping table, which is maintained by the commissioning personnel during the system configuration phase, recording the correspondence between logical device identifiers and communication addresses. The local control unit queries the mapping table to replace the unique identifier of each target device with address information that can be directly used for communication. Based on this, the target operating settings parameters are encoded into control commands according to different device types and communication protocols. For example, the outlet water temperature setpoint is converted into a value written to an analog register, the start / stop status is converted into a Boolean value written to the coil, and the valve target opening degree is converted into a percentage or step position. This ultimately forms a control command set corresponding to each device, with each control command containing a timestamp, communication address, written value, and necessary check fields.

[0064] After generating the control command set, the local control terminal invokes preset equipment control logic to constrain the execution order of the control commands. This equipment control logic reflects the physical dependencies between the cooling and refrigeration circuits, clearly defining the cascading order where cooling circuit valve actions take precedence over cooling water pump actions, and cooling water pump actions take precedence over chiller unit actions. To implement this logic, the local control terminal internally maintains an equipment cascading dependency graph, abstracting devices such as valves, pumps, and chillers as nodes, and abstracting "must-act first" relationships as directed edges. For example, if the cooling circuit inlet valve node points to the cooling water pump node, and the cooling water pump node then points to the chiller unit node, this indicates that when executing control commands, the valve action must be completed first, followed by the pump action, and finally the unit start / stop or setpoint adjustment. The dependency graph can simultaneously contain the topology of multiple circuits, thus maintaining overall sequence constraints when multiple units, pumps, and circuits coexist.

[0065] The local control terminal sorts the control instructions in the control instruction set according to the device cascading dependencies. The sorting process can employ a topological sorting algorithm, dividing all control instructions into several execution levels based on their dependencies: the first level contains instructions that do not depend on other actions, such as opening cooling circuit valves; the second level contains instructions that depend on the completion of actions in the previous level, such as starting the cooling water pump; the third level contains instructions that can only be executed after a stable flow rate is established in the water circuit, such as starting the chiller unit or adjusting the outlet water temperature. For control instructions involving multiple devices within each time slice, the local control terminal first divides the hierarchy based on device cascading dependencies, and then determines the specific order within the same level based on factors such as communication load and device response time, generating a time series that satisfies dependency constraints and has a low risk of conflict.

[0066] After completing the timing sequence, the local control terminal inserts status check steps between adjacent control commands. These status check steps are used to confirm the status of critical equipment after executing the previous control command and before initiating the next. For example, after executing the cooling circuit valve opening command, the status check step can read the valve's on / off feedback or analog opening feedback to verify whether the valve has reached the target state; after executing the cooling water pump start command, the status check step can check whether the motor current has entered a stable range, whether the flow sensor has detected effective flow, and whether the pressure sensor displays a reasonable differential pressure. Only when the status check results meet preset conditions will the command execution sequence advance to the next control command; otherwise, exception handling logic can be triggered in subsequent stages. By inserting status check steps between commands, the command execution sequence is expanded from a simple list of commands to a combined sequence of "action + confirmation," providing a structural foundation for safety control and rollback control in subsequent execution stages.

[0067] This embodiment parses the global optimization control strategy at the local control terminal, constructing a complete mapping chain from strategy objects to control instruction sets and then to instruction execution sequences. On the one hand, it utilizes the mapping relationship between the unique identifier of the target device and the underlying communication address of the physical device to reliably sink the abstracted optimization results to the field execution layer, avoiding configuration errors and response delays caused by manual conversion. On the other hand, by introducing the cascading dependency relationship of devices including cooling circuit valves, cooling water pumps, and chiller unit sequence constraints, and combining the status check step in the timing sequence, the physical topology of the devices and safe operation constraints are explicitly reflected in the instruction execution sequence. This enables the cooling source system to automatically follow the correct action sequence and perform status confirmation at key nodes when performing start-up, shutdown, and setting adjustments, thereby reducing the risk of misoperation and equipment damage. At the same time, it provides a clear and controllable execution trajectory basis for subsequent anomaly detection and rollback control.

[0068] In one embodiment, step S50 above includes: S501, retrieve the control instruction to be executed sequentially from the instruction execution sequence, send the control instruction to the corresponding target device, and start the timer; S502, within a preset time of the timer, continuously monitor the device response status of the target device and compare the device response status with the expected status; S503, when the device response status is not received within the preset time, or the device response status is inconsistent with the expected status, it is determined that the device response status indicates an abnormality, and the subsequent execution of the instruction execution sequence is terminated; S504 triggers a rollback operation, generating a corresponding equipment stop instruction sequence based on the successfully executed control instruction record; S505, in the reverse order of execution, each stop instruction in the device stop instruction sequence is executed sequentially, so that each device is restored to the state before the start of the instruction execution sequence, so that the device returns to a safe state.

