Lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power
By deploying edge computing devices near lithium bromide heat pump equipment, establishing a global equipment topology map, and performing phased calculations, the data transmission delay and computational isolation problems of lithium bromide heat pump units were solved, achieving efficient distributed collaborative management and optimization analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHAANXI TOPSAIL ELECTRIC TECH CO LTD
- Filing Date
- 2026-01-20
- Publication Date
- 2026-05-12
AI Technical Summary
In existing technologies, the data acquisition and processing systems of lithium bromide heat pump units suffer from problems such as high network bandwidth pressure, severe analysis latency, isolated computing tasks, and difficulty in data integration. They also lack effective distributed computing power collaborative management methods, resulting in low computing efficiency.
A collaborative management system for waste heat recovery from lithium bromide heat pumps based on distributed computing power is adopted. By deploying edge computing devices near each device, a global device topology map is established, a unified data acquisition template is implemented, collaborative analysis tasks are decomposed into multiple computing stages, and directional flow computing is performed between devices. Sub-task coordination nodes are used to manage the computing process, enabling on-demand data transmission and fusion.
It reduces network transmission volume, lowers the computational load on the central management device, improves the uniformity of data structure and the synchronization of time series, enhances computing efficiency and the reliability of analysis, and achieves more efficient distributed collaborative optimization.
Smart Images
Figure CN121539917B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of waste heat recovery collaborative computing technology, specifically a lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power. Background Technology
[0002] In the field of industrial waste heat recovery, the operation monitoring and energy efficiency optimization of lithium bromide heat pump units typically rely on centralized data acquisition and processing systems. Such systems remotely transmit the operating status data of all equipment to a central server for unified analysis. Another common approach is to use independent edge computing units, each responsible only for local data acquisition and simple processing, lacking inter-device collaboration.
[0003] In a centralized architecture, the long-distance transmission of massive amounts of real-time data puts enormous pressure on network bandwidth, significantly delays in analysis and decision-making, and makes it difficult to meet the demands for rapid response under complex operating conditions. Centralized data processing also makes the central server a performance bottleneck and a single point of failure risk. While a simple independent edge computing model alleviates the pressure on data transmission, the computing tasks of each node are isolated, making it impossible to leverage the physical connections and data logic between device clusters for deeper collaborative analysis and optimization, thus limiting the dimension and depth of data analysis.
[0004] Current technologies lack a method to effectively organize distributed computing power and perform collaborative computing based on the actual relationships between devices. Furthermore, the timing and type of data acquisition by each device are determined locally, leading to a significant burden of data cleaning, alignment, and integration when the backend performs multi-source data fusion analysis. This results in low computational efficiency and hinders the generation and application of real-time collaborative optimization strategies based on system-wide data. Summary of the Invention
[0005] The purpose of this invention is to provide a collaborative management system for waste heat recovery from lithium bromide heat pumps based on distributed computing power, so as to solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides a collaborative management system for waste heat recovery from lithium bromide heat pumps based on distributed computing power, the system comprising:
[0007] Multiple edge computing devices and a collaborative management device are deployed near a lithium bromide heat pump device to collect the operating status data of the corresponding lithium bromide heat pump device.
[0008] The collaborative management device establishes a global device topology graph. Nodes in the global device topology graph correspond to a lithium bromide heat pump device and its edge computing device. Edges in the global device topology graph represent the physical or logical connection relationship between two lithium bromide heat pump devices.
[0009] The collaborative management device issues a data acquisition template to each edge computing device. The data acquisition template specifies the acquisition type and timing of the operating status data. The edge computing device acquires and stores the operating status data based on the data acquisition template to form a local status dataset.
[0010] The collaborative management device initiates a collaborative analysis task based on the global device topology map. The collaborative analysis task includes multiple calculation stages. Each calculation stage is assigned to a node in the global device topology map. The edge computing device corresponding to the node executes the calculation stage. The calculation stage uses the local state dataset and intermediate data transmitted from adjacent nodes in the global device topology map to generate intermediate calculation results and transmits the intermediate calculation results to the next node specified by the collaborative management device.
[0011] Preferably, before initiating a collaborative analysis task, the collaborative management device divides the global device topology graph to obtain multiple topology subgraphs. Each topology subgraph contains at least one node, and the nodes within the topology subgraphs are directly connected by edges.
