Edge intelligent agent collaborative optimization method based on cloud-edge collaboration

Abstract

A cloud edge collaborative based edge intelligent agent collaborative optimization method, the method comprises: obtaining the device running state from each edge node, and constructing the correlation state representation of the physical coupling between devices based on the device running state. According to the correlation state representation, the physical coupling relationship strength between devices is determined, and the device control partition is reconstructed based on the size of the physical coupling relationship strength between devices. The operation constraints of each device in the reconstructed device control partition are jointly represented to generate the joint control space corresponding to the device control partition. Each edge node is called to execute collaborative control actions on the devices in the corresponding device control partition according to the joint control space, and the device response state after the execution of the collaborative control actions is output. The application makes the devices in the same control partition have joint regulation and control, avoids the error division of strong coupling objects into independent decision units, improves the decision feasibility and eliminates the problem of physical layer transient strong coupling impact.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a collaborative optimization method for edge intelligent agents based on cloud-edge collaboration. Background Technology

[0002] In scenarios with strong physical coupling but fixed control partition logic, the partitioning assumption of the collaborative control strategy fails. Understandably, in large data center cooling systems and centralized industrial fluid control systems, there are strong physical coupling relationships between the physical objects corresponding to different edge agents (e.g., adjacent cooling units, adjacent air ducts, series and parallel pipelines). That is, the control actions of one device can directly affect other devices in a very short time through physical media (airflow, heat diffusion, pressure transmission).

[0003] Currently, in existing technical solutions, the local states and local feasible decision spaces under the jurisdiction of edge agents are relatively independent and separable in engineering. However, in this scenario, this separation does not hold at the physical level, making it difficult to achieve true feasibility in the local feasible decision space. Furthermore, collaborative correction can only be adjusted at the logical level, but it is difficult to eliminate the transient strong coupling impact at the physical layer, resulting in oscillating behavior where control actions are continuously "amplified-cancelled-re-amplified" in the physical system. Summary of the Invention

[0004] This application provides a cloud-edge collaborative optimization method for edge intelligent agents, which addresses the technical problems of poor decision-making feasibility and difficulty in eliminating transient strong coupling impacts at the physical layer in existing technologies.

[0005] The present invention adopts the following technical solution.

[0006] The first aspect of this invention discloses an edge agent collaborative optimization method based on cloud-edge collaboration, the method comprising: The device operating status from each edge node is obtained, and a correlation state representation of the physical coupling between devices is constructed based on the device operating status. The strength of the physical coupling relationship between devices is determined based on the associated state characterization, and the device control partition is reconstructed based on the magnitude of the physical coupling relationship between devices; The operational constraints of each device within the reconstructed device control partition are jointly characterized to generate the joint control space corresponding to the device control partition; Each edge node is invoked to perform collaborative control actions on the devices within the corresponding device control partition according to the joint control space, and the device response status after the collaborative control actions are executed is output.

[0007] Furthermore, the step of acquiring the device operating status from each edge node and constructing a relational state representation of the physical coupling between devices based on the device operating status includes: The control quantity and corresponding response quantity of each device are obtained according to a unified sampling period, and the response change of the response quantity at each time moment is calculated, so as to generate the device operating status based on the control quantity, response quantity and response change. The operating status of the equipment is time-aligned, and the instantaneous influence intensity coefficient of the change in the control quantity of the first equipment on the change in the response quantity of the second equipment at the same moment is calculated to determine the average influence intensity within the set window length. Wherein, the first device and the second device are any two different devices, and the average influence intensity is the result of averaging the instantaneous influence intensity coefficients at multiple moments within the set window length.

[0008] Furthermore, the step of acquiring the device operating status from each edge node and constructing a relational state representation of the physical coupling between devices based on the device operating status also includes: The equipment coupling strength coefficient is obtained by normalizing each device based on the average influence intensity. Based on the device coupling strength coefficient, the effective coupling strength between different devices and the coupling influence range of each device on other devices are determined, and the associated state characterization of physical coupling between devices is constructed according to the effective coupling strength and the coupling influence range.

[0009] The further step of determining the strength of physical coupling between devices based on the associated state characterization, and reconstructing the device control partition based on the magnitude of the strength of physical coupling between devices, includes: The effective coupling strength between different devices at the same time is read, and the relationship between the effective coupling strength and a set coupling threshold is compared. When the effective coupling strength is not less than the set coupling threshold, it is determined that there is a strong coupling relationship between the devices. Using each device as a node, strong coupling candidate edges are generated based on the strong coupling relationships between the devices to construct a strongly coupled connected cluster, and the merging benefit index of the strongly coupled connected cluster is calculated. When the merged revenue index is not less than the set merged revenue threshold, the strongly coupled connected cluster is written into the partition merge candidate set. The partition merge candidate set consists of multiple candidate connected clusters, and the candidate connected clusters are strongly coupled connected clusters whose merged revenue index is not less than the set merged revenue threshold.