[0069] In this embodiment, after the instruction execution sequence has been constructed at the local control end according to the device cascading dependencies and status check order, the execution control module needs to retrieve the control instructions one by one. The execution control module typically maintains an ordered queue or index pointer and, based on signaling triggering or periodic scheduling mechanisms, retrieves the currently pending control instructions from the instruction execution sequence in sequence. Each control instruction contains the target device's address information, target operating settings, operation type, and necessary verification fields. After retrieving the control instructions, the execution control module encapsulates them into field-recognizable messages according to the communication protocol used by the target device. For example, it constructs write-to-single-register, write-to-multiple-register, coil write, or vendor-specific protocol frames, and sends them to the corresponding control unit of the target device via fieldbus, industrial Ethernet, or serial bus. After transmission, to constrain the instruction execution process and monitor its status, the execution control module immediately starts a timer, records the current time as the start time, and sets a timeout threshold. The timer can be implemented using an operating system-level timer, a real-time task scheduler's internal timer, or a dedicated hardware timing unit.

[0070] During the timer's operation, the status acquisition module continuously monitors the device's response status. The device response status is not limited to a single signal but can be composed of a combination of various feedback quantities. For example, for motor-type devices, the device response status may include the status of operating indicator contacts, current detection values, and speed or frequency feedback values; for valve devices, the device response status may include switch feedback signals, valve position percentage feedback, actuator current, etc.; for chiller units, the device response status may include compressor operating flags, evaporation pressure, chilled water outlet and return temperatures, alarm flags, etc. The status acquisition module periodically reads these feedback quantities from the field I / O modules, PLC registers, or the internal status registers of the intelligent device through polling or interrupt triggering, and combines the acquisition results to form the device response status at the current moment. Simultaneously, the execution control module determines the expected status based on the target operating setpoint of the control command, the device type, and the operating scenario. For example, it may expect the operating flag to be closed, the current to be within the allowable range, the valve position to be close to the target opening, and the temperature or pressure to gradually converge towards the target value. Then, the collected device response status is compared with the expected status. The comparison method can be discrete state consistency judgment, analog quantity range judgment, or trend change direction judgment, so as to obtain the judgment result of whether the current control command is effectively executed by the device.

[0071] If a valid device response status cannot be obtained from the status acquisition module before the timer reaches the preset time threshold, or if multiple consecutive acquisition results do not match the expected status, the anomaly detection module will mark the current situation as an abnormal device response status indication. Abnormal device response status indications include communication layer problems such as no feedback or lost feedback signals, as well as device layer problems such as motor start-up failure, valve jamming, and unit self-test failure. Once an anomaly is determined, the execution control module immediately suspends the subsequent execution of the instruction execution sequence and marks any unexecuted instructions after the current control instruction as pending cancellation. This prevents the continued sending of start / stop or setting adjustment instructions to downstream equipment when critical equipment is not ready, thus avoiding unsafe equipment combination states.

[0072] To reliably restore the system to a safe state after an anomaly occurs, the rollback control module continuously maintains a record of successfully executed control commands during operation. This record can be designed as a stack structure or a sequential list with timestamps. When a control command receives normal feedback and is determined to have been executed successfully, the control command, its target device, operation type, and original setpoint are pushed into the record structure. In the event of an anomaly, the rollback control module generates a sequence of device stop commands based on the record of successfully executed control commands. The generation process includes constructing a logically reversed stop command for each historical control command. For example, a stop command is constructed for start-type commands, a command to restore to a safe setpoint is constructed for setpoint increase commands, and a command to close or adjust to a safe opening is constructed for valve opening adjustment commands. The device stop command sequence is arranged in the reverse order of the original execution order; that is, the control command executed later generates its corresponding stop command first, and the control command executed earlier generates its corresponding stop command later, thus forming a command chain mirroring the original execution order.

[0073] The rollback execution process follows the sequence of equipment stop commands. Each stop command is confirmed before being sent using a timer and status acquisition module. The execution control module retrieves the current stop command from the sequence, encapsulates it for communication, and sends it to the target device. It then monitors the new device response status to see if the stop action has been completed, such as the running flag being off, current dropping to no-load or zero, valve position feedback returning to the closed position, or alarm status being cleared. Only when the stop action is confirmed to be complete does the rollback control module proceed to the next higher-level stop command, reversing the already executed equipment actions step by step. This rollback method, reversing the execution sequence, ensures that upstream equipment dependent on downstream conditions stops last, and downstream equipment stops first, maintaining the physical link under control throughout the rollback process. This avoids dangerous combinations such as pumps stopping while valves remain wide open, or units being under load while cooling circuits are not established. After the entire equipment stop command sequence is executed, the system operating state is restored to the state before the command execution sequence began or the preset safe state, such as all related equipment being shut down, valves returning to safe opening degrees, and current and pressure being within safe ranges, providing a stable foundation for subsequent fault diagnosis or restart.