[0012] The collaborative management device assigns a set of computation stages to each topology subgraph, and the computation stages in the set of computation stages correspond one-to-one with the nodes in the topology subgraph.
[0013] The collaborative management device sends the start command for the collaborative analysis task and the corresponding set of computation stages to the node with the smallest load value in each topology subgraph. This node is designated as the subtask coordination node of the topology subgraph.
[0014] Preferably, after receiving the start command and the set of computation stages, the subtask coordination node determines the execution order of each computation stage in the set of computation stages according to the connection relationship between nodes in the topology subgraph, thus forming a subtask execution chain.
[0015] The subtask coordination node activates the edge computing devices of the corresponding nodes in the topology subgraph in sequence according to the subtask execution chain to perform the calculation phase. When the edge computing device completes its calculation phase, the intermediate calculation results generated are passed along the subtask execution chain to the edge computing device corresponding to the next calculation phase.
[0016] Preferably, when the calculation phase requires intermediate data from other topological subgraphs in the global device topology graph, the subtask coordination node sends a data request to the collaborative management device.
[0017] The collaborative management device locates the subtask coordination nodes of other topological subgraphs that hold the required intermediate data according to the data request, and establishes a data channel between the current subtask coordination node and the target subtask coordination node.
[0018] Data is transmitted through the data channel, and the receiving subtask coordination node injects the received intermediate data into the designated computation stage of the subtask execution chain it maintains.
[0019] Preferably, after each topology subgraph is calculated, its subtask coordination node will generate a topology subgraph result and upload the topology subgraph result to the collaborative management device;
[0020] The collaborative management device collects all topology subgraph results and merges the results of each topology subgraph based on the global device topology graph to generate global analysis results.
[0021] Preferably, the collaborative management device calculates the load value of each node based on the historical computing power consumption data and real-time operating status data of the edge computing power device corresponding to each node;
[0022] The node with the lowest load value is defined as one in which the sum of the average of its historical computing power consumption data and the current computing power utilization rate reflected by its real-time operating status data is lower than a preset load threshold.
[0023] Preferably, the collaborative management device determines the fusion weight between the results of each topology subgraph based on the connection attributes of the edges in the global device topology graph. The connection attributes include connection type and connection strength.
[0024] The collaborative management device performs weighted calculations on the topology subgraph results according to the fusion weights, and then normalizes the weighted calculation results to obtain the global analysis results.
[0025] Preferably, the edge computing device periodically checks the data integrity of the local status dataset. When data loss is detected, the edge computing device re-acquires or interpolates the data for the missing period according to the acquisition sequence in the data acquisition template.
[0026] Once the data integrity requirements are met, the edge computing device compresses the local state dataset and synchronizes the metadata of the compressed local state dataset to the collaborative management device.
[0027] Preferably, before activating the edge computing device of a node within its topology subgraph to perform the computation phase, the subtask coordination node will verify the data integrity status of the edge computing device.
[0028] If the data integrity status is incomplete, the subtask coordination node will delay the activation of this computation phase and instruct the edge computing device to prioritize the data integrity repair operation.
[0029] Preferably, the collaborative management device maps the global analysis results back to the global device topology map and generates a control parameter adjustment suggestion for each node;
[0030] The collaborative management device sends control parameter adjustment suggestions to the edge computing devices of the corresponding nodes, and the edge computing devices drive the local lithium bromide heat pump equipment to perform parameter adjustments.
[0031] Compared with the prior art, the beneficial effects of the present invention are:
[0032] By establishing a global device topology map and decomposing collaborative analysis tasks into computational stages corresponding to topology nodes, the directional flow and phased execution of computational tasks within the device network are achieved. Each edge computing device, acting as a topology node, directly utilizes locally stored data and receives intermediate results from adjacent nodes for computation when executing its assigned computational stage. The resulting intermediate results then flow along the topology path to the next designated node. This approach ensures that computation occurs near the data generation or flow location, reducing the amount of network transmission of raw data to the central convergence point and lowering the computational load on the central management device. The computation process proceeds along the physical or logical connections between devices, enabling data analysis to align with actual thermal processes or technological links, achieving more physically consistent and efficient distributed collaborative reasoning and optimization analysis.