[0010] Furthermore, the step of determining the strength of the physical coupling relationship between devices based on the associated state characterization, and reconstructing the device control partition based on the magnitude of the physical coupling relationship strength between devices, further includes: Obtain the initial control partition of each device in the candidate connected cluster, and mark the initial control partition as a strongly coupled sensing control partition when all devices in the candidate connected cluster are located in the same initial control partition. When the devices in the candidate connected cluster are located in different initial control partitions, the degree of merging between the different initial control partitions is calculated, and when the degree of merging is not less than a set merging threshold, the different initial control partitions are merged into a strongly coupled sensing control partition.

[0011] Furthermore, the joint characterization of the operational constraints of each device within the reconstructed device control partition to generate the joint control space corresponding to the device control partition includes: Read the operating constraints of each device in the reconstructed device control partition, and standardize the operating constraints to obtain the initial constraints of the device. The operating constraints are the upper and lower limits of the control quantity and the allowable rate of change of the control quantity for each device. Based on the strength of the physical coupling relationship between different devices within the device control zone, and in conjunction with the initial constraints of the devices, the coupling constraint weights between different devices are defined, and an upper bound is imposed on the coupling constraint weights. Based on the initial constraints and coupling constraint weights of the devices, a joint control interval for each device within the device control partition is defined, and the initial constraints of the devices are coupled, corrected, and verified according to the upper and lower limits of the joint control interval to generate the joint control space.

[0012] Furthermore, the step of invoking each edge node to perform coordinated control actions on the devices within the corresponding device control partition according to the joint control space, and outputting the device response status after the coordinated control actions are executed, includes: The edge node is invoked to read the joint control space corresponding to the device control partition, and the device collaborative control quantity is calculated based on the upper and lower limits of the joint control space; Based on the device collaborative control quantity, a partition disturbance intensity and a partition disturbance intensity upper limit are defined for the device control partition. When the partition disturbance intensity exceeds the product of the partition disturbance intensity upper limit and the disturbance ratio coefficient, the device collaborative control quantity is recalculated.

[0013] Furthermore, the step of invoking each edge node to perform collaborative control actions on the devices within the corresponding device control partition according to the joint control space, and outputting the device response status after the collaborative control actions are executed, also includes: Before executing the collaborative control action, the edge node is invoked to verify the collaborative control amount of the device, so that the collaborative control amount of the device is determined to pass the verification when the collaborative control amount of the device is not lower than the lower limit of the control amount and does not exceed the upper limit of the control amount; When the device collaborative control quantity fails the verification, the device collaborative control quantity is truncated according to the upper and lower limits of the control quantity; After the collaborative control action is executed, the response change index of the device response before and after the execution is calculated by the edge node, and when the response change index does not exceed the set response threshold, it is determined that the device control partition is in a stable response state.

[0014] The second aspect of this invention discloses an edge agent collaborative optimization device based on cloud-edge collaboration, used to implement the edge agent collaborative optimization method based on cloud-edge collaboration as described in any of the first aspects of this invention, the device comprising: The device coupling association module is used to obtain the device operating status from each edge node and construct an association status representation of the physical coupling between devices based on the device operating status. The control partition reconstruction module is used to determine the strength of the physical coupling relationship between devices based on the associated state characterization, and to reconstruct the device control partition based on the magnitude of the physical coupling relationship between devices; The control space generation module is used to jointly characterize the operating constraints of each device within the reconstructed device control partition in order to generate a joint control space corresponding to the device control partition. The collaborative control response module is used to call each edge node to perform collaborative control actions on the devices in the corresponding device control partition according to the joint control space, and output the device response status after the collaborative control actions are executed.

[0015] A third aspect of the present invention discloses a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method described in the first aspect.

[0016] A fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0017] Compared with the prior art, this application has the following advantages: (1) During system operation, the cloud constructs a system-related state expression reflecting the degree of physical mutual influence between devices based on the device operating status reported by each edge node. This system-related state expression no longer relies on fixed control partitions, but describes the degree of influence of any device's control action on the state changes of other devices, thus characterizing the physical coupling structure under the current operating conditions and forming a system-level related state that reflects the strength and scope of coupling. Subsequently, the control boundary of the edge agent is dynamically adjusted. When a strong physical coupling relationship is detected between multiple devices, the originally independent control partitions are merged or reconstructed into a coupling-aware control partition, so that the devices within the same control partition have the rationality of joint control in an engineering sense. This effectively avoids the erroneous division of strongly coupled objects into independent decision-making units, thereby solving the technical problems of poor decision-making feasibility and difficulty in eliminating transient strong coupling impacts at the physical layer in the prior art.

