Machine room air conditioner alarm cooperative processing method and system for equipment group control

Abstract

The invention provides a machine room air conditioner alarm cooperative processing method and system oriented to equipment group control, and relates to the technical field of control, and the method comprises the steps: obtaining a plurality of pieces of air conditioner alarm information, carrying out the spatial correlation analysis based on an equipment topological relation, mapping the alarm information to a group control cooperative graph structure comprising equipment nodes, an alarm propagation edge and a control dependence edge, and a reverse path tracing algorithm is adopted to identify a root cause device node, and a priority processing instruction is generated. The alarm root cause can be accurately positioned, the false alarm rate is reduced, and the operation efficiency and the energy utilization rate of the air conditioner system are improved.

Description

Technical Field

[0001] This invention relates to the field of control technology, and in particular to a collaborative processing method and system for alarms of computer room air conditioning for equipment group control. Background Technology

[0002] Air conditioning equipment in data center server rooms is a critical infrastructure for ensuring the normal operation of information equipment, and its stable operation directly affects the reliability and energy efficiency of the data center. As data centers continue to expand, multiple air conditioning units are often deployed within a single server room, working collaboratively to create complex airflow organization and temperature field distribution. These units interact with each other through supply and return air systems, forming a tightly coupled spatial network. In actual operation, server room air conditioning systems frequently generate various alarms, such as temperature exceeding limits, abnormal airflow, and pressure fluctuations. These alarms are crucial for the timely detection and resolution of equipment failures.

[0003] Currently, most data center air conditioning alarm handling systems primarily adopt an independent processing mode, treating each air conditioning unit as an independent entity for monitoring and alarm processing. This approach typically triggers alarms based on fixed thresholds and then processes them according to alarm level or time sequence.

[0004] Existing technologies lack in-depth analysis of the spatial correlation between multiple air conditioning devices in a computer room, and cannot effectively identify the fault propagation and alarm cascading effects generated by airflow paths between devices. As a result, when dealing with alarms, we often only focus on the symptoms and ignore the root cause. Summary of the Invention

[0005] This invention provides a collaborative processing method and system for air conditioning alarms in computer rooms for group control of equipment, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a collaborative processing method for alarms in a computer room air conditioning system for group control of equipment, comprising:

[0007] The alarm information generated by multiple air conditioning devices in the computer room within the current time window is obtained. Based on the preset device topology relationship, the spatial correlation analysis of the alarm information is performed to obtain an alarm set with spatial coupling characteristics. The device topology relationship includes the supply air path dependency relationship and return air path dependency relationship between devices.

[0008] Based on the device identifier in the alarm information and the spatial coupling feature, the alarm information is mapped to a pre-constructed device group control collaborative graph structure. The device group control collaborative graph structure includes device nodes, alarm propagation edges, and control dependency edges between devices. The weight of the alarm propagation edge is determined by the airflow influence degree between devices and the alarm timing correlation degree.

[0009] Based on the device group control cooperative graph structure, a reverse path tracing algorithm is used to identify a root cause device node causing the alarm set to be generated, and a priority processing instruction for the root cause device node is generated according to the root cause device node and the control dependency edge associated with the root cause device node, wherein the priority processing instruction includes a device control parameter adjustment action and an alarm suppression strategy; and the priority processing instruction is issued to the air conditioning device corresponding to the root cause device node.

[0010] Based on the preset device topology relationship, spatial correlation analysis is performed on the alarm information to obtain an alarm set with spatial coupling characteristics, including:

[0011] Based on the device identifier in the alarm information, the spatial position coordinates and the air flow direction vector of the device corresponding to each alarm information are extracted from the device topology relationship, and the alarm time stamp is combined to calculate the alarm time sequence difference between adjacent devices, thereby obtaining an alarm space-time feature matrix containing spatial distance characteristics and time sequence interval characteristics;

[0012] According to the air supply path dependence relationship and the return air path dependence relationship, the alarm space-time feature matrix is matched with the air flow conduction path, and the spatial coupling degree between devices on the same air flow conduction path is calculated, wherein the spatial coupling degree is determined by the connectivity of the air flow conduction path and the spatial distance characteristics;

[0013] Alarm information with a spatial coupling degree exceeding a preset coupling threshold is grouped into the same alarm set, thereby obtaining the alarm set with spatial coupling characteristics.

[0014] According to the air supply path dependence relationship and the return air path dependence relationship, the alarm space-time feature matrix is matched with the air flow conduction path, and the spatial coupling degree between devices on the same air flow conduction path is calculated, including:

[0015] Based on the air supply path dependence relationship and the return air path dependence relationship, a directed graph of the air flow conduction path is constructed, wherein the nodes of the directed graph of the air flow conduction path correspond to the device identifiers in the alarm space-time feature matrix, the directed edges of the directed graph of the air flow conduction path represent the air flow conduction relationship in the air supply direction or the return air direction, and the edge weight of the directed edge is determined by the consistency degree of the air flow direction vector and the reciprocal of the spatial distance characteristics;

[0016] For each pair of alarm information, connectivity detection is performed in the directed graph of the air flow conduction path to determine whether there is a directed connected path between the device nodes corresponding to the two alarm information;

[0017] If there is, the edge weight values of all directed edges on the directed connected path are accumulated and attenuated to obtain the spatial coupling degree of the device node pair, and the accumulated attenuation calculation is realized by weighted summation of the edge weight values of each directed edge on the path and application of a path length attenuation factor, and the path length attenuation factor decreases with the increase of the number of nodes contained in the directed connected path.

[0018] Based on the device group control collaborative graph structure, a reverse path tracing algorithm is used to identify the root cause device node that causes the alarm set to include:

[0019] All device nodes in the device group control collaborative graph structure are traversed, and according to the alarm information in the alarm set, the device nodes that generate alarms are marked as alarm nodes and are assigned an initial alarm propagation contribution degree, and the device nodes that do not generate alarms are marked as candidate root cause nodes.

[0020] Iterative traversal is performed along the reverse direction of the control dependency edge in the device group control collaborative graph structure, for each alarm node, the upstream device nodes pointed to by the alarm node through the control dependency edge are identified, and the alarm propagation contribution degree of the alarm node is reversely transmitted to the upstream device nodes according to the weight of the alarm propagation edge, the reverse transmission is realized by combining the alarm propagation contribution degree with the attenuation coefficient of the alarm propagation edge weight, and the cumulative alarm propagation contribution degree of each upstream device node is obtained.

[0021] During the iterative traversal process, only the propagation path that satisfies the directionality constraint of the control dependency edge is retained, and the directionality constraint requires that the direction of the control dependency edge on the propagation path is consistent with the causal direction of the alarm propagation.

[0022] When the iterative traversal terminates, the device node with the largest cumulative alarm propagation contribution degree is selected from the candidate root cause nodes as the root cause device node.

[0023] The reverse transmission of the alarm propagation contribution degree of the alarm node to the upstream device nodes according to the weight of the alarm propagation edge includes:

[0024] The weight of the alarm propagation edge is decomposed into an airflow influence degree component and an alarm timing correlation degree component, the airflow influence degree component represents the airflow conduction influence intensity of the upstream device node on the alarm node, and the alarm timing correlation degree component represents the timing correlation intensity between the change of the running state of the upstream device node and the alarm occurrence time of the alarm node.