[0033] By uniformly defining and distributing data acquisition templates through a collaborative management device, the type and timing of data collected by all edge nodes are mandated. Each edge computing device collects and stores data locally based on the same template, ensuring the uniformity of data structure and the synchronization of time series across the entire system from the source of data generation. This allows subsequent collaborative computing tasks across multiple nodes to directly obtain standardized data sets with consistent structure and timestamp alignment, eliminating the need for complex and time-consuming data format conversions and time synchronization calibrations before or during computation. This eliminates data heterogeneity issues, provides ready-to-use and well-organized data input for distributed pipelined computing, reduces the overall preprocessing overhead and complexity of collaborative computing tasks, and improves the reliability and execution efficiency of distributed analysis processes. Attached Figure Description
[0034] Figure 1 This is a schematic diagram of the working principle of the lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power described in this invention.
[0035] Figure 2 A flowchart for topology graph partitioning and task allocation;
[0036] Figure 3 A flowchart for cross-topology subgraph data request and processing;
[0037] Figure 4 A bar chart showing the load distribution grouping of edge computing devices in a lithium bromide heat pump collaborative management system;
[0038] Figure 5A bar chart showing the data integrity ratio of the local status dataset for lithium bromide heat pump equipment. Detailed Implementation
[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0040] Please see Figure 1 This invention provides a collaborative management system for waste heat recovery from lithium bromide heat pumps based on distributed computing power. The system includes multiple edge computing devices and a collaborative management device. Each edge computing device is deployed near a lithium bromide heat pump device to collect operational status data. The collaborative management device establishes a global device topology graph, where nodes correspond to a lithium bromide heat pump device and its corresponding edge computing device, and edges represent physical or logical connections between two lithium bromide heat pump devices. The collaborative management device issues a data acquisition template to each edge computing device, specifying the acquisition type and timing of operational status data. The edge computing devices collect and store operational status data based on the data acquisition template to form a local status dataset. The collaborative management device initiates a collaborative analysis task based on the global device topology graph. This task includes multiple computation stages, each assigned to a node in the global device topology graph. The edge computing device corresponding to each node executes the computation stage. The computation stage uses the local status dataset and intermediate data transmitted from adjacent nodes in the global device topology graph to generate intermediate computation results, which are then transmitted to the next node specified by the collaborative management device.
[0041] In one embodiment of the present invention, see [reference] Figure 2In practical implementation, taking the waste heat recovery system of a chemical plant as an example scenario, the plant has deployed twelve lithium bromide heat pump devices, numbered BHP-01 to BHP-12. Each lithium bromide heat pump device is deployed near an edge computing device, which corresponds one-to-one with the lithium bromide heat pump device and collects its operating status data. The collaborative management device establishes a global device topology diagram containing twelve nodes based on the steam pipe connection relationship and process coupling relationship between the lithium bromide heat pump devices. The edges in the diagram represent the logical association between direct physical pipe connections or high-temperature wastewater discharge between the lithium bromide heat pump devices. In its implementation, before initiating a collaborative analysis task aimed at optimizing the waste heat distribution efficiency across the entire plant, the collaborative management device divides the global equipment topology graph. The division is based on the density of the connecting edges between the lithium bromide heat pump devices. For example, lithium bromide heat pump devices BHP-01, BHP-02, BHP-03, and BHP-04 have densely connected edges and are grouped into the same topology subgraph, labeled as subgraph Alpha. Lithium bromide heat pump devices BHP-05 to BHP-08 are grouped into topology subgraph Beta, and lithium bromide heat pump devices BHP-09 to BHP-12 are grouped into topology subgraph Gamma. Nodes within each topology subgraph are directly connected by edges. The collaborative management device assigns a set of computational stages to each topology subgraph. For topology subgraph Alpha, the allocated set of computational stages contains four computational stages, numbered S1-A, S2-A, S3-A, and S4-A. These four computational stages correspond one-to-one with the four nodes in topology subgraph Alpha.