[0018] (2) After obtaining the coupled sensing control partition, the present invention reconstructs the original local control constraints within the corresponding partition by each edge node, and jointly describes the operating constraints of all devices within the partition to form a joint feasible control space that can simultaneously satisfy the physical constraints of the devices within the partition. This avoids the unexpected amplification effect across devices caused by the control actions of a single device in the physical system. In addition, during the control execution process, all control actions are limited to the joint feasible control space, thereby ensuring that the adjustment of any device will not cause unexpected disturbances to other devices through physical coupling. Ultimately, the system state can still converge smoothly in the strong physical coupling scenario, complete the collaborative optimization control, further improve the feasibility of decision-making and eliminate the transient strong coupling impact at the physical layer. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] Figure 1 This is a flowchart illustrating the edge agent collaborative optimization method based on cloud-edge collaboration provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the edge intelligent agent collaborative optimization device based on cloud-edge collaboration provided by the present invention.

[0022] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0024] like Figure 1 As shown, in one embodiment, an edge agent collaborative optimization method based on cloud-edge collaboration includes the following steps: Step S110: Obtain the device operating status from each edge node, and construct a relational state representation of the physical coupling between devices based on the device operating status.

[0025] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes the following steps in step S110: Step S111: Obtain the control quantity and corresponding response quantity of each device according to a unified sampling period, and calculate the response change of the response quantity at each time point to generate the device operating status based on the control quantity, response quantity and response change.

[0026] Step S112: Time-align the device operating status and calculate the instantaneous influence intensity coefficient of the change in the control quantity of the first device on the change in the response quantity of the second device at the same moment, so as to determine the average influence intensity within the set window length.

[0027] Wherein, the first device and the second device are any two different devices, and the average influence intensity is the result of averaging the instantaneous influence intensity coefficients at multiple moments within a set window length.

[0028] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention further includes the following steps in step S110: Step S113: Normalize each device according to the average influence intensity to obtain the device coupling strength coefficient.

[0029] Step S114: Based on the device coupling strength coefficient, determine the effective coupling strength between different devices and the coupling influence range of each device on other devices, and construct the association state characterization of physical coupling between devices according to the effective coupling strength and the coupling influence range.

[0030] In a specific embodiment, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes steps 1 to 4: Step 1: Construct a system-related state representation that reflects the strength of physical coupling.

[0031] During system operation, the cloud constructs a system-level correlation state representation reflecting the degree of physical interaction between devices based on the device operating status reported by each edge node. This system correlation state representation no longer relies on fixed control partitions but describes the impact of any device's control action on the state changes of other devices, characterizing the physical coupling structure under the current operating conditions, thus forming a system-level correlation state that reflects the strength and scope of coupling. This includes the following sub-steps: Sub-step 1.1: Equipment status acquisition and preprocessing.

[0032] Specifically, a uniform sampling period is set for each device, ranging from 0.5 seconds to 10 seconds. The control quantities and response quantities of each device at each moment are collected. The change in response quantity of each device at each moment is calculated, and this change in response quantity is equal to the difference in response quantity of a device at adjacent sampling moments. Response quantities include selected temperature, pressure, flow rate, and air volume, while control quantities include valve opening, fan speed, and cooling output setpoint. Then, the reported sequences of each device are aligned in the cloud according to a uniform time frame. If a device lacks data at a certain moment, the most recent valid data is used as a placeholder, ensuring that all devices form a comparable aligned sequence at the same time, i.e., a state aligned sequence under a unified time reference.

[0033] Sub-step 1.2: Calculate the transient influence intensity coefficient.

[0034] Specifically, for any two devices (devices) and equipment ), computing devices at the same time Changes in control quantity trigger equipment The intensity of transient effects of response changes. First, it is necessary to calculate the equipment. The change in the control quantity is equal to the difference between the control quantity at the current moment and the control quantity at the previous moment. Then, the influence intensity coefficient is constructed, expressed as: In the formula, For equipment At any moment The change in the control quantity; In response to changes; For equipment For equipment At any moment The transient influence intensity coefficient; To prevent constants with a denominator of zero, the value is taken as 0.001-0.05.

[0035] Subsequently, to avoid misjudgment caused by a single fluctuation, the transient influence intensity coefficient was averaged over a time window of length W (ranging from 3 to 30) to obtain the window-averaged influence intensity.

[0036] Sub-step 1.3: Calculate the coupling strength coefficient.

[0037] Specifically, the different units and amplitudes of control quantities for different devices result in the average influence intensity of the window not being directly comparable. Therefore, for each group of devices (equipment...), and equipment Normalization is performed to obtain the coupling strength coefficient between 0 and 1, expressed as: In the formula, For equipment For equipment At any moment The coupling strength coefficient; For equipment For equipment At any moment The average influence intensity of the window; The normalization scaling constant, ranging from 0.1 to 10, is used to control the degree of compression of the normalization curve; the closer the coupling strength coefficient is to 1, the stronger the coupling, and the closer it is to 0, the weaker the coupling.