[0025] The airflow influence degree component and the alarm timing correlation degree component are nonlinearly fused to obtain a comprehensive transmission coefficient of the incident alarm propagation edge, by performing product operation on the airflow influence degree component and the alarm timing correlation degree component and applying normalization processing;

[0026] The alarm propagation contribution degree of the alarm node is multiplied by the comprehensive transmission coefficient to obtain a contribution degree increment transmitted to the corresponding upstream device node through the incident alarm propagation edge;

[0027] The contribution degree increment is added to the current cumulative alarm propagation contribution degree of the corresponding upstream device node to complete the update of the cumulative alarm propagation contribution degree of the upstream device node.

[0028] Generating a priority processing instruction for the root cause device node according to the root cause device node and its associated control dependency edge includes:

[0029] Starting from the root cause device node, breadth-first traversal is performed along the forward direction of the control dependency edge in the device group control collaboration graph structure to identify all downstream device nodes directly or indirectly controlled by the root cause device node, and an impact domain topology subgraph containing the root cause device node and all downstream device nodes is constructed;

[0030] For each downstream device node in the impact domain topology subgraph, the corresponding regulation and control constraint condition of the downstream device node is extracted from the edge attribute of the control dependency edge;

[0031] The regulation and control constraint condition includes the running parameter adjustment amplitude allowed by the downstream device node and the state switching timing requirement of the downstream device node, the running parameter adjustment amplitude is determined by the control strength component in the control dependency edge weight, and the state switching timing requirement is determined by the response time delay component in the control dependency edge weight;

[0032] According to the topological level relationship of the impact domain topology subgraph, a main regulation and control instruction is generated for the root cause device node and a slave regulation and control instruction is generated for each downstream device node in the control conduction order from the root cause device node to the downstream device node.

[0033] The method further includes:

[0034] The running parameter adjustment amount in the main regulation and control instruction and the collaborative adjustment amount in the slave regulation and control instruction are substituted into the regulation and control constraint condition for feasibility verification,

[0035] When it is detected that the adjustment amount exceeds the operating parameter adjustment range or violates the state switching timing requirement, the over-limit adjustment amount is reduced and corrected and redistributed to other device nodes in the influence domain topology subgraph, to obtain optimized master control instructions and optimized slave control instructions that satisfy the control constraints, and the optimized master control instructions are output as the priority processing instructions.

[0036] In a second aspect of the embodiment of the present application, a device group control-oriented machine room air conditioner alarm cooperative processing system is provided, comprising:

[0037] A first unit is configured to acquire alarm information generated by multiple air conditioner devices in a machine room within a current time window, perform spatial correlation analysis on the alarm information based on a preset device topology relationship, and obtain an alarm set with spatial coupling characteristics, wherein the device topology relationship includes air supply path dependency relationships and return air path dependency relationships between devices;

[0038] A second unit is configured to map the alarm information to a pre-constructed device group control cooperative graph structure according to device identifiers in the alarm information and the spatial coupling characteristics, wherein the device group control cooperative graph structure includes device nodes, alarm propagation edges, and control dependency edges between devices, and a weight of the alarm propagation edge is determined by air flow influence degrees between devices and alarm timing correlation degrees.

[0039] A third unit is configured to identify root cause device nodes causing the alarm set based on the device group control cooperative graph structure by using a reverse path tracing algorithm, generate priority processing instructions for the root cause device nodes according to the root cause device nodes and associated control dependency edges, wherein the priority processing instructions include device control parameter adjustment actions and alarm suppression strategies, and issue the priority processing instructions to air conditioner devices corresponding to the root cause device nodes.

[0040] In a third aspect of the embodiment of the present application,

[0041] An electronic device is provided, comprising:

[0042] a processor;

[0043] a memory for storing processor-executable instructions;

[0044] The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0045] In a fourth aspect of the embodiment of the present application,

[0046] A computer-readable storage medium is provided, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method described above.

[0047] The beneficial effects of the present application are as follows:

[0048] The present application realizes intelligent collaborative processing of machine room air conditioner alarm, effectively avoids alarm storm, and improves fault processing efficiency by spatial correlation analysis based on device topology relationship and device group control collaborative graph structure.

[0049] The present application introduces a topology analysis method considering the dependence relationship of air supply path and return air path, and combines an alarm propagation edge weight calculation mechanism to accurately identify the root cause device node, solve the multi-device cascade alarm problem from the source, and reduce the work burden of operation and maintenance personnel.

[0050] The priority processing instruction provided by the present application includes device control parameter adjustment and alarm suppression strategy, realizes intelligent hierarchical processing of alarm, reduces the number of invalid alarms, and improves the stability of the machine room environment and the operation efficiency of the air conditioning system through automatic control means. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 A flowchart of the machine room air conditioner alarm collaborative processing method for device group control of the present application embodiment is shown.

[0052] Figure 2 A flowchart of the reverse path tracing algorithm for identifying the root cause device node causing the alarm set is shown. DETAILED DESCRIPTION

[0053] To make the purpose, technical scheme and advantages of the present application embodiment clearer, the technical scheme of the present application embodiment will be described clearly and completely in combination with the drawings of the present application embodiment. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0054] The technical scheme of the present application will be described in detail in specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in some embodiments.

[0055] Reference Figure 1 and Figure 2 The machine room air conditioner alarm collaborative processing method for device group control of the present application embodiment comprises:

[0056] Obtain alarm information generated by multiple air conditioning devices in a machine room within a current time window, perform spatial correlation analysis on the alarm information based on a preset device topology relationship, obtain an alarm set with a spatial coupling characteristic, and the device topology relationship includes a supply air path dependency relationship and a return air path dependency relationship between devices.

[0057] Map the alarm information to a pre-constructed device group control collaboration graph structure according to the device identifier in the alarm information and the spatial coupling characteristic, the device group control collaboration graph structure includes a device node, an alarm propagation edge, and a control dependency edge between devices, wherein the weight of the alarm propagation edge is determined by the air flow influence degree between devices and the alarm time sequence correlation degree.

[0058] Based on the device group control collaboration graph structure, a reverse path tracing algorithm is used to identify the root cause device node that causes the alarm set, and a priority processing instruction is generated for the root cause device node according to the root cause device node and its associated control dependency edge, the priority processing instruction includes a device control parameter adjustment action and an alarm suppression strategy; the priority processing instruction is issued to the air conditioning device corresponding to the root cause device node.

[0059] In an optional implementation, the spatial correlation analysis on the alarm information based on the preset device topology relationship to obtain the alarm set with the spatial coupling characteristic includes:

[0060] Based on the device identifier in the alarm information, the spatial position coordinates and the air flow direction vector of the device corresponding to each alarm information are extracted from the device topology relationship, and the alarm time sequence difference between adjacent devices is calculated by combining the alarm time stamp, to obtain an alarm space-time feature matrix including spatial distance features and time interval features;

[0061] According to the supply air path dependency relationship and the return air path dependency relationship, the air flow conduction path matching of each alarm information in the alarm space-time feature matrix is performed, and the spatial coupling degree between devices on the same air flow conduction path is calculated, the spatial coupling degree is determined by the connectivity of the air flow conduction path and the spatial distance feature;

[0062] Alarm information with a spatial coupling degree exceeding a preset coupling threshold is classified into the same alarm set to obtain the alarm set with the spatial coupling characteristic.