[0042] In some embodiments, the collaborative management device calculates the load value of each node based on the historical computing power consumption data and real-time operating status data of the edge computing power device corresponding to each node. The historical computing power consumption data is the average value over the most recent 24 hours, and the real-time operating status data reflects the current computing power utilization rate. The formula used by the collaborative management device to calculate the load value is:
[0043]
[0044] in: This represents the load value of node n. This represents the average historical computing power consumption data of the edge computing device corresponding to node n. This represents the current computing power utilization rate of the edge computing device corresponding to node n, as reflected by real-time operating status data. In specific implementation, for the four nodes of the topology subgraph Alpha, the collaborative management device calculates the load value of the edge computing device corresponding to the lithium bromide heat pump equipment BHP-02. The load values of the edge computing devices corresponding to the lithium bromide heat pump devices BHP-01, BHP-03, and BHP-04 are 0.45, 0.38, and 0.40, respectively, and the preset load threshold is 0.85. The load value of the edge computing device corresponding to the lithium bromide heat pump device BHP-02 meets the condition of being lower than the load threshold and is the smallest among the four nodes. Therefore, the collaborative management device sends the start command and calculation stage set of the collaborative analysis task to the node corresponding to the lithium bromide heat pump device BHP-02 and designates this node as the subtask coordination node of the topology subgraph Alpha.
[0045] It can be understood that after the subtask coordination node, i.e. the edge computing device corresponding to the lithium bromide heat pump device BHP-02, receives the start command and the set of computing stages, it determines the execution order of each computing stage according to the connection relationship between the nodes in the topology subgraph Alpha. The lithium bromide heat pump devices BHP-01 are connected to BHP-02, BHP-02 are connected to BHP-03, and BHP-03 are connected to BHP-04. Based on this connection path, the subtask execution chain formed by the subtask coordination node is as follows: computing stage S1-A (corresponding to the lithium bromide heat pump device BHP-01 node) -> computing stage S2-A (corresponding to the lithium bromide heat pump device BHP-02 node) -> computing stage S3-A (corresponding to the lithium bromide heat pump device BHP-03 node) -> computing stage S4-A (corresponding to the lithium bromide heat pump device BHP-04 node). The subtask coordination node activates the edge computing device of the corresponding node in the topology subgraph Alpha in sequence according to the execution chain of this subtask. First, the edge computing device corresponding to the lithium bromide heat pump device BHP-01 is activated to execute the calculation stage S1-A. The calculation stage S1-A uses the local state dataset of the edge computing device corresponding to the lithium bromide heat pump device BHP-01 to perform calculations and generate intermediate calculation results M1.
[0046] In specific implementation, after the edge computing device corresponding to the lithium bromide heat pump device BHP-01 completes the calculation phase S1-A, it passes the intermediate calculation result M1 along the subtask execution chain to the edge computing device corresponding to the next calculation phase S2-A, namely the edge computing device corresponding to the lithium bromide heat pump device BHP-02. The edge computing device corresponding to the lithium bromide heat pump device BHP-02 executes the calculation phase S2-A. The calculation phase S2-A uses the local state dataset of the edge computing device corresponding to the lithium bromide heat pump device BHP-02 and the received intermediate calculation result M1 to perform the next stage calculation, generating the intermediate calculation result M2. After the edge computing device corresponding to the lithium bromide heat pump device BHP-02 completes the calculation phase S2-A, it passes the intermediate calculation result M2 to the edge computing device corresponding to the lithium bromide heat pump device BHP-03 to execute the calculation phase S3-A. The subsequent process follows this pattern until all subtask execution chains within the topology subgraph Alpha are completed. For topological subgraphs Beta and Gamma, the collaborative management device uses the same mechanism to select subtask coordination nodes and drive them to form and execute their respective subtask execution chains.
[0047] Optionally, at the data comparison level, if the topology partitioning and subtask coordination node mechanism in this embodiment is not adopted, the collaborative management device needs to directly manage and coordinate all computational stage dependencies and data transmission relationships between the twelve nodes, resulting in complex communication and control links. With this embodiment, the collaborative management device only needs to interact with the three subtask coordination nodes of the three topology subgraphs. The execution chain within the topology subgraphs is autonomously managed by the subtask coordination nodes, reducing the number of managed objects from twelve to three, thus simplifying the scheduling logic of collaborative analysis tasks. It can be understood that by dividing the global device topology graph into multiple topology subgraphs and selecting the node with the lowest load value as the subtask coordination node within each subgraph, the collaborative management device delegates the scheduling responsibility of the computational stage, achieving distributed sharing of management pressure. Simultaneously, using the node with the lowest load value as the coordination starting point helps balance the computational startup load within the subgraph.