[0038] Next, in order to limit the scope of "effective coupling", a coupling judgment threshold is set, with a value of 0.3-0.8. When the above coupling strength coefficient is less than the coupling judgment threshold, it is judged as weak coupling and set to zero, thereby obtaining the effective coupling strength.

[0039] Sub-step 1.4: Construct the system-related state representation.

[0040] Specifically, in the cloud, a system association state representation is constructed using devices as nodes and effective coupling strength as the correlation quantity. This system association state representation includes at least the effective coupling strength between device groups and a coupling influence range index for each device (used to describe the size of the coupling effect range, representing the overall coupling degree of a device to other devices). Then, the cloud packages all effective coupling strengths and coupling influence range indices to obtain a system association state representation that simultaneously expresses both "coupling strength" and "coupling coverage" at a given time.

[0041] Step S120: Determine the strength of physical coupling between devices based on the associated state characterization, and reconstruct the device control partition based on the magnitude of the strength of physical coupling between devices.

[0042] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes the following steps in step S120: Step S121: Read the effective coupling strength between different devices at the same time, and compare the effective coupling strength with the set coupling threshold. If the effective coupling strength is not less than the set coupling threshold, it is determined that there is a strong coupling relationship between the devices.

[0043] Step S122: Using each device as a node, generate strongly coupled candidate edges based on the strong coupling relationships between devices, construct a strongly coupled connected cluster, and calculate the merging benefit index of the strongly coupled connected cluster.

[0044] Step S123: When the merged revenue index is not less than the set merged revenue threshold, the strongly coupled connected clusters are written into the partition merge candidate set. The partition merge candidate set consists of multiple candidate connected clusters, and the candidate connected clusters are strongly coupled connected clusters whose merged revenue index is not less than the set merged revenue threshold.

[0045] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention further includes the following steps in step S120: Step S124: Obtain the initial control partition of each device in the candidate connected cluster, and mark the initial control partition as a strongly coupled sensing control partition when all devices in the candidate connected cluster are located in the same initial control partition.

[0046] Step S125: When each device in the candidate connected cluster is located in a different initial control partition, calculate the merging degree between the different initial control partitions, and merge the different initial control partitions into a strongly coupled sensing control partition when the merging degree is not less than the set merging threshold.

[0047] In a specific embodiment, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes step 2, dynamically generating coupled-aware control partitions based on system association states. Based on the system association state expression set output in step 1, the cloud dynamically adjusts the control boundary of the edge agent. When a strong physical coupling relationship is detected between multiple devices, the originally independent control partitions are merged or reconstructed into coupled-aware control partitions, making the devices within the same control partition reasonably operable in an engineering sense, thereby avoiding the erroneous division of strongly coupled objects into independent decision-making units. This includes the following sub-steps: Sub-step 2.1: Construct a set of strongly coupled candidate edges.

[0048] Specifically, the cloud can read data from any two devices at any given time. and equipment The effective coupling strength is determined, and a strong coupling threshold is set, ranging from 0.4 to 0.9. If the device... and equipment If the effective coupling strength is not less than the set strong coupling judgment threshold, then the judgment device... and equipment A strong coupling relationship exists, and a strong coupling candidate edge is generated based on this relationship. This strong coupling candidate edge is then added to the set of strong coupling candidate edges. Subsequently, to avoid frequent partition changes caused by occasional fluctuations, a "consecutive determination count" is introduced, with a value of 2-10. Only when the above strong coupling condition (effective coupling strength not less than the set strong coupling determination threshold) is met for a consecutive "consecutive determination count" of time points, is the strong coupling candidate edge designated as a valid candidate edge.

[0049] Sub-step 2.2: Generation of strongly coupled connected clusters.

[0050] Specifically, N devices are considered as a set of nodes, and strongly coupled candidate edges are considered as connections between nodes. Based on these candidate edges, the cloud platform forms several strongly coupled connected clusters. Each connected cluster represents a set of devices that have a strong mutual influence in engineering and should be considered "jointly controlled objects." To ensure the construction results of the connected clusters are quantifiable and verifiable, the intra-cluster coupling strength of a device within a cluster is used to determine whether it should remain in the current connected cluster. Simultaneously, the average coupling strength within the connected cluster is calculated to measure whether the cluster as a whole is sufficiently "tight." If the average intra-cluster coupling strength is lower than a set tightness threshold (ranging from 0.2 to 0.7), the cluster is considered to lack strong coupling consistency in an engineering sense and needs to be split or not merged.

[0051] Sub-step 2.3: Partition merging.