[0074] This embodiment constructs a closed-loop control chain from instruction issuance to equipment feedback verification by executing control instructions according to the instruction execution sequence and continuously monitoring the equipment response status during execution. When the equipment response status indicates an anomaly, a sequence of equipment stop instructions is generated using the successfully executed control instruction records, and each stop instruction is executed sequentially in the reverse order of execution, causing the equipment operating state to converge backward to a safe state according to the dependency relationship. In this way, on the one hand, anomalies can be detected in time when critical equipment does not act as expected, and subsequent instructions can be blocked, avoiding the continued accumulation of operations in an uncertain state and reducing the risk transmission between linked equipment; on the other hand, through stack-based rollback control, each start, stop, or setting adjustment operation is bound to a corresponding reverse undo path, so that the system can orderly undo the effective actions when an anomaly occurs, automatically restoring the operating state to a safe range, thereby improving the safety and controllability of the automated control process and reducing the impact of manual intervention on the fault handling sequence.

[0075] In one embodiment, step S60 above includes: S601, sample the target equipment in the operating equipment group to obtain the real-time operating current value, output frequency value and inlet / outlet pressure difference value, and combine the real-time operating current value, output frequency value and inlet / outlet pressure difference value as a parameter characterizing the operating status of the equipment. S602, the parameters characterizing the operating status of the equipment are input into the isolated forest anomaly detection model pre-installed on the local control terminal for feature separation calculation, and anomaly scores reflecting the health status of the target equipment are output in real time. S603, compare the abnormal score with the preset alarm threshold in real time. When the abnormal score is continuously higher than the alarm threshold within the preset fault confirmation time window, determine that the parameter meets the preset abnormal conditions, and determine the fault type of the target device based on the abnormal feature analysis results. S604, when the fault type is a frequency converter drive module fault, trigger the operation mode adjustment operation to control the target equipment to switch from frequency converter operation mode to power frequency operation mode; S605, when the fault type is a mechanical fault of the equipment body, trigger the equipment switching operation, stop the operation of the target equipment and start the redundant backup equipment of the same type.

[0076] In this embodiment, while the equipment group remains operational, the local control terminal collects multi-dimensional measurement data reflecting the operational health status from the target equipment using a predetermined sampling strategy. The target equipment can be a single unit such as a chilled water pump, cooling water pump, or chiller unit. The sampling process is typically executed with a fixed sampling period or event-triggered method. For example, every few seconds, the real-time operating current value, output frequency value, and inlet / outlet pressure difference value are read from the sensor link or driver interface. The real-time operating current value originates from the current transformer or the current detection channel inside the frequency converter, reflecting the load size and motor stress. The output frequency value is typically provided by the frequency converter drive module, characterizing the motor speed command and actual execution level. The inlet / outlet pressure difference value is obtained through pressure sensors installed on the equipment's inlet and outlet pipelines; the pressure difference reflects whether the flow rate is sufficient, whether the pipeline is blocked, and whether heat exchange is obstructed. Within each sampling period, the sampling module combines the three types of values ​​according to the equipment identifier to construct a parameter vector characterizing the equipment's operating status. This vector includes both instantaneous amplitude and statistical quantities within a short time window, such as average, variance, or rate of change, providing input for subsequent anomaly detection.

[0077] An isolated forest anomaly detection model is pre-deployed within the local control unit. The parameter vectors collected each time are input into this model to perform feature separation calculations. The isolated forest model consists of multiple randomly constructed tree structures. By randomly partitioning the feature space, samples are progressively divided into different regions. The feature separation calculation process can be understood as traversing downwards from the root node in each tree, randomly partitioning based on dimensions such as current, frequency, and voltage difference, until the parameter vector falls into a leaf node or reaches a predetermined depth, thus obtaining the corresponding path length. During the training phase, the model uses a large number of historical normal samples for fitting, ensuring that normally operating regions correspond to longer average paths. Data deviating from normal regions are more easily partitioned early in most trees, resulting in shorter paths. During online operation, the local control unit calculates the average path length in the forest for each newly sampled parameter vector and generates anomaly scores using a predetermined transformation function. The higher the anomaly score, the more the operating state deviates from the historical normal distribution from the model's perspective, indicating a lower level of equipment health. Because the model is deployed on the local control unit, parameter acquisition and anomaly score calculation can be completed within milliseconds to seconds, enabling real-time assessment of the target equipment's health status.