[0063] The device topology relationship is a digital representation of the spatial position and air flow conduction relationship of each device in the data center cooling system, which includes the identifier, three-dimensional coordinate position, air flow direction vector, and supply air and return air path dependency relationship of each device. The alarm information at least includes the device identifier, alarm type, alarm level and alarm time stamp.

[0064] In the spatial correlation analysis phase, all alarm information triggered within the last 24 hours is extracted from the alarm log database. For each alarm record, the spatial location information of the corresponding device is queried from the device topology relationship database through the device identifier. Specifically, the X, Y, Z coordinate values of each alarm device and the main airflow direction vector of the device are extracted. Assuming that a data center has a total of 15 alarm records within a certain time period, which are respectively from air conditioning devices numbered CRAC-01 to CRAC-05 and server cabinets numbered RACK-01 to RACK-10.

[0065] For each pair of devices with alarms, the spatial distance feature between them is calculated, which uses the three-dimensional Euclidean distance formula, that is, the straight-line distance between two points. For example, the coordinates of CRAC-01 are (10, 5, 0), and the coordinates of RACK-03 are (12, 8, 0), so the spatial distance between them is 5 meters. At the same time, the time stamp of each alarm is extracted, and the time interval between any two alarms is calculated in seconds. For example, the alarm time of CRAC-01 is 2023-07-15 14:30:25, and the alarm time of RACK-03 is 2023-07-15 14:33:42, so the time interval is 197 seconds.

[0066] The spatial location coordinates, airflow direction, and time interval between alarms of all alarms are organized into an alarm space-time feature matrix. This matrix is a multi-dimensional data structure, where each row represents a pair of alarm combinations, and the columns contain device identifiers, spatial distance values, time interval values, and other related attributes.

[0067] According to the pre-defined air supply path dependency relationship and return air path dependency relationship, the airflow conduction path matching analysis is performed on each alarm device in the alarm space-time feature matrix. The air supply path dependency relationship describes the path network of cold air flowing from the air conditioning device to the server cabinet, and the return air path dependency relationship describes the path network of hot air returning from the server cabinet to the air conditioning device.

[0068] For any two alarm devices, it is checked whether they are on the same airflow conduction path, and the specific implementation method is to determine whether there is airflow connectivity between the two devices through the path query algorithm in the graph data structure. For example, the cold air of CRAC-01 is directly sent to RACK-01, RACK-02, and RACK-03, while RACK-04 receives cold air from CRAC-02, so CRAC-01 and RACK-03 are on the same airflow path, while CRAC-01 and RACK-04 are not on the same airflow path.

[0069] For the pair of devices located on the same airflow path, further calculate the spatial coupling degree between them. The spatial coupling degree is a comprehensive index reflecting the degree of association of two devices in space and airflow conduction. In specific calculation, the connectivity strength of the airflow conduction path is determined, which is related to the number of devices on the path and the connection complexity. For example, the connectivity strength of directly connected devices is 1.0, the connectivity strength of devices connected through an intermediate device is 0.8, the connectivity strength of devices connected through two intermediate devices is 0.6, and so on.

[0070] Subsequently, the connectivity strength is combined with the spatial distance feature between devices to calculate the final spatial coupling degree. Specifically, the spatial coupling degree is equal to the connectivity strength multiplied by the decreasing function of the spatial distance. When the distance between two devices is 0, the value of the decreasing function is 1; as the distance increases, the function value gradually decreases. For example, the connectivity strength of CRAC-01 and RACK-03 is 1.0, the spatial distance is 5 meters, and the corresponding decreasing function value is 0.7, and the final spatial coupling degree is 0.7.

[0071] The system presets the spatial coupling threshold value as 0.5. When the spatial coupling degree of two devices exceeds the threshold value, the alarms of the two devices are classified into the same alarm set. In this way, the alarm groups with causal relationship due to airflow conduction relationship can be identified.

[0072] Through the above implementation process, alarm sets with spatial coupling characteristics can be extracted from a large number of dispersed alarm information. These alarm sets represent a group of devices that are closely related in physical space and airflow conduction, providing an important basis for subsequent root cause analysis. For example, in the example scenario, the alarms of CRAC-01, RACK-01, RACK-02 and RACK-03 are identified to form an alarm set, indicating that these devices have a common fault cause or fault propagation relationship.

[0073] In an alternative embodiment, according to the air supply path dependence relationship and the return air path dependence relationship, the airflow conduction path matching is performed on each alarm information in the alarm space-time feature matrix, and the spatial coupling degree between devices on the same airflow conduction path is calculated.

[0074] Based on the air supply path dependence relationship and the return air path dependence relationship, a airflow conduction path directed graph is constructed, the nodes of the airflow conduction path directed graph correspond to the device identifiers in the alarm space-time feature matrix, the directed edges of the airflow conduction path directed graph represent the airflow conduction relationship in the air supply direction or the return air direction, and the edge weight of the directed edge is determined by the consistency degree of the airflow direction vector and the reciprocal of the spatial distance feature.

[0075] For each pair of alarm information, connectivity detection is performed in the airflow conduction path directed graph to determine whether there is a directed connected path between the device nodes corresponding to the two alarm information.

[0076] If there is, the edge weight values of all directed edges on the directed connected path are accumulated and attenuated to obtain the spatial coupling degree of the device node pair. The accumulated and attenuated calculation is realized by weighted summation of the edge weight values of the directed edges on the path and application of a path length attenuation factor, which decreases with the increase of the number of nodes included in the directed connected path.

[0077] Based on the obtained air supply path dependence and air return path dependence, an airflow conduction path directed graph is constructed. In the directed graph, the nodes correspond to the device identifiers appearing in the alarm spatio-temporal feature matrix, such as air conditioner device A and temperature sensor B. The directed edges represent the direction relationship of airflow conduction, including air supply direction or air return direction. For example, the directed edge from air conditioner device A to temperature sensor B indicates that the airflow of air conditioner A will affect the environment of sensor B.

[0078] The edge weight calculation of the directed edge combines the consistency degree of the airflow direction vector and the reciprocal of the spatial distance feature. Specifically, for the directed edge connecting device i and device j, its edge weight can be determined in the following way: first, calculate the cosine value of the angle between the air supply direction vector of device i and the spatial position vector from device i to device j. The closer this value is to 1, the more consistent the airflow direction is with the line direction of the two devices. Then calculate the reciprocal of the spatial distance between the two devices. The closer the distance, the larger the value. Finally, combine the two feature values by weighting to obtain the edge weight. For example, if the air supply direction of air conditioner A and the line direction from air conditioner A to sensor B form an angle of 30 degrees, and the distance between the two devices is 5 meters, then the consistency degree is about 0.866, and the distance reciprocal is 0.2. The comprehensive edge weight can be set to 0.866×0.7+0.2×0.3=0.6662.

[0079] After the airflow conduction path directed graph is constructed, connectivity detection is performed for each pair of alarm information in the alarm spatio-temporal feature matrix to determine whether there is a directed connected path between the device nodes corresponding to the two alarm information. This step can be realized by using depth-first search or breadth-first search algorithm. For example, if there are alarm information of device C and device E in the matrix, all possible paths from C to E or from E to C in the directed graph will be searched.