[0052] Specifically, in engineering, strong coupling does not necessarily require "unconditional merging," otherwise it would lead to excessively large partitions and overly coarse control boundaries. Therefore, the cloud calculates a merging benefit index for each connected cluster. This index measures the necessity of "merging for suppressing oscillations and avoiding erroneous splits." The merging benefit index comprehensively considers the average coupling strength within the cluster and the coupling influence range of devices within the cluster, and its expression is: In the formula, For connected clusters At any moment Consolidated earnings metrics; This is the weighting coefficient, ranging from 0.3 to 0.7; The average coupling strength within the cluster; For equipment The range of coupling influence; The cluster size.

[0053] Next, a merge benefit threshold is set, ranging from 0.3 to 0.8. When the merge benefit index is not less than the set merge benefit threshold, the current connected cluster is added to the partition merge candidate set, indicating that the cluster should be included in the same control partition as a whole or used as the core skeleton of partition reconstruction.

[0054] Sub-step 2.4: Partition merging and adjustment.

[0055] Specifically, the cloud-based system uses the original set of control partitions as a benchmark and performs partition reconstruction on each candidate connected cluster in the candidate set for partition merging. During the partition reconstruction process, if a device in a candidate connected cluster spans multiple original partitions, these original partitions are merged according to the coverage relationship of the candidate connected cluster. If a candidate connected cluster falls entirely within a single original partition, the partition remains unchanged, but the cluster is marked as a "strongly coupled joint control subset." To ensure the verifiability of the partition reconstruction, a partition merging degree is defined for any two partitions to quantify the cross-partition coupling strength between them, serving as a check metric for the merging decision. The expression is: In the formula, For two different control partitions; , These represent the number of devices within each of the two control partitions, ranging from 1 to N. For these two control partitions at time The average coupling strength; For equipment and equipment In time The effective coupling strength.

[0056] Next, a threshold for merging intervals is set, ranging from 0.2 to 0.8. If a candidate cluster spans both of the above control partitions... If the average coupling strength of the two partitions is not less than the set partition merging threshold, the two control partitions are merged to obtain a new partition. After merging, the verification is repeated until all candidate connected clusters are completely contained within a certain partition.

[0057] Step S130: Jointly characterize the operating constraints of each device within the reconstructed device control partition to generate a joint control space corresponding to the device control partition.

[0058] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes the following steps in step S130: Step S131: Read the operating constraints of each device in the reconstructed device control partition, and standardize the operating constraints to obtain the initial constraints of the device. The operating constraints are the upper and lower limits of the control quantity and the allowable rate of change of the control quantity for each device.

[0059] Step S132: Based on the strength of the physical coupling relationship between different devices within the device control zone and combined with the initial constraints of the devices, define the coupling constraint weights between different devices and impose an upper bound on the coupling constraint weights.

[0060] Step S133: Based on the initial constraints and coupling constraint weights of the equipment, define the joint control interval of each equipment within the equipment control partition, and perform coupling correction and verification on the initial constraints of the equipment according to the upper and lower limits of the joint control interval to generate the joint control space.

[0061] In a specific embodiment, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention, step 3, generates a joint feasible control space within the coupled-aware control partition. After obtaining the set of coupled-aware control partitions, each edge node reconstructs the original local control constraints within the corresponding partition. The operational constraints of all devices within the partition are jointly described to form a joint feasible control space that can simultaneously satisfy the physical constraints of the devices within the partition, avoiding unintended amplification effects across devices caused by single-device control actions in the physical system. This includes the following sub-steps: Sub-step 3.1: Define the original constraints of the device.

[0062] Specifically, for any coupled sensing control zone, the set and number of devices within that zone are determined. At a given moment, the cloud reads the operational constraints of each device within the zone and organizes them into a standardized format. These operational constraints include at least upper and lower limits for control quantities and allowable rates of change, describing the control range that the device can execute at the physical level. For example, the control quantity of a device may be greater than or equal to the minimum allowable control quantity under the current operating condition and less than or equal to the maximum allowable control quantity. The maximum and minimum control quantities are determined by the device's nameplate parameters, current operating status, and field configuration, with the value range depending on the device type. Simultaneously, to suppress transient overshoot, control rate of change constraints are defined for each device. For example, the absolute value of the difference between the control quantity at the current moment and the previous moment may be less than or equal to the device's maximum allowable control change amplitude (ranging from 1% to 20% of the full-scale control quantity). Finally, the cloud aggregates the above constraints for all devices within the control zone to form a zone-level original constraint set.

[0063] Sub-step 3.2 transforms the "coupling effect" into constraint weights.