[0048] In one embodiment of the present invention, see [reference] Figure 3In practical implementation, continuing the scenario of a chemical plant waste heat recovery system, the global equipment topology has been divided into topology subgraphs Alpha, Beta, and Gamma. Each topology subgraph has a subtask coordination node driving its internal subtask execution chain. In practice, when the subtask coordination node of topology subgraph Alpha is activated and executes the set of computation stages in sequence, the logic design of computation stage S3-A requires obtaining intermediate data on the flow rate of high-temperature wastewater. This intermediate data is not generated internally by topology subgraph Alpha, but depends on the intermediate calculation results generated by the lithium bromide heat pump device BHP-11 node in topology subgraph Gamma after completing its computation stage S2-G. When the subtask coordination node of topology subgraph Alpha is scheduled to the pre-stage of computation stage S3-A in the subtask execution chain, it identifies the dependency of computation stage S3-A on intermediate data of topology subgraph Gamma. The subtask coordination node of topology subgraph Alpha, i.e. the edge computing power device corresponding to lithium bromide heat pump device BHP-02, sends a data request to the collaborative management device. The data request contains the identifier of the required intermediate data as well as information on the source topology subgraph Gamma and the target computation stage S3-A.
[0049] In some embodiments, after receiving a data request from the Alpha subtask coordinating node of the topology subgraph, the collaborative management device parses the intermediate data identifier in the data request. Based on the global device topology map and the collaborative analysis task decomposition mapping relationship it maintains, the collaborative management device locates the subtask coordinating node of the Gamma topology subgraph that holds the required intermediate data. This subtask coordinating node is the node with the smallest load value within the Gamma topology subgraph. The collaborative management device then establishes a direct data channel between the Alpha and Gamma subtask coordinating nodes. The establishment process involves negotiating and opening a dedicated data transmission link between the network addresses of the two edge computing devices. The data channel establishment time is related to the network hop count and the current load of the node.
[0050] It is understandable that after the collaborative management device completes the data channel establishment, the subtask coordination node of the Gamma topology will directly transmit the required intermediate data generated by the BHP-11 node of the lithium bromide heat pump device within the Gamma topology to the subtask coordination node of the Alpha topology through this data channel. The subtask coordination node of Alpha, as the receiver, after obtaining the intermediate data from the Gamma topology through the data channel, injects this intermediate data into the designated computation stage of the subtask execution chain it maintains, that is, provides this data as an input parameter to the computation stage S3-A that is about to be executed or is waiting to be executed. Only after obtaining the required cross-topology subgraph intermediate data can the edge computing device corresponding to computation stage S3-A execute the logic of computation stage S3-A in conjunction with its local state dataset, generate new intermediate computation results, and continue to pass them along the subtask execution chain of Alpha.
[0051] Optionally, if multiple data dependencies exist across topological subgraphs in the collaborative analysis task, the collaborative management device will process data requests from multiple subtask coordination nodes in parallel and establish an independent data channel for each pair of subtask coordination nodes with data supply and demand relationships. In some embodiments, the data channel supports bidirectional or unidirectional data transmission, and the specific mode is configured by the collaborative management device according to the directionality of the data request. For example, data transmission from topological subgraph Gamma to topological subgraph Alpha is configured as a unidirectional channel, allowing only subtask coordination nodes of topological subgraph Gamma to send data to subtask coordination nodes of topological subgraph Alpha.
[0052] At the data comparison level, without the cross-topology subgraph data request and channel establishment mechanism described in this embodiment, when external data is needed during the computation phase, all intermediate data must first be uploaded to the collaborative management device for centralized storage, and then distributed to the requesting party by the collaborative management device. This will create a data bottleneck at the collaborative management device and increase transmission latency. With this embodiment, the collaborative management device only acts as a coordinator for data request routing and channel establishment; actual data transmission occurs directly between subtask coordination nodes, avoiding centralized data transfer at the collaborative management device. It can be understood that by locating and establishing direct data channels between subtask coordination nodes through the collaborative management device, this embodiment achieves on-demand, point-to-point data transmission across topology subgraphs, resulting in shorter data flow paths and reducing unnecessary network traffic and the consumption of storage and bandwidth resources on the collaborative management device.