[0080] When it is confirmed that there is a directed connected path between the two device nodes, the spatial coupling degree of the device node pair is calculated. The spatial coupling degree is obtained by accumulated and attenuated calculation of the edge weight values of the directed edges on the path. In specific implementation, all edges on the directed path from the starting node to the ending node are traversed, the edge weight values of each edge are collected, and then a path length attenuation factor is applied for calculation.

[0081] The path length attenuation factor is designed as a function that decreases as the number of nodes contained in the directed connected path increases. For example, for a path of length n (containing n+1 nodes), an attenuation factor of the form 1 / (1+α(n-1)) can be used, where α is an attenuation rate parameter, typically taking a value between 0.1 and 0.5. This design ensures that the longer the path, the more obvious the cumulative attenuation effect, reflecting the physical characteristic that the air conduction effect weakens with the propagation distance.

[0082] An entire calculation process is illustrated with a specific case: suppose that the data center has alarm devices A, B, C, D, and E, and in the directed graph of the air conduction path, there are paths A→B→C→D and A→E→D. Now the spatial coupling degree between device A and device D needs to be calculated. First, find the two paths from A to D in the directed graph: A→B→C→D (path 1) and A→E→D (path 2).

[0083] For path 1, suppose that the edge weights of edges A→B, B→C, and C→D are 0.8, 0.7, and 0.6 respectively, the path length is 3 (containing 4 nodes), and the attenuation factor is 1 / (1+0.2×(3-1))=0.71. The cumulative calculation obtains a spatial coupling degree of (0.8+0.7+0.6)×0.71=1.491.

[0084] For path 2, suppose that the edge weights of edges A→E and E→D are 0.75 and 0.65 respectively, the path length is 2 (containing 3 nodes), and the attenuation factor is 1 / (1+0.2×(2-1))=0.83. The cumulative calculation obtains a spatial coupling degree of (0.75+0.65)×0.83=1.162.

[0085] Compare the spatial coupling degree values of the two paths, and select the larger value 1.491 as the final spatial coupling degree between device A and device D. This indicates that the air conduction influence from device A to device D is stronger, and the alarm of the two devices has higher relevance.

[0086] Through the above method, the device alarm correlation caused by air conduction can be effectively identified, providing more accurate fault diagnosis basis for data center operation and maintenance personnel, reducing misjudgment and omission, and improving operation and maintenance efficiency. At the same time, the quantitative calculation of spatial coupling degree also provides an important feature input for subsequent alarm correlation analysis and root cause positioning.

[0087] In an optional implementation, based on the device group control cooperation graph structure, the root cause device node causing the alarm set is identified by using a reverse path tracing algorithm, including:

[0088] Traverse all device nodes in the device group control collaborative graph structure, and according to the alarm information in the alarm set, mark the device nodes generating alarms as alarm nodes and assign initial alarm propagation contribution degrees, and mark the device nodes not generating alarms as candidate root cause nodes;

[0089] Iterative traverse along the reverse direction of the control dependency edges in the device group control collaborative graph structure, for each alarm node, identify the upstream device nodes pointing to the alarm node through the control dependency edges, and according to the weights of the alarm propagation edges, perform reverse transfer of the alarm propagation contribution degree of the alarm node to the upstream device nodes, the reverse transfer is realized by combining the alarm propagation contribution degree with the attenuation coefficient of the weight of the alarm propagation edge, to obtain the cumulative alarm propagation contribution degrees of the upstream device nodes;

[0090] During the iterative traverse process, only the propagation paths meeting the directionality constraint of the control dependency edges are retained, the directionality constraint requires that the direction of the control dependency edge on the propagation path is consistent with the causal direction of the alarm propagation;

[0091] When the iterative traverse terminates, select the device node with the largest cumulative alarm propagation contribution degree from the candidate root cause nodes as the root cause device node.

[0092] Traverse all device nodes in the device group control collaborative graph structure, taking a network containing 10 device nodes as an example, when it is detected that 3 of the nodes (device A, device C and device F) generate alarm information, the 3 nodes are marked as alarm nodes and are assigned initial alarm propagation contribution degree values. For example, based on the severity of the alarms, device A is assigned an alarm contribution degree of 0.8, device C is assigned an alarm contribution degree of 0.6, and device F is assigned an alarm contribution degree of 0.7. The remaining 7 device nodes (device B, device D, device E, device G, device H, device I and device J) not generating alarms are marked as candidate root cause nodes, and their initial alarm propagation contribution degrees are all set to 0.

[0093] Iterative traverse is started along the reverse direction of the control dependency edges in the device group control collaborative graph structure. For example, it is detected that device A has control dependency edges from device B and device D with weights of 0.9 and 0.5 respectively, device C has a control dependency edge from device D with a weight of 0.8, and device F has a control dependency edge from device E with a weight of 0.7. For each alarm node, identify the upstream device nodes pointing to the alarm node through the control dependency edges.

[0094] For the alarm node device A, devices B and D are identified as its upstream nodes. The alarm propagation contribution of device A (0.8) is transferred to devices B and D through the backward transfer mechanism. Assuming the attenuation coefficient is 0.9, the contribution increment obtained by device B is 0.8x0.9x0.9=0.648, and the contribution increment obtained by device D is 0.8x0.9x0.5=0.36.

[0095] For the alarm node device C, device D is identified as its upstream node, and the alarm propagation contribution of device C (0.6) is transferred to device D. Using the same attenuation coefficient 0.9, the contribution increment obtained by device D is 0.6x0.9x0.8=0.432.

[0096] For the alarm node device F, device E is identified as its upstream node, and the alarm propagation contribution of device F (0.7) is transferred to device E. Using the same attenuation coefficient 0.9, the contribution increment obtained by device E is 0.7x0.9x0.7=0.441.

[0097] After this round of iteration, the cumulative alarm propagation contribution of device B is 0.648, the cumulative alarm propagation contribution of device D is 0.36+0.432=0.792, and the cumulative alarm propagation contribution of device E is 0.441. Continue the next round of iteration to check whether these nodes with obtained contribution have upstream nodes.

[0098] Suppose it is found that device B has a control dependency edge from device G with a weight of 0.6; device D has control dependency edges from devices G and H with weights of 0.7 and 0.4 respectively; device E has a control dependency edge from device H with a weight of 0.8. The contribution will continue to be transferred upstream according to the same mechanism.

[0099] During the transfer process, the directionality constraint of the control dependency edge is strictly followed to ensure that the alarm propagation path is consistent with the causal direction of the control dependency. For example, if there is a control dependency edge from device I to device J, but in the actual network topology, the alarm propagation direction is from device J to device I, which is opposite to the direction of the control dependency, then this propagation path will be excluded.

[0100] The iteration process continues until no new upstream nodes are found or all transferred contribution values are less than a preset threshold (such as 0.01). In this example, after multiple rounds of iteration, the cumulative alarm propagation contribution of device G finally reaches 1.253, and the cumulative alarm propagation contribution of device H reaches 0.896.

[0101] When the iteration traversal terminates, the device node with the largest accumulated alarm propagation contribution degree is selected from the candidate root cause nodes (device B, device D, device E, device G, device H, device I, device J) as the root cause device node. In this example, the cumulative alarm propagation contribution degree of device G is the highest (1.253), so it is identified as the root cause device node.