[0064] Specifically, in scenarios with strong physical coupling, changes in the control quantity of a single device can affect other devices through physical media. Therefore, the "coupling effect" needs to be transformed into constraint weights and introduced into the calculation of the joint control boundary. For any device within a control zone... and equipment Define the coupling constraint weights between the two devices to describe the devices. Control actions on equipment The degree of constraint erosion is numerically equal to the effective coupling strength as a molecule-device interaction. The ratio of the coupling influence range index to the value of 1 is used to avoid amplifying extreme small values.

[0065] Subsequently, to ensure that the weights can be used for joint constraint calculations, an upper bound is applied to the aforementioned coupled constraint weights to obtain the final constraint weights. A larger weight value indicates a higher level of equipment... Control changes on equipment The stronger the impact.

[0066] Sub-step 3.3 generates a partition-level joint feasible control interval.

[0067] Specifically, in any control partition Internal to any device In the context of joint control, the control range of a single device needs to be "shrunken" to prevent its control actions from causing uncontrollable amplification to other devices through strong coupling. Therefore, it is necessary to define the device... Within the joint feasible control interval of the partition, the upper and lower limits of this interval (upper limit: Lower limit: The expression is obtained by coupling and correcting the original operational constraints: In the formula, , respectively equipment The lower and upper limits of the joint feasible control interval within the partition; , Representing the equipment and equipment The maximum permissible control change within a control cycle; , Representing the equipment For equipment and equipment For equipment The coupling weights. Furthermore, in the equation... , All of them need to be normalized.

[0068] The above formula describes the equipment. The adjustable margin needs to reserve a "buffer margin" for the potential coupling effects of other devices, thereby avoiding system-level oscillations caused by multiple devices simultaneously adjusting to their limits. Finally, for partitioning... All devices within the system perform the above calculations to obtain a set of jointly feasible control intervals at the partition level.

[0069] Sub-step 3.4, joint feasible control space verification.

[0070] Specifically, to avoid excessive contraction of joint constraints leading to control space degradation, the cloud verifies the aforementioned set of partition-level joint feasible control intervals. For any partition in the set, its partition feasibility index is defined, expressed as: In the formula, For partitioning At any moment The partition feasibility index is used to determine the joint constraints of a partition. When the partition feasibility index is less than the preset minimum feasible threshold (ranging from 5% to 30% of the total width of the original control space of the partition), the joint constraints of the partition are considered too tight, and the partition is reverted to a conservative joint mode, retaining only the top few coupling constraints with the largest weights until the feasibility requirements are met. When the partition feasibility index is greater than or equal to the preset minimum feasible threshold, the set of joint feasible control intervals at the partition level is confirmed as an effective joint feasible control space, and the confirmed joint feasible control space is finally formed.

[0071] Step S140: Invoke each edge node to perform collaborative control actions on the devices in the corresponding device control partition according to the joint control space, and output the device response status after the collaborative control actions are executed.

[0072] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes the following steps in step S140: Step S141: Call the edge node to read the joint control space corresponding to the device control partition, and calculate the device collaborative control quantity based on the upper and lower limits of the joint control space.

[0073] Step S142: Based on the equipment cooperative control quantity, define the partition disturbance intensity and the upper limit of the partition disturbance intensity for the equipment control partition, and recalculate the equipment cooperative control quantity when the partition disturbance intensity exceeds the product of the upper limit of the partition disturbance intensity and the disturbance ratio coefficient.

[0074] In some embodiments, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention further includes the following steps in step S140: Step S143: Before executing the collaborative control action, the edge node is called to verify the collaborative control quantity of the device. If the collaborative control quantity of the device is not lower than the lower limit of the control quantity and does not exceed the upper limit of the control quantity, the collaborative control quantity of the device is determined to pass the verification.

[0075] Step S144: When the equipment collaborative control quantity fails the verification, the equipment collaborative control quantity is truncated according to the upper and lower limits of the control quantity.

[0076] Step S145: After the collaborative control action is executed, the response change index of the device response quantity before and after the execution is calculated through the edge node, and when the response change index does not exceed the set response threshold, it is determined that the device control zone is in a stable response state.

[0077] In a specific embodiment, the edge agent collaborative optimization method for cloud-edge collaboration provided by the present invention includes step 4, which involves executing collaborative control based on a joint feasible control space and stabilizing the system response. Based on the joint feasible control space set output in step 3, edge nodes execute collaborative control actions on devices within a partition. During the control execution process, all control actions are confined within the joint feasible control space, thereby ensuring that the adjustment of any device will not cause unexpected disturbances to other devices through physical coupling, ultimately enabling the system state to converge smoothly even in strongly physically coupled scenarios, thus completing the collaborative optimization control. This includes the following sub-steps: Sub-step 4.1: Construct the initial cooperative control variables for devices within the partition.