[0078] After anomaly scores are generated, the local control unit continuously compares them with preset alarm thresholds using comparison logic. The alarm thresholds are determined during the configuration phase based on historical operating data and maintenance experience, corresponding to an acceptable maximum level of anomaly. To avoid misjudgments caused by momentary interference, the local control unit introduces a fault confirmation time window, analyzing the anomaly score sequence over multiple consecutive sampling periods. If the anomaly score remains consistently higher than the alarm threshold throughout the time window, and no samples recover to the normal range, it is considered that the parameter vector has stably fallen within the abnormal region for a period of time, and the parameters characterizing the equipment's operating status have met the preset anomaly conditions. Upon reaching this condition, the fault diagnosis module further performs feature analysis based on parameter anomaly patterns. For example, a significant increase in current accompanied by output frequency fluctuations, but with minimal pressure difference changes, can be identified as motor overload or drive control anomaly; current and frequency remaining near the command level while the inlet and outlet pressure difference decreases significantly, can be identified as water circuit blockage, pump damage, or impeller degradation. By classifying different parameter combinations and trend characteristics, fault types can be categorized into variable frequency drive module faults, equipment mechanical faults, etc., providing input for subsequent handling logic.

[0079] When the fault diagnosis indicates a fault in the frequency converter drive module, the local control terminal triggers an operation mode adjustment. During this adjustment, control logic sends control commands to the frequency converter drive module and bypass control unit. First, the frequency converter output is deactivated or the frequency command is gradually reduced to zero. Then, the pre-configured power frequency bypass contactor is closed, switching the power supply to the target equipment to the power frequency circuit. Necessary dead time and interlocking conditions can be set during the switching process, such as ensuring the power frequency contactor is only allowed to close after the frequency converter output is completely disconnected, preventing the two power supply circuits from operating in parallel. Simultaneously, the number of starts and the start interval can be limited during the initial stage of power frequency operation to avoid grid impact caused by frequent switching when the drive module is faulty. When the fault diagnosis indicates a mechanical fault in the equipment itself, the local control terminal no longer maintains equipment operation through operation mode adjustment but instead initiates an equipment switching operation. The equipment switching logic first sends a shutdown command to the target equipment. After key quantities such as current, frequency, and differential voltage drop to safe ranges, the equipment is marked as unavailable to prevent subsequent scheduling modules from selecting it again. Simultaneously, select the same type of standby equipment from the redundant resource pool, allocate matching operating conditions according to the current system load and pipeline topology, and then execute the startup sequence for the standby equipment, including valve opening pre-adjustment, pump start-up and speed-up, unit commissioning and other operations, to ensure that the system can quickly recover to an acceptable cooling output level after a mechanical failure.

[0080] This embodiment continuously collects multi-dimensional parameters such as current, frequency, and inlet / outlet pressure difference during the operation of the equipment group. These parameters are combined into a unified state vector and input into an isolated forest anomaly detection model for feature separation calculation. This enables the construction of real-time health assessment capabilities without relying on a large number of labeled fault samples. Anomaly scores and alarm thresholds are continuously compared within the fault confirmation time window, so that anomaly judgment no longer relies on single-point thresholds, but combines the time dimension for stability verification, thereby reducing the probability of false alarms caused by short-term disturbances. When the parameter mode meets the anomaly conditions, feature analysis distinguishes between frequency converter drive module faults and equipment mechanical faults, triggering operation mode adjustment or equipment switching operations respectively. This allows the system to maintain cooling output through power frequency operation in drive failure scenarios, while quickly switching to redundant equipment to maintain continuous system operation in mechanical failure scenarios.

[0081] In one embodiment, step S602 includes: S6021, Perform data standardization preprocessing on the parameters characterizing the operating status of the equipment to generate standardized data with unified dimensions; S6022, The normalized data is input into the pre-trained isolated forest anomaly detection model that is pre-installed on the local control terminal. The isolated forest anomaly detection model contains multiple isolated tree structures. S6023, In the isolated forest anomaly detection model, the length of the isolated path of the normalized data in each isolated tree structure is calculated; S6024, calculate the arithmetic mean of the isolated path lengths obtained in all isolated tree structures to obtain the average path length value; S6025, Based on the average path length value, an anomaly score reflecting the health status of the target device is calculated using a preset anomaly score conversion function.