[0102] This root cause analysis method based on reverse path tracing can effectively trace the source of the fault by quantifying the influence relationship between devices and combining the actual alarm situation. Especially in a complex network environment, multiple alarms may be caused by the same root cause, and this method can accurately identify the root cause, rather than simply dealing with surface phenomena, thereby improving the accuracy and efficiency of fault diagnosis.

[0103] In an optional implementation, the method further includes:

[0104] The weight of the alarm propagation edge is decomposed into an airflow influence degree component and an alarm time sequence correlation component, the airflow influence degree component representing the airflow conduction influence intensity of the upstream device node on the alarm node, and the alarm time sequence correlation component representing the time sequence correlation intensity between the running state change of the upstream device node and the alarm occurrence time of the alarm node.

[0105] For each incident alarm propagation edge of the alarm node, the airflow influence degree component and the alarm time sequence correlation component are calculated by nonlinear fusion to obtain a comprehensive transmission coefficient of the incident alarm propagation edge, and the nonlinear fusion calculation is realized by product operation on the airflow influence degree component and the alarm time sequence correlation component and normalization processing.

[0106] The alarm propagation contribution degree of the alarm node is multiplied by the comprehensive transmission coefficient to obtain a contribution degree increment transmitted to the corresponding upstream device node through the incident alarm propagation edge.

[0107] The contribution degree increment is added to the current cumulative alarm propagation contribution degree of the corresponding upstream device node to complete the update of the cumulative alarm propagation contribution degree of the upstream device node.

[0108] The present embodiment provides a method for reverse transmission of alarm propagation contribution degree based on the weight of the alarm propagation edge. This method is suitable for device fault root cause positioning in a data center air conditioning system, and can accurately trace the upstream root cause device causing the alarm by analyzing the airflow conduction relationship and alarm time sequence correlation.

[0109] In a data center air conditioning system, the airflow influence relationship between devices is represented by an alarm propagation graph. The alarm propagation graph includes device nodes and alarm propagation edges, where the alarm propagation edges represent the influence path of upstream devices to downstream devices. When an alarm occurs at a device node, it is necessary to determine which upstream devices may be the root cause of the alarm. The present embodiment identifies the most likely root cause device by reversely transferring the contribution degree of the alarm node along the alarm propagation edges to the upstream devices.

[0110] For weight decomposition of the alarm propagation edges, the weight of each alarm propagation edge is divided into two key components: an airflow influence degree component and an alarm timing correlation degree component. The airflow influence degree component reflects the airflow conduction influence strength of the upstream device on the downstream device, which can be calculated by an airflow model or historical operation data. For example, the airflow influence degree component of a certain air conditioning unit A on a downstream hot spot area B is 0.75, indicating that the change in the operating state of A has a strong influence on the temperature of B. The alarm timing correlation degree component reflects the time sequence correlation degree between the change in the state of the upstream device and the alarm occurrence time of the downstream device, which is calculated by analyzing the time sequence relationship between the change in the state of the device and the triggering of the alarm. For example, if the fan speed of an upstream device C changes significantly 10 minutes before the alarm of a downstream device D, the alarm timing correlation degree component between them is 0.85, indicating that they have a strong time sequence correlation.

[0111] For each incident alarm propagation edge of the alarm node, nonlinear fusion calculation is required for the airflow influence degree component and the alarm timing correlation degree component to obtain a comprehensive transfer coefficient. The specific implementation method is to nonlinearly combine the two components by multiplication operation, and then apply normalization processing. For example, if the airflow influence degree component of a certain incident alarm propagation edge is 0.8 and the alarm timing correlation degree component is 0.6, the initial product is 0.48. Assuming that the current alarm node has three incident edges, their initial products are 0.48, 0.32, and 0.2, respectively, and the normalized comprehensive transfer coefficients are 0.48, 0.32, and 0.2, respectively. Normalization ensures that the sum of the comprehensive transfer coefficients of all incident edges is 1, maintaining the total conservation of the contribution degree in the reverse transfer process.

[0112] When calculating the contribution degree increment, the alarm propagation contribution degree of the alarm node is multiplied by the comprehensive transfer coefficient. Assuming that the alarm propagation contribution degree of a certain alarm node is 100, and the comprehensive transfer coefficients of the three incident edges are 0.48, 0.32, and 0.2, respectively, then the contribution degree increments transmitted to the upstream devices through the three edges are 48, 32, and 20, respectively. This indicates that the first upstream device has the largest contribution to the current alarm and is most likely the root cause of the fault.

[0113] Finally, the calculated contribution increment is added to the current cumulative alarm propagation contribution of the corresponding upstream device node. For example, if the current cumulative alarm propagation contribution of a certain upstream device node is 25, and the contribution increment received through the incident alarm propagation edge is 48, then the updated cumulative alarm propagation contribution is 73. After the system completes the reverse contribution transfer of the entire alarm propagation graph, the device node with the highest cumulative alarm propagation contribution is identified as the most likely root cause of the fault.

[0114] In actual application scenarios, this method can handle complex alarm propagation situations. For example, a data center has deployed 10 air conditioning units and 200 temperature sensors, of which 5 temperature sensors simultaneously trigger high-temperature alarms. By constructing an alarm propagation graph, the air flow influence degree component and the alarm timing correlation component between each alarm sensor and the upstream air conditioning device are analyzed. After the contribution reverse transfer calculation, it is found that the cumulative alarm propagation contribution of air conditioning unit G reaches 285, which is much higher than that of other air conditioning units (the second highest is air conditioning unit H, with a contribution of 127), so air conditioning unit G is determined as the most likely root cause device.

[0115] This method can also handle complex situations of multi-level propagation. When the alarm propagates from the downstream device to the upstream device, the cumulative contribution of the upstream device will be further transferred to the more upstream device. Through this multi-level transfer mechanism, even if there are multiple intermediate devices between the root cause and the alarm node, the system can still accurately trace back to the real fault source. For example, in a certain fault analysis, the system finally locates the root cause of the temperature anomaly in multiple areas to the control valve failure of a fresh air unit through three-level reverse transfer, rather than the directly connected air conditioning unit.

[0116] The advantage of this embodiment is that both the physical air flow influence relationship and the timing correlation are considered, and the influence of the two is balanced through a nonlinear fusion mechanism, improving the accuracy and interpretability of fault root cause positioning.

[0117] In an optional embodiment, generating a priority processing instruction for the root cause device node according to the root cause device node and its associated control dependency edge includes:

[0118] Starting from the root cause device node, breadth-first traversal is performed along the forward direction of the control dependency edge in the device group control collaboration graph structure, all downstream device nodes directly or indirectly controlled by the root cause device node are identified, and an influence domain topology subgraph containing the root cause device node and all downstream device nodes is constructed;

[0119] For each downstream device node in the influence domain topology subgraph, the corresponding regulation and control constraint condition of the downstream device node is extracted from the edge attribute of the control dependency edge;

[0120] The regulation constraint includes an operation parameter adjustment range allowed by the downstream device node and a state switching timing requirement of the downstream device node, the operation parameter adjustment range is determined by a control strength component in the control dependent edge weight, and the state switching timing requirement is determined by a response time delay component in the control dependent edge weight;

[0121] According to the topological hierarchical relationship of the influence domain topological subgraph, a main regulation instruction is generated for the root cause device node and a slave regulation instruction is generated for each downstream device node in a control conduction order from the root cause device node to the downstream device node.