[0078] Specifically, for any coupled sensing control partition, the edge node reads the joint feasible control space corresponding to that partition at any given time. For each device within the partition, the joint feasible control space provides its upper and lower control limits. Based on this, the edge node generates an initial cooperative control quantity according to the "minimum disturbance principle," minimizing the control adjustment amplitude to avoid triggering potential coupled oscillations. Therefore, the initial control quantity is equal to the sum of the control adjustment quantity and the control quantity at the previous time, and the sum of the control adjustment quantity and the control quantity at the previous time is greater than or equal to the lower limit of the control interval and less than or equal to the upper limit of the control interval. Simultaneously, to limit the amplitude of single-cycle disturbances, a variation amplitude limit is introduced, meaning the absolute value of the control adjustment quantity is less than or equal to the maximum allowable control variation amplitude of the device.

[0079] Sub-step 4.2, constrain consistency correction control quantity.

[0080] Specifically, due to the strong coupling between devices within a partition, even if each initial coordinated control variable individually satisfies the joint feasible control interval, simultaneous adjustment by multiple devices may still generate superimposed disturbances during the transient phase. Therefore, partition-level consistency correction is introduced. For any partition, a partition disturbance intensity is defined to assess the overall disturbance level of the current control combination, expressed as: In the formula, For equipment At any moment The initial collaborative control quantity; For equipment The control quantity at the previous moment; For partitioning The intensity of the partition perturbation at time k.

[0081] Simultaneously, an upper limit for permissible disturbances within a partition is defined, determined by the combined maximum permissible control variation of all devices within the partition. When the disturbance intensity of a partition is less than or equal to the product of the upper limit for permissible disturbances and the disturbance proportionality coefficient (ranging from 0.4 to 0.9), the current control combination is considered to meet the consistency requirements; otherwise, the control adjustment amounts of all devices are scaled proportionally to obtain the corrected control adjustment amounts, and the corrected control amounts are recalculated, ultimately forming a set of constraint consistency corrected control amounts within the partition.

[0082] Sub-step 4.3: Joint feasibility verification and generation of final execution control quantities.

[0083] Specifically, before execution, the edge nodes perform a joint feasibility check on the set of control quantities for constraint consistency correction to ensure that the correction process does not cause any device to go out of bounds. For any device, if its coordinated control quantity is greater than or equal to the lower limit of the control interval and less than or equal to the upper limit of the control interval, the check is considered passed; otherwise, the check is considered failed. For control quantities that fail the check, the corresponding control quantities are truncated to the nearest boundary value according to the upper and lower limits.

[0084] Sub-step 4.4: Control action execution and response result output.

[0085] Specifically, after the control command is executed, the edge node will perform the following sampling at the next sampling time. Data acquisition equipment Device response And calculate the partition-level response change index. The expression is: In the formula, respectively equipment The response amount at adjacent time points.

[0086] Simultaneously, a partition response stability threshold is set, the value of which is determined by engineering experience and is typically the sum of 5% to 25% of the equipment's rated response variation. When the partition-level response variation index is less than or equal to the set partition response stability threshold, the partition response is determined to be stable, and the current state is recorded as a stable response result; otherwise, the partition is marked as being in an unstable state, and in the next control cycle, the process returns to step 3.

[0087] The edge agent collaborative optimization device based on cloud-edge collaboration provided by the present invention is described below. The edge agent collaborative optimization device based on cloud-edge collaboration described below and the edge agent collaborative optimization method based on cloud-edge collaboration described above can be referred to in correspondence.

[0088] like Figure 2As shown, in one embodiment, an edge intelligent agent collaborative optimization device based on cloud-edge collaboration includes a device coupling association module, a control partition reconstruction module, a control space generation module, and a collaborative control response module.

[0089] The device coupling association module is used to obtain the device operating status from each edge node and construct a physical coupling association status representation between devices based on the device operating status.

[0090] The control partition reconstruction module is used to determine the strength of physical coupling between devices based on the associated state characterization, and to reconstruct the device control partition based on the magnitude of the physical coupling strength between devices.

[0091] The control space generation module is used to jointly characterize the operating constraints of each device within the reconstructed device control partition in order to generate the joint control space corresponding to the device control partition.

[0092] The collaborative control response module is used to call each edge node to perform collaborative control actions on the devices in the corresponding device control partition according to the joint control space, and output the device response status after the collaborative control actions are executed.