[0082] In this embodiment, the parameters characterizing the equipment's operating status have already been constructed into a multi-dimensional vector in the previous stage through sampling of current, frequency, and inlet / outlet pressure difference. When this vector enters the anomaly detection calculation process within the local control terminal, it first undergoes data standardization preprocessing. Data standardization preprocessing addresses the differences in dimensions and numerical ranges between different physical quantities by mapping the values ​​of each dimension to a unified interval or a unified statistical distribution through linear or nonlinear transformations. For example, current values ​​might be in amperes, pressure difference in kilopascals, and frequency in hertz. Directly inputting these into the model could easily lead to a certain physical quantity dominating distance calculations due to its large numerical range. Standardization can employ mean-variance standardization, subtracting the historical mean from each dimension and dividing by the standard deviation, or interval scaling, mapping the sampled values ​​to a closed interval between zero and one. The standardized data generated through this process has a unified dimensional meaning across all dimensions, thus making subsequent calculations based on distance or depth of division more stable. The standardized data exists in array or vector form, with the element order consistent with the original parameter dimensions, facilitating dimensional access within the model.

[0083] The pre-trained and pre-installed isolated forest anomaly detection model on the local control unit has already undergone parameter fitting using a large amount of historical operational data during the deployment phase. Internally, it contains multiple isolated tree structures. An isolated tree structure can be understood as a binary tree constructed for multi-dimensional data. At each level, the data range is divided into two sub-regions by randomly selecting a dimension and a splitting threshold. Each tree in the forest uses different random seeds and sample subsets during construction, resulting in diverse partitioning methods. The model pre-training phase primarily completes two tasks: first, determining the depth distribution and splitting threshold of each isolated tree; and second, recording the average path depth of normally operating data in the forest, providing a benchmark for subsequent online computation. After the model is pre-installed on the local control unit, no further structure updates are performed. Instead, these tree structures are reused during online inference to evaluate newly collected normalized data, enabling rapid response.

[0084] Within the Isolation Forest anomaly detection model, normalized data is traversed according to a tree structure, and the length of the isolated path in each isolated tree is calculated. The path length can be defined as follows: starting from the root node, based on the current node's split dimension and threshold, the normalized data's value in a specific dimension is determined layer by layer to see if it falls into the left or right sub-region, until a leaf node or the maximum depth specified in the tree construction is reached, thus obtaining the isolated path length. For different isolated tree structures, due to differences in split order and thresholds, the path length of the same normalized data may vary. Isolation Forest uses this difference to assess the ease with which a sample is "isolated"; samples in sparse regions often exhibit shorter path lengths across multiple trees.

[0085] The arithmetic mean of the isolated path lengths obtained from all isolated tree structures is used to obtain the average path length value, which is a comprehensive evaluation of the results of each tree in the forest. The arithmetic mean operation integrates the local judgments of each tree into a global indicator, avoiding the excessive influence of the accidental partitioning of a single tree on the result. In implementation, a path length array can be allocated for each input normalized data, with the array length equal to the number of isolated trees. When traversing the forest, the path length of each tree is filled in sequentially. Finally, a single addition and division operation is performed on the local control terminal to obtain the average path length value. The average path length value is closely related to the feature space partitioning of the forest during training. Generally, the closer it is to the typical depth of most samples in the training dataset, the closer the running state is to the historical normal state.

[0086] The anomaly score conversion function takes the average path length as input and outputs an anomaly score reflecting the health status of the target device. The conversion function can employ the exponential form given by the isolated forest theory, comparing the path length with the expected path length of the corresponding subsample and mapping it to a score between zero and one. For example, a shorter path outputs an anomaly score close to one, while a longer path outputs an anomaly score close to zero. Alternatively, based on field operation and maintenance experience, a piecewise linear function or interpolation function can be used to map the average path length range to multiple anomaly levels. The conversion function is implemented in the local control terminal using mathematical expressions or lookup tables, completing the numerical conversion with a single function call. The anomaly score reflecting the health status of the target device, as a dimensionless indicator, is independent of device type and only relates to the relative position of the operating status in the feature space. It can be used for subsequent threshold determination, trend analysis, or health score curve plotting. Through this process, the originally collected multidimensional physical quantities are compressed into a single numerical indicator, while preserving the distribution information of multidimensional features in high-dimensional space.

[0087] This embodiment eliminates the interference of differences in dimensions and numerical scales between different physical quantities on model calculations by performing data standardization preprocessing on parameters characterizing the equipment's operating status. This allows the random partitioning within the isolated forest to be carried out more fairly across all dimensions, thereby improving the sensitivity of anomaly detection to multidimensional features. The pre-trained isolated forest anomaly detection model, pre-installed on the local control terminal, maps the complex multidimensional operating status to an average path length value through path length calculation and arithmetic averaging of multiple isolated tree structures. Then, it combines this with anomaly score conversion function to generate a single anomaly score, transforming health assessment from a multi-parameter intuitive judgment into a unified numerical indicator.