[0122] In the device group control collaborative system, when the root cause device node is identified, a targeted processing instruction needs to be generated to minimize system loss and achieve efficient recovery. The core of the method is to determine the accurate regulation strategy by deeply analyzing the control dependency relationship between devices.

[0123] Exemplarily, the root cause device node identified is taken as a traversal starting point, and a breadth-first traversal operation is performed in the positive direction of the control dependent edge in the device group control collaborative graph structure. The traversal operation starts from the root cause device node, sequentially accesses the first layer of downstream device nodes directly connected through the control dependent edge, and adds these first layer of downstream device nodes to a to-be-accessed queue. Then, for each first layer of downstream device node in the queue, the second layer of downstream device nodes connected through the control dependent edge is accessed, and the second layer of downstream device nodes is added to the queue. In this way, the expansion is carried out layer by layer outwardly until all device nodes that can be reached from the root cause device node through the control dependent edge path are traversed. In the traversal process, the node identifier of each accessed device node and the hierarchical depth of the device node in the topological structure relative to the root cause device node are recorded, and the hierarchical depth is determined by the number of control dependent edges passed through in the traversal path. After the traversal is completed, a set of all downstream device nodes affected by the root cause device node directly or indirectly is obtained.

[0124] Based on the obtained downstream device node set, an influence domain topology subgraph is constructed. The influence domain topology subgraph contains the root cause device node and all downstream device nodes, and retains the control dependency edges between these device nodes and their complete edge attribute information. In the specific construction process, the root cause device node is taken as the root node of the influence domain topology subgraph, and the downstream device nodes are organized in layers according to their hierarchical depth. Device nodes of the same hierarchical depth belong to the same topology level, and different topology levels are connected by control dependency edges. For each control dependency edge retained in the influence domain topology subgraph, the weight information of the edge is extracted and saved, which includes a control strength component representing the control ability of the upstream device node to the downstream device node, and a response time delay component representing the time required for the control instruction to be transmitted from the upstream device node to the downstream device node. Through this construction process, a device influence range topology structure is formed, which is centered on the root cause device node and organized by control transmission levels.

[0125] For each downstream device node in the influence domain topology subgraph, the regulation constraint condition is extracted from the edge attribute of the incident control dependency edge connected to the downstream device node. For any downstream device node, its regulation constraint condition includes two dimensions of operating parameter adjustment amplitude and state switching timing requirement. The extraction of operating parameter adjustment amplitude is realized by reading the control strength component in the weight of the incident control dependency edge corresponding to the downstream device node. The larger the control strength component, the stronger the control influence of the upstream device node on the downstream device node, and correspondingly, the larger the operating parameter adjustment amplitude allowed by the downstream device node. The operating parameter adjustment amplitude is quantified as the maximum percentage range of the current operating parameter value of the downstream device node that can be changed, which is positively correlated with the control strength component. The extraction of state switching timing requirement is realized by reading the response time delay component in the weight of the incident control dependency edge corresponding to the downstream device node, which reflects the time required for the control instruction to be transmitted from the upstream device node to the downstream device node and take effect. The state switching timing requirement is quantified as the minimum response interval length that must be met after the downstream device node receives the regulation instruction, which ensures that the device has enough time to complete state conversion without control conflict. By performing this extraction operation on all downstream device nodes in the influence domain topology subgraph, the regulation constraint condition set of each downstream device node is obtained.

[0126] Based on the topological hierarchical relationship of the influence domain topology subgraph, the control conduction order from the root cause device node to the downstream device node is generated for each device node. First, the root cause device node generates a main control instruction, which contains the adjustment content for the running state of the root cause device node. The generation of the main control instruction is based on the device type and alarm characteristics of the root cause device node to determine the specific control action, which includes the adjustment amount of the running parameter and the switching target of the state mode. The running parameter adjustment amount is determined according to the alarm severity and the historical running data of the root cause device node. When the alarm severity is high, the running parameter adjustment amount is set to a large value to quickly eliminate the fault; when the alarm severity is low, the running parameter adjustment amount is set to a small value to smoothly transition. The state mode switching target is determined according to the alarm type and the functional positioning of the root cause device node to ensure that the switched state mode can directly eliminate the root cause of the alarm.

[0127] After generating the main control instruction, the topological hierarchical order of the influence domain topology subgraph is followed to generate the slave control instruction for the downstream device node from the first level. For the downstream device node of the first level, according to the expected state change after the execution of the main control instruction of the root cause device node, combined with the control dependency edge weight between the downstream device node and the root cause device node, the cooperative adjustment amount that needs to be executed by the downstream device node is calculated. The calculation of the cooperative adjustment amount considers the influence degree of the main control instruction on the running state of the downstream device node, which is determined by the control intensity component of the control dependency edge. The specific calculation method is to multiply the running parameter adjustment amount in the main control instruction and the control intensity component to obtain the parameter change amount that the downstream device node needs to respond. At the same time, according to the state switching time sequence requirement of the downstream device node, the execution time of the slave control instruction is determined, which is set as the execution time of the main control instruction plus the response delay component, to ensure the time sequence coordination of the control action. In the same way, the slave control instruction is generated for all downstream device nodes of the second level, the third level, and the last level. The slave control instruction of each level of downstream device node is calculated based on the control instruction of its previous level device node and the weight information of the corresponding control dependency edge.

[0128] After generating all the control instructions, constraint checking is performed on the master control instruction and each slave control instruction. The constraint checking process checks whether the running parameter adjustment amount in each control instruction exceeds the running parameter adjustment range allowed by the corresponding device node. For the master control instruction, its running parameter adjustment amount is compared with the running parameter adjustment range constraint of the root cause device node itself; for each slave control instruction, its cooperative adjustment amount is compared with the running parameter adjustment range constraint of the corresponding downstream device node. When it is detected that the adjustment amount of a control instruction exceeds the constraint range, the over-limit adjustment amount is corrected by reducing it to the boundary value of the allowed running parameter adjustment range. The difference in the reduced adjustment amount needs to be redistributed to other device nodes in the influence domain topology subgraph to maintain the overall control effect. The target device nodes for redistribution are selected from device nodes at the same topology level or adjacent topology levels as the current device node, and the selection criteria are that the target device nodes have remaining capacity in the running parameter adjustment range constraint and have similar functions to the current device node. The reduced adjustment amount difference is distributed according to the proportion of the remaining capacity of the target device nodes, and the slave control instructions of the target device nodes are updated accordingly.

[0129] At the same time, the constraint checking process also checks whether the execution timing of each control instruction meets the state switching timing requirement. The checking method is to verify whether the difference between the execution time of the slave control instruction of the subsequent level device node and the execution time of the control instruction of the previous level device node is greater than or equal to the response time delay component of the corresponding control dependency edge. When it is detected that the timing requirement is violated, the execution time of the slave control instruction of the subsequent level device node is adjusted to the earliest time that meets the response time delay component requirement. After adjustment amount constraint checking and timing constraint checking, the optimized master control instruction and the optimized slave control instruction set that meet all the control constraint conditions are obtained.