[0053] In one embodiment of the present invention, continuing the scenario of a chemical plant waste heat recovery system, topological subgraphs Alpha, Beta, and Gamma complete all calculation stages of their internal subtask execution chains under the drive of their respective subtask coordination nodes. After each topological subgraph completes its calculation, its subtask coordination node integrates the final intermediate calculation results generated by all calculation stages within the topological subgraph to generate a topological subgraph result that characterizes the overall operating status of the topological subgraph. For example, the subtask coordination node of topological subgraph Alpha generates a topological subgraph result R_Alpha, which may be a vector or matrix containing the comprehensive heat recovery efficiency improvement value of the four lithium bromide heat pump devices within topological subgraph Alpha. The subtask coordination node of topological subgraph Alpha uploads the topological subgraph result R_Alpha to the collaborative management device. Similarly, the subtask coordination node of the topology subgraph Beta generates the topology subgraph result R_Beta, and the subtask coordination node of the topology subgraph Gamma generates the topology subgraph result R_Gamma. The topology subgraph results R_Beta and R_Gamma are then uploaded to the collaborative management device.
[0054] In some embodiments, the collaborative management device collects all topology subgraph results, including topology subgraph results R_Alpha, R_Beta, and R_Gamma. The collaborative management device fuses the results of each topology subgraph based on a global device topology map to generate a global analysis result. The global device topology map defines the connection relationships between topology subgraphs, and the fusion process needs to determine the contribution and mutual influence of each topology subgraph result at the global level based on these connection relationships. The collaborative management device calculates the load value of each node based on the historical computing power consumption data and real-time operating status data of the edge computing power device corresponding to each node. Historical computing power consumption data refers to the record of computing resources consumed by the edge computing power device in processing computing tasks within a defined time window, such as the average CPU cycles per second usage in the past hour. Real-time operating status data refers to the current operating indicators of the edge computing power device, such as the current CPU utilization and memory utilization. The collaborative management device uses an aggregation function to calculate the load value, which combines historical consumption trends with real-time status.
[0055] It is understandable that after collecting the results of each topology subgraph, the collaborative management device performs fusion based on the global device topology graph. The fusion operation is not simply concatenating the topology subgraph results R_Alpha, R_Beta, and R_Gamma; rather, it needs to consider the physical or logical coupling strength represented by the connecting edges between the topology subgraphs. For example, in the global device topology graph, there are two connecting edges between topology subgraphs Alpha and Beta, one connecting edge between Beta and Gamma, and no direct connecting edge between Alpha and Gamma. Based on these connections, the collaborative management device constructs a graph model describing the dependencies between the topology subgraph results and executes the fusion algorithm on this model, ultimately outputting a global analysis result applicable to the entire lithium bromide heat pump equipment network of the plant. This global analysis result may be a system-wide optimization strategy vector or an overall energy efficiency assessment index.
[0056] Optionally, at the data comparison level, if the method described in this embodiment—where the subtask coordination node generates the topology subgraph results and uploads them, then the collaborative management device merges them—is not adopted, the traditional method might require the collaborative management device to directly receive the original intermediate calculation results or final state data from all twelve nodes, and then centrally perform full, unaggregated global calculations. With this embodiment, the initial calculations and aggregations within the topology subgraph are completed by the subtask coordination node at the edge, and the collaborative management device only needs to process three highly aggregated topology subgraph results. It can be understood that the method of calculating the load value based on the historical computing power consumption data and real-time operating status data of the edge computing device corresponding to each node provides a quantitative basis for selecting the node with the lightest load to assume the responsibility of subtask coordination. In terms of data traffic, the topology subgraph results, as aggregated data, have a much smaller data volume than the sum of the original intermediate calculation results from all nodes. This significantly reduces the amount of data uploaded to the collaborative management device, lowering network transmission pressure and the data processing overhead of the collaborative management device. In terms of computational pressure, the global-level fusion computation is based on the aggregated topological subgraph results, which has a lower computational complexity than directly processing the raw data of all nodes, thus reducing the central computing burden of the collaborative management device.