[0088] like Figure 3 As shown, the device dynamic control method based on edge-cloud collaboration for remote computing nodes proposed in this invention includes the following steps: S701 receives real-time operating status data and environmental parameter data of the device group periodically collected by the sensor network deployed at the equipment site, sent by the local control terminal. S702, Based on the real-time operating status data and the environmental parameter data, the predicted value of the system cooling load demand is calculated through a preset load prediction model; S703: Based on the predicted system cooling load demand, a preliminary global optimization control strategy is generated using a multi-objective optimization method. This preliminary global optimization control strategy is then sent to the local control terminal, which parses the strategy to generate control instruction sets corresponding to each device. The control instruction sets are then sorted according to preset device control logic to form an instruction execution sequence. Each control instruction is executed according to the instruction execution sequence, and the device response status is monitored during execution. If the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. During the operation of the device group, parameters characterizing the device operating status are continuously monitored. If these parameters meet preset abnormal conditions, a device switching or operating mode adjustment operation is initiated.

[0089] In this embodiment, the remote computing node receives real-time operating status data and environmental parameter data collected by the field sensor network and processed by the local control terminal through a communication link established with the local control terminal. Before transmission, the data has been timestamped, encapsulated with device and measurement point numbers, and marked with basic quality tags locally. After receiving the data, the remote computing node organizes it according to time windows and device groups, loading the continuous historical input sequence into the input interface of the load forecasting model. The load forecasting model is pre-trained based on historical operating records and external meteorological data to estimate the system cooling load demand during the future forecast period. The model structure can adopt a time series regression structure, a recurrent neural network structure, or an integrated structure based on gradient boosting. By introducing features such as device load rate, supply and return water temperature, flow rate, outdoor temperature and humidity, calendar attributes, and operating mode markers, it outputs the cooling load demand values ​​corresponding to each time slice on the forecast time axis, thereby forming a sequence of predicted system cooling load demand values.

[0090] After obtaining the predicted system cooling load demand, the remote computing node constructs an optimization calculation scenario based on the predicted cooling load sequence. The cooling load demand, upper and lower limits of equipment capacity, start-stop frequency constraints, outlet water temperature and flow boundaries, and energy consumption indicators are input into the multi-objective optimization solution module. After setting multiple objectives such as energy consumption, operational stability, and load tracking accuracy, the multi-objective optimization solution module iteratively searches for combined operation schemes of multiple devices using linear programming, mixed integer programming, or other numerical optimization algorithms to generate a preliminary global optimization control strategy covering the predicted period. This strategy is expressed in the form of a time-series configuration table, including time slices, equipment identifiers, start-stop status, temperature setpoints, flow target values, and necessary operating mode fields. After calculation, the remote computing node sends the preliminary global optimization control strategy back to the local control terminal via a downlink communication channel. The local control terminal then continues to perform strategy parsing, control instruction set generation, instruction execution sequence construction, equipment response monitoring and rollback in abnormal situations, and equipment switching or operating mode adjustment. This allows the remote node to centrally undertake the prediction and optimization calculation tasks, while the local node performs execution control tightly coupled with the on-site operating conditions.

[0091] This embodiment utilizes load forecasting models and multi-objective optimization to generate a preliminary global optimization control strategy on the remote computing node side. This decouples the computationally intensive forecasting and optimization process from the field end, reducing the burden on the local control end and establishing unified constraints among energy consumption, equipment start-up and shutdown frequency, and load tracking accuracy. This makes the control commands subsequently executed by the local control end closer to the globally optimal configuration, thereby improving the energy efficiency and operational stability of multi-device collaborative operation and reducing the risk of uneven energy distribution and control deviation caused by relying on manual experience for adjustment.