[0130] In an optional implementation, the method further includes:

[0131] The running parameter adjustment amount in the master control instruction and the cooperative adjustment amount in the slave control instruction are substituted into the control constraint condition for feasibility checking,

[0132] When it is detected that the adjustment amount exceeds the running parameter adjustment range or violates the state switching timing requirement, the over-limit adjustment amount is corrected by reducing it and redistributed to other device nodes in the influence domain topology subgraph, to obtain the optimized master control instruction and the optimized slave control instruction that meet the control constraint condition, and the optimized master control instruction is output as the priority processing instruction.

[0133] The feasibility check also verifies the state switching timing requirement. The state switching timing requirement check is performed from the execution time of the control instruction to ensure that the control actions of the device nodes at each level meet the timing sequence of the control transmission. For each control dependency edge in the domain topology subgraph, the control dependency edge connects an upstream device node and a downstream device node, and the response time component is included in the edge attribute of the control dependency edge. The response time component defines the minimum time interval required for the downstream device node to start responding after the upstream device node control instruction takes effect. The execution time of the upstream device node control instruction and the execution time of the downstream device node control instruction are extracted, and the time difference between the two execution times is calculated. The time difference is compared with the response time component of the corresponding control dependency edge. When the time difference is less than the response time component, it is determined that the downstream control instruction violates the state switching timing requirement, and the downstream control instruction is marked as a control instruction that needs timing adjustment. By traversing all control dependency edges and the device node pairs connected by the control dependency edges in the domain topology subgraph, a set of downstream control instructions that violate the state switching timing requirement is obtained.

[0134] For the control instructions that violate the running parameter adjustment amplitude constraint, a reduction correction operation is performed. For the main control instruction, when the running parameter adjustment amount exceeds the upper limit of the running parameter adjustment amplitude of the root cause device node, the running parameter adjustment amount is reduced to the upper limit of the running parameter adjustment amplitude, and the reduction method is to set the value of the running parameter adjustment amount to the upper limit of the running parameter adjustment amplitude while keeping the adjustment direction unchanged. The difference between the running parameter adjustment amount before and after the reduction is calculated, which represents the adjustment amount that cannot be implemented on the root cause device node due to the constraint limitation, and is recorded as the to-be-distributed adjustment amount. For the downstream control instruction, when the cooperative adjustment amount exceeds the upper limit of the running parameter adjustment amplitude of the corresponding downstream device node, the same reduction operation is performed, the cooperative adjustment amount is reduced to the upper limit of the running parameter adjustment amplitude of the corresponding device node, and the to-be-distributed adjustment amount generated by the reduction is calculated. Through the reduction correction operation, it is ensured that the adjustment amount of all control instructions does not exceed the running parameter adjustment amplitude constraint of the corresponding device node.

[0135] The to-be-distributed adjustment amount generated by the reduction correction needs to be redistributed to other device nodes in the domain topology subgraph to maintain the equivalence of the overall control effect. The redistribution operation first determines a set of candidate device nodes that can receive the to-be-distributed adjustment amount. The candidate device nodes need to meet two conditions: first, the candidate device nodes are at the same topological level or adjacent topological levels as the original device nodes that generate the to-be-distributed adjustment amount in the domain topology subgraph, to ensure the timing coordination of the control actions; second, the current control instruction adjustment amount of the candidate device nodes has not reached the upper limit of the running parameter adjustment amplitude, i.e., the candidate device nodes still have adjustment capacity to accept additional adjustment amount. By traversing all device nodes in the domain topology subgraph and checking the above two conditions, a set of candidate device nodes is selected.

[0136] For each candidate device node in the candidate device node set, a remaining adjustment capacity of the candidate device node is calculated, the remaining adjustment capacity being equal to an upper limit of an adjustment range of an operating parameter of the candidate device node minus an absolute value of an adjustment amount of a current control instruction of the candidate device node. The remaining adjustment capacities of all the candidate device nodes are accumulated to obtain a total remaining adjustment capacity. An allocation weight of each candidate device node is calculated, the allocation weight being equal to the remaining adjustment capacity of the candidate device node divided by the total remaining adjustment capacity, the allocation weight reflecting a relative proportion of an adjustment amount that can be assumed by the candidate device node among all the candidate device nodes. The adjustment amount to be allocated is proportionally allocated according to the allocation weights of the candidate device nodes, and a specific allocation amount is equal to the adjustment amount to be allocated multiplied by the allocation weight of the corresponding candidate device node. The allocated adjustment amount is added to the control instruction of the corresponding candidate device node to update a numerical value of the adjustment amount of the control instruction of the candidate device node. Through the re-allocation mechanism, equivalent compensation of the reduced adjustment amount in the influence domain is ensured, and a decline in overall control effect caused by single-point constraint is avoided.

[0137] The embodiment of the application faces a machine room air conditioner alarm cooperative processing system for device group control, comprising:

[0138] A first unit is configured to acquire alarm information generated by a plurality of air conditioner devices in a machine room in a current time window, perform spatial correlation analysis on the alarm information based on a preset device topology relationship, and obtain an alarm set with spatial coupling characteristics, wherein the device topology relationship includes air supply path dependency relationships and return air path dependency relationships between devices.

[0139] A second unit is configured to map the alarm information to a pre-constructed device group control cooperative graph structure according to device identifiers in the alarm information and the spatial coupling characteristics, wherein the device group control cooperative graph structure includes device nodes, alarm propagation edges, and control dependency edges between devices, and a weight of the alarm propagation edge is determined by air flow influence degrees between devices and alarm time sequence correlation degrees.

[0140] A third unit is configured to identify a root cause device node causing the alarm set based on the device group control cooperative graph structure by using a reverse path tracing algorithm, generate a priority processing instruction for the root cause device node according to the root cause device node and associated control dependency edges, wherein the priority processing instruction includes a device control parameter adjustment action and an alarm suppression strategy, and the priority processing instruction is sent to an air conditioner device corresponding to the root cause device node.

[0141] In a third aspect of the embodiment of the application, an electronic device is provided, comprising:

[0142] a processor;

[0143] a memory for storing processor-executable instructions;

[0144] The processor is configured to invoke instructions stored in the memory to perform the method described above.

[0145] In a fourth aspect, the present application provides a computer readable storage medium, having stored thereon computer program instructions, which when executed by a processor implement the method described above.

[0146] The present application can be a method, apparatus, system, and / or computer program product. Computer program products can include computer readable storage media having computer readable program instructions thereon for performing various aspects of the present application.

[0147] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A collaborative alarm processing method for computer room air conditioning for equipment group control, characterized in that, include: The alarm information generated by multiple air conditioning devices in the computer room within the current time window is obtained. Based on the preset device topology relationship, the spatial correlation analysis of the alarm information is performed to obtain an alarm set with spatial coupling characteristics. The device topology relationship includes the supply air path dependency relationship and return air path dependency relationship between devices. Based on the device identifier in the alarm information and the spatial coupling feature, the alarm information is mapped to a pre-constructed device group control collaborative graph structure. The device group control collaborative graph structure includes device nodes, alarm propagation edges, and control dependency edges between devices. The weight of the alarm propagation edge is determined by the airflow influence degree between devices and the alarm timing correlation degree. Based on the device group control collaborative graph structure, a reverse path tracing algorithm is used to identify the root cause device node that causes the alarm set. According to the root cause device node and its associated control dependency edge, a priority processing instruction is generated for the root cause device node. The priority processing instruction includes device control parameter adjustment actions and alarm suppression strategies. The priority processing instruction is then sent to the air conditioning equipment corresponding to the root cause device node.