[0093] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0094] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0095] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0096] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0097] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0098] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0099] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A collaborative optimization method for edge intelligent agents based on cloud-edge collaboration, characterized in that, The method includes: The device operating status from each edge node is obtained, and a correlation state representation of the physical coupling between devices is constructed based on the device operating status. The strength of the physical coupling relationship between devices is determined based on the associated state characterization, and the device control partition is reconstructed based on the magnitude of the physical coupling relationship between devices; The operational constraints of each device within the reconstructed device control partition are jointly characterized to generate the joint control space corresponding to the device control partition; Each edge node is invoked to perform coordinated control actions on the devices within the corresponding device control partitions according to the joint control space, and the device response status after the coordinated control actions are executed is output; specifically including: The edge node is invoked to read the joint control space corresponding to the device control partition, and the device collaborative control quantity is calculated based on the upper and lower limits of the joint control space; Based on the device cooperative control quantity, a partition disturbance intensity and a partition disturbance intensity upper limit are defined for the device control partition. When the partition disturbance intensity exceeds the product of the partition disturbance intensity upper limit and the disturbance ratio coefficient, the device cooperative control quantity is recalculated. Before executing the collaborative control action, the edge node is invoked to verify the collaborative control amount of the device, so that the collaborative control amount of the device is determined to pass the verification when the collaborative control amount of the device is not lower than the lower limit of the control amount and does not exceed the upper limit of the control amount; When the device collaborative control quantity fails the verification, the device collaborative control quantity is truncated according to the upper and lower limits of the control quantity; After the collaborative control action is executed, the response change index of the device response before and after the execution is calculated by the edge node, and when the response change index does not exceed the set response threshold, it is determined that the device control partition is in a stable response state.

2. The edge agent collaborative optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The process of acquiring the device operating status from each edge node and constructing a relational state representation of the physical coupling between devices based on the device operating status includes: The control quantity and corresponding response quantity of each device are obtained according to a unified sampling period, and the response change of the response quantity at each time moment is calculated, so as to generate the device operating status based on the control quantity, response quantity and response change. The operating status of the equipment is time-aligned, and the instantaneous influence intensity coefficient of the change in the control quantity of the first equipment on the change in the response quantity of the second equipment at the same moment is calculated to determine the average influence intensity within the set window length. Wherein, the first device and the second device are any two different devices, and the average influence intensity is the result of averaging the instantaneous influence intensity coefficients at multiple moments within the set window length.

3. The edge agent collaborative optimization method based on cloud-edge collaboration according to claim 2, characterized in that, The step of acquiring the device operating status from each edge node and constructing a relational state representation of the physical coupling between devices based on the device operating status further includes: The equipment coupling strength coefficient is obtained by normalizing each device based on the average influence intensity. Based on the device coupling strength coefficient, the effective coupling strength between different devices and the coupling influence range of each device on other devices are determined, and the associated state characterization of physical coupling between devices is constructed according to the effective coupling strength and the coupling influence range.

4. The edge agent collaborative optimization method based on cloud-edge collaboration according to claim 3, characterized in that, The step of determining the strength of physical coupling between devices based on the associated state characterization, and reconstructing the device control partition based on the magnitude of the strength of physical coupling between devices, includes: The effective coupling strength between different devices at the same time is read, and the relationship between the effective coupling strength and a set coupling threshold is compared. When the effective coupling strength is not less than the set coupling threshold, it is determined that there is a strong coupling relationship between the devices. Using each device as a node, strong coupling candidate edges are generated based on the strong coupling relationships between the devices to construct a strong coupling connected cluster, and the merging benefit index of the strong coupling connected cluster is calculated. When the merged revenue index is not less than the set merged revenue threshold, the strongly coupled connected cluster is written into the partition merge candidate set. The partition merge candidate set consists of multiple candidate connected clusters, and the candidate connected clusters are strongly coupled connected clusters whose merged revenue index is not less than the set merged revenue threshold.

5. The edge agent collaborative optimization method based on cloud-edge collaboration according to claim 4, characterized in that, The step of determining the strength of physical coupling between devices based on the associated state characterization, and reconstructing the device control partition based on the magnitude of the strength of physical coupling between devices, further includes: Obtain the initial control partition of each device in the candidate connected cluster, and mark the initial control partition as a strongly coupled sensing control partition when all devices in the candidate connected cluster are located in the same initial control partition. When the devices in the candidate connected cluster are located in different initial control partitions, the degree of merging between the different initial control partitions is calculated, and when the degree of merging is not less than a set merging threshold, the different initial control partitions are merged into a strongly coupled sensing control partition.

6. The edge agent collaborative optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The process of jointly representing the operational constraints of each device within the reconstructed device control partition to generate a joint control space corresponding to the device control partition includes: Read the operating constraints of each device in the reconstructed device control partition, and standardize the operating constraints to obtain the initial constraints of the device. The operating constraints are the upper and lower limits of the control quantity and the allowable rate of change of the control quantity for each device. Based on the strength of the physical coupling relationship between different devices within the device control zone, and in conjunction with the initial constraints of the devices, the coupling constraint weights between different devices are defined, and an upper bound is imposed on the coupling constraint weights. Based on the initial constraints and coupling constraint weights of the devices, a joint control interval for each device within the device control partition is defined, and the initial constraints of the devices are coupled, corrected, and verified according to the upper and lower limits of the joint control interval to generate the joint control space.

7. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-6.

Images