[0092] In one embodiment, a device dynamic control apparatus based on edge-cloud collaboration is provided, which corresponds one-to-one with the device dynamic control method based on edge-cloud collaboration described in the above embodiments. (Refer to...) Figure 4 , Figure 4 This is a schematic diagram of the functional modules of a preferred embodiment of the device dynamic control device based on edge-cloud collaboration according to the present invention. The modules include a sensor acquisition module 10, a strategy generation module 20, an instruction parsing module 30, an instruction execution module 40, and an anomaly monitoring module 50. Detailed descriptions of each functional module are as follows: The sensor acquisition module 10 is used to periodically collect real-time operating status data and environmental parameter data of the equipment group through a sensor network deployed at the equipment site; The strategy generation module 20 is used to generate a global optimization control strategy in collaboration between the local control terminal and the remote computing node based on the real-time running status data and environmental parameter data. The instruction parsing module 30 is used to parse the global optimization control strategy on the local control terminal to generate control instruction sets corresponding to each device, and sort the control instruction sets according to the preset device control logic to form an instruction execution sequence. The instruction execution module 40 is used to execute each control instruction according to the instruction execution sequence, and monitor the device response status during the execution process. When the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. The anomaly monitoring module 50 is used to continuously monitor parameters characterizing the operating status of the equipment during the operation of the equipment group, and to initiate equipment switching or operating mode adjustment operations when the parameters meet preset anomaly conditions.

Claims

1. A device dynamic control method based on edge-cloud collaboration, characterized in that, When applied to a local control terminal, the following steps are included: The sensor network deployed at the equipment site periodically collects real-time operating status data and environmental parameter data of the equipment group; The real-time operating status data and the environmental parameter data are sent to a remote computing node; Receive the preliminary global optimization control strategy generated by the remote computing node based on the real-time operating status data and the environmental parameter data; The global optimization control strategy is parsed to generate control instruction sets corresponding to each device, and the control instruction sets are sorted according to the preset device control logic to form an instruction execution sequence; The system executes each control instruction according to the instruction execution sequence, and monitors the device response status during the execution process. When the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. During the operation of the equipment group, parameters characterizing the operating status of the equipment are continuously monitored. When the parameters meet preset abnormal conditions, equipment switching or operating mode adjustment operations are initiated.

2. The device dynamic control method based on edge-cloud collaboration as described in claim 1, characterized in that, Real-time operating status data and environmental parameter data of the equipment group are periodically collected by a sensor network deployed at the equipment site, including: Based on the preset data polling cycle, the data acquisition command is sent to the sensor network deployed at the equipment site via the fieldbus; In response to the acquisition command, outdoor temperature and humidity data are acquired through the temperature and humidity sensors in the sensor network, and the temperature and humidity data are encapsulated into environmental parameter data. In response to the acquisition command, the operating parameter data of each cold source device in the device group is acquired through the current sensor, temperature sensor, pressure sensor and flow sensor in the sensor network. The operating parameter data includes current data, temperature data, pressure data and flow data. The operating parameter data is mapped to real-time operating status data.

3. The device dynamic control method based on edge-cloud collaboration as described in claim 1, characterized in that, The remote computing node generates a preliminary global optimization control strategy based on the real-time operating status data and the environmental parameter data, including: The remote computing node calculates the predicted value of the system cooling load demand based on the received real-time operating status data and environmental parameter data through a preset load prediction model. Based on the predicted cooling load demand of the system, a preliminary global optimization control strategy is generated using a multi-objective optimization method.

4. The device dynamic control method based on edge-cloud collaboration as described in claim 1, characterized in that, After receiving the preliminary global optimization control strategy generated by the remote computing node based on the real-time operating status data and the environmental parameter data, the system further includes: The preliminary global optimization control strategy is revised based on the latest real-time operating status data to generate the final global optimization control strategy.

5. The device dynamic control method based on edge-cloud collaboration as described in claim 1, characterized in that, The global optimization control strategy is parsed to generate control instruction sets corresponding to each device, and the control instruction sets are sorted according to preset device control logic to form an instruction execution sequence, including: The local control terminal decodes the global optimization control strategy and extracts the target device unique identifier and target operating setting parameters contained in the global optimization control strategy. The local control terminal maps the unique identifier of the target device to the underlying communication address of the physical device, and converts the target running setting parameters into specific control instructions to generate control instruction sets corresponding to each device. The local control terminal calls the preset equipment control logic, which includes a cascading dependency relationship where the action of the cooling circuit valve takes precedence over the action of the cooling water pump, and the action of the cooling water pump takes precedence over the action of the chiller unit. The local control terminal sorts the control instructions in the control instruction set according to the device cascading dependency relationship, and inserts a status check step between adjacent control instructions to form an instruction execution sequence.

6. The device dynamic control method based on edge-cloud collaboration as described in claim 1, characterized in that, Execute each control instruction according to the instruction execution sequence, and monitor the device response status during execution. When the device response status indicates an abnormality, trigger a rollback operation to return the device to a safe state, including: The control command to be executed is retrieved sequentially from the instruction execution sequence, the control command is sent to the corresponding target device, and a timer is started. Within a preset time of the timer, the device response status of the target device is continuously monitored, and the device response status is compared with the expected status. If the device response status is not received within the preset time, or if the device response status is inconsistent with the expected status, the device response status is determined to be abnormal, and the subsequent execution of the instruction execution sequence is suspended. Trigger the rollback operation and generate the corresponding device stop instruction sequence based on the successfully executed control instruction record; In the reverse order of execution, each stop instruction in the device stop instruction sequence is executed sequentially, so that each device is restored to the state before the start of the instruction execution sequence, so that the device returns to a safe state.