2. The method according to claim 1, characterized in that, Based on a preset device topology, spatial correlation analysis is performed on the alarm information to obtain an alarm set with spatial coupling characteristics, including: Based on the device identifier in the alarm information, the spatial location coordinates and airflow direction vector of the device corresponding to each alarm information are extracted from the device topology relationship. Combined with the alarm timestamp, the alarm timing difference between adjacent devices is calculated to obtain an alarm spatiotemporal feature matrix that includes spatial distance features and timing interval features. Based on the supply air path dependency and the return air path dependency, airflow conduction path matching is performed on each alarm information in the alarm spatiotemporal feature matrix, and the spatial coupling degree between devices on the same airflow conduction path is calculated. The spatial coupling degree is determined by the connectivity of the airflow conduction path and the spatial distance feature. Alarm information whose spatial coupling exceeds a preset coupling threshold is grouped into the same alarm set to obtain the alarm set with spatial coupling characteristics.

3. The method according to claim 2, characterized in that, Based on the supply air path dependency and the return air path dependency, airflow conduction path matching is performed on each alarm information in the alarm spatiotemporal feature matrix, and the spatial coupling degree between devices on the same airflow conduction path is calculated, including: Based on the supply air path dependency and the return air path dependency, a directed graph of airflow conduction paths is constructed. The nodes of the directed graph of airflow conduction paths correspond to the device identifiers in the alarm spatiotemporal feature matrix. The directed edges of the directed graph of airflow conduction paths represent the airflow conduction relationship in the supply air direction or the return air direction. The edge weight of the directed edge is determined by the consistency degree of the airflow direction vector and the reciprocal of the spatial distance feature. For each pair of alarm messages, a connectivity test is performed on the directed graph of the airflow conduction path to determine whether a directed connected path exists between the device nodes corresponding to the two alarm messages; If it exists, the cumulative attenuation calculation is performed on the edge weights of all directed edges on the directed connected path to obtain the spatial coupling degree of the device node pair. The cumulative attenuation calculation is achieved by weighted summation of the edge weights of each directed edge on the path and applying a path length attenuation factor. The path length attenuation factor decreases as the number of nodes contained in the directed connected path increases.

4. The method according to claim 1, characterized in that, Based on the device group control collaborative graph structure, the root cause device node leading to the alarm set is identified using a reverse path tracing algorithm, including: Traverse all device nodes in the device group control collaboration graph structure, and based on the alarm information in the alarm set, mark the device nodes that generate alarms as alarm nodes and assign them an initial alarm propagation contribution, and mark the device nodes that do not generate alarms as candidate root cause nodes; The system iterates along the reverse direction of the control dependency edges in the device group control collaboration graph structure. For each alarm node, it identifies the upstream device node that is pointed to by the control dependency edge, and propagates the alarm propagation contribution of the alarm node to the upstream device node in reverse according to the weight of the alarm propagation edge. The reverse propagation is achieved by combining the alarm propagation contribution with the attenuation coefficient of the alarm propagation edge weight, so as to obtain the cumulative alarm propagation contribution of each upstream device node. During the iterative traversal, only propagation paths that satisfy the control dependency edge directionality constraint are retained. The directionality constraint requires that the direction of the control dependency edge on the propagation path be consistent with the causal direction of alarm propagation. When the iterative traversal terminates, the device node with the largest cumulative alarm propagation contribution is selected from the candidate root cause nodes as the root cause device node.

5. The method according to claim 4, characterized in that, Based on the weight of the alarm propagation edge, the alarm propagation contribution of the alarm node is propagated back to the upstream device node, including: The weights of the alarm propagation edge are decomposed into an airflow influence component and an alarm timing correlation component. The airflow influence component represents the intensity of the airflow transmission influence of the upstream device node on the alarm node, and the alarm timing correlation component represents the temporal correlation strength between the change in the operating state of the upstream device node and the alarm occurrence time of the alarm node. For each incident alarm propagation edge of an alarm node, a nonlinear fusion calculation is performed on the airflow influence component and the alarm timing correlation component to obtain the comprehensive transmission coefficient of the incident alarm propagation edge. The nonlinear fusion calculation is achieved by multiplying the airflow influence component and the alarm timing correlation component and applying normalization processing. The alarm propagation contribution of the alarm node is multiplied by the comprehensive transmission coefficient to obtain the contribution increment transmitted to the corresponding upstream device node through the incident alarm propagation edge; The contribution increment is added to the current cumulative alarm propagation contribution of the corresponding upstream device node to complete the update of the cumulative alarm propagation contribution of the upstream device node.

6. The method according to claim 1, characterized in that, Based on the root cause device node and its associated control dependency edges, priority processing instructions are generated for the root cause device node, including: Starting from the root cause device node, a breadth-first traversal is performed along the positive direction of the control dependency edges in the device group control collaboration graph structure to identify all downstream device nodes directly or indirectly affected by the root cause device node, and to construct an influence domain topology subgraph containing the root cause device node and all downstream device nodes; For each downstream device node in the influence domain topology subgraph, the control constraint conditions corresponding to the downstream device node are extracted from the edge attributes of the control-dependent edge. The control constraints include the allowable adjustment range of the operating parameters of the downstream device node and the state switching timing requirements of the downstream device node. The adjustment range of the operating parameters is determined by the control strength component in the control dependency edge weights, and the state switching timing requirements are determined by the response delay component in the control dependency edge weights. Based on the topological hierarchy of the influence domain topology subgraph, and in accordance with the control transmission sequence from the root cause device node to the downstream device node, a master control command is generated for the root cause device node, and a slave control command is generated for each downstream device node.

7. The method according to claim 6, characterized in that, The method further includes: The feasibility of the operation parameter adjustment amounts in the main control command and the coordinated adjustment amounts in the secondary control command is verified by substituting them into the control constraints. When the adjustment amount is detected to exceed the adjustment range of the operating parameters or violate the state switching timing requirements, the excessive adjustment amount is reduced and corrected and redistributed to other device nodes in the influence domain topology subgraph to obtain optimized master control instructions and optimized slave control instructions that satisfy the control constraints. The optimized master control instructions are then output as the priority processing instructions.

8. A collaborative processing system for alarms in computer room air conditioning systems for equipment group control, used to implement the method as described in any one of claims 1-7, characterized in that, include: The first unit is used to acquire alarm information generated by multiple air conditioning devices in the computer room within the current time window, and to perform spatial correlation analysis on the alarm information based on the preset device topology relationship to obtain an alarm set with spatial coupling characteristics. The device topology relationship includes the supply air path dependency relationship and return air path dependency relationship between devices. The second unit is used to map the alarm information to a pre-constructed device group control collaborative graph structure based on the device identifier in the alarm information and the spatial coupling feature. The device group control collaborative graph structure includes device nodes, alarm propagation edges, and control dependency edges between devices. The weight of the alarm propagation edge is determined by the airflow influence degree between devices and the alarm timing correlation degree. The third unit is used to identify the root cause device node that causes the alarm set by using a reverse path tracing algorithm based on the device group control collaborative graph structure. Based on the root cause device node and its associated control dependency edge, it generates a priority processing instruction for the root cause device node. The priority processing instruction includes device control parameter adjustment actions and alarm suppression strategies. The priority processing instruction is then sent to the air conditioning equipment corresponding to the root cause device node.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images