7. The device dynamic control method based on edge-cloud collaboration as described in claim 1, characterized in that, During the operation of the equipment group, parameters characterizing the equipment's operating status are continuously monitored. When these parameters meet preset abnormal conditions, equipment switching or operating mode adjustment operations are initiated, including: Sampling is performed on the target equipment in the operating equipment group to obtain real-time operating current value, output frequency value and inlet / outlet pressure difference value, and the real-time operating current value, output frequency value and inlet / outlet pressure difference value are combined as parameters characterizing the operating status of the equipment. The parameters representing the operating status of the equipment are input into the isolated forest anomaly detection model pre-installed on the local control terminal for feature separation calculation, and anomaly scores reflecting the health status of the target equipment are output in real time. The abnormal score is compared with a preset alarm threshold in real time. When the abnormal score is continuously higher than the alarm threshold within a preset fault confirmation time window, the parameter is determined to meet the preset abnormal conditions, and the fault type of the target device is determined based on the abnormal feature analysis results. When the fault type is a frequency converter drive module fault, an operation mode adjustment operation is triggered to control the target equipment to switch from frequency converter operation mode to power frequency operation mode. When the fault type is a mechanical fault of the equipment itself, a equipment switching operation is triggered to stop the operation of the target equipment and start a redundant backup equipment of the same type.

8. The device dynamic control method based on edge-cloud collaboration as described in claim 7, characterized in that, The parameters characterizing the device's operating status are input into a pre-installed isolated forest anomaly detection model on the local control terminal for feature separation calculation. The model then outputs anomaly scores reflecting the health status of the target device in real time, including: The parameters characterizing the operating status of the equipment are subjected to data standardization preprocessing to generate standardized data with uniform dimensions; The normalized data is input into the pre-trained isolated forest anomaly detection model that is pre-installed on the local control terminal. The isolated forest anomaly detection model contains multiple isolated tree structures. In the isolated forest anomaly detection model, the length of the isolated path in each isolated tree structure of the normalized data is calculated; Calculate the arithmetic mean of the isolated path lengths obtained from all isolated tree structures to obtain the average path length value; Based on the average path length value, an anomaly score reflecting the health status of the target device is calculated using a preset anomaly score conversion function.

9. A device dynamic control method based on edge-cloud collaboration, characterized in that, Applied to remote computing nodes, including the following steps Receive real-time operating status data and environmental parameter data of the equipment group periodically collected by the sensor network deployed at the equipment site from the local control terminal; Based on the real-time operating status data and the environmental parameter data, the predicted value of the system cooling load demand is calculated through a preset load prediction model. Based on the predicted system cooling load demand, a preliminary global optimization control strategy is generated using a multi-objective optimization method. This preliminary global optimization control strategy is then sent to the local control terminal, which parses the strategy to generate control instruction sets corresponding to each device. The control instruction sets are then sorted according to preset device control logic to form an instruction execution sequence. Each control instruction is executed according to the instruction execution sequence, and the device response status is monitored during execution. If the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. During the operation of the equipment group, parameters characterizing the operating status of the equipment are continuously monitored. When the parameters meet preset abnormal conditions, equipment switching or operating mode adjustment operations are initiated.

10. A device dynamic control apparatus based on edge-cloud collaboration, characterized in that, The device dynamic control device based on edge-cloud collaboration includes: The sensor acquisition module is used to periodically collect real-time operating status data and environmental parameter data of the equipment group through a sensor network deployed at the equipment site; The strategy generation module is used to generate a global optimization control strategy in collaboration between the local control terminal and the remote computing node based on the real-time running status data and environmental parameter data. The instruction parsing module is used to parse the global optimization control strategy on the local control terminal to generate control instruction sets corresponding to each device, and sort the control instruction sets according to the preset device control logic to form an instruction execution sequence. The instruction execution module is used to execute each control instruction according to the instruction execution sequence, and monitor the device response status during the execution process. When the device response status indicates an abnormality, a rollback operation is triggered to return the device to a safe state. An anomaly monitoring module is used to continuously monitor parameters characterizing the operating status of the equipment during the operation of the equipment group. When the parameters meet preset anomaly conditions, the module initiates equipment switching or operating mode adjustment operations.

Images