[0057] In one embodiment of the present invention, in a specific implementation, after the collaborative management device collects the topology subgraph results R_Alpha from topology subgraph Alpha, R_Beta from topology subgraph Beta, and R_Gamma from topology subgraph Gamma, it needs to fuse these topology subgraph results based on the global device topology map to generate a global analysis result. The collaborative management device determines the fusion weight between each topology subgraph result based on the connection attributes of the edges in the global device topology map. The connection attributes include connection type and connection strength; the connection type describes the physical or logical connection between topology subgraphs, such as "direct connection of steam pipes," "shared high-temperature wastewater source," or "adjacent heating network loops"; the connection strength is a quantitative value used to characterize the tightness of the association, which can be calculated based on factors such as pipe diameter, flow rate, temperature difference, or process coupling degree. The collaborative management device maintains a mapping table recording the connection relationships and attributes between topology subgraphs, and assigns fusion weights to each pair of connected topology subgraph results according to this table. See Table 1 for a description of the connection relationships and weights between topology subgraphs:
[0058] Table 1: Topological Subgraph Connection Attributes and Fusion Weights
[0059]
[0060] In some embodiments, the collaborative management device performs a weighted calculation on the topology subgraph results according to the fusion weights determined in the table. The weighted calculation is not a simple linear weighting, but rather considers the mutual influence between the topology subgraph results. The calculation process can be represented as allocating and superimposing a topology subgraph result according to its influence weights on other topology subgraphs. The collaborative management device uses a fusion function, whose inputs are all topology subgraph results and their pairwise fusion weights. The formula for calculating the fusion weights is expressed as:
[0061]
[0062] in: Represents the topological subgraph Results to topological subgraph The fusion weights of the results Representing a topological subgraph With topological subgraph The type influence coefficient corresponding to the connection type between them. Representing a topological subgraph With topological subgraph The connection strength value between them, function This represents specific rules for calculating weights based on connection type and connection strength, such as weighted product or rule-based mapping. The collaborative management device calculates weights based on fusion weights. Weighted calculations are performed on the topological subgraph results, for example, on a specific component of the global analysis results. It may be due to the corresponding components of each topological subgraph result. The weighted results are calculated based on the weights and their associated weights. After the weighted calculation is completed, the collaborative management device normalizes the weighted calculation results to ensure the scale consistency and interpretability of the output results, ultimately obtaining the global analysis results. The global analysis results may be an overall energy efficiency index or a vector containing operational recommendations for each topology subgraph.
[0063] It is understandable that while the edge computing device performs data acquisition and local computation, it periodically checks the data integrity of its local state dataset. For example, an edge computing device deployed near the lithium bromide heat pump unit BHP-05 checks every minute the sixty operational status data records that should have been collected in the past minute, based on the acquisition sequence and data fields specified in the data acquisition template. When missing data is detected, such as the absence of temperature data records at timestamp T1, the edge computing device immediately attempts to re-acquire the data for the missing period according to the acquisition sequence in the data acquisition template, i.e., it re-initiates a data read request to the sensors of the lithium bromide heat pump unit BHP-05. If the re-acquisition fails or cannot be obtained due to a momentary equipment failure, the edge computing device uses an interpolation imputation method, utilizing the valid data from adjacent times T0 and T2 before and after the missing time T1, to calculate an estimated value for time T1 using a linear interpolation algorithm and fill it into the local state dataset.
[0064] Optionally, after data integrity requirements are met—that is, all missing data within a check cycle has been processed through re-acquisition or interpolation imputation—the edge computing device compresses the local state dataset. The compression algorithm can be lossy or lossless; for example, a rotating door compression algorithm can be used for time-series data such as temperature and pressure, reducing the number of storage points while maintaining data trend characteristics. The edge computing device synchronizes the metadata of the compressed local state dataset to the collaborative management device. The metadata includes the dataset's time range, compressed size, key statistics, and data integrity identifiers. The collaborative management device receives and updates the metadata catalogs of each edge computing device, but does not store the complete local state dataset, thus maintaining a global data view.
[0065] At the data comparison level, if the method of determining fusion weights based on connection attributes in this embodiment is not adopted, the generation of global analysis results may only rely on the simple averaging or equal-weight superposition of topology subgraph results, ignoring the different effects brought about by actual physical connections and process coupling, resulting in deviations between the analysis results and the actual operating state of the system. After adopting this embodiment, the fusion weights accurately reflect the actual correlation strength between different parts of the equipment, making the global analysis results more physically meaningful and valuable for guidance. At the data processing level, if the edge computing device does not perform periodic data integrity checks and repairs, the local state dataset may have gaps. Collaborative analysis based on incomplete data will lead to distortion of intermediate calculation results, and errors will be amplified as they are passed along the computation chain. The edge computing device completes data integrity repair locally, ensuring the quality of the data used for computation. At the same time, the mechanism of local data compression and only synchronizing metadata, compared with the scheme of continuously uploading all raw data to the collaborative management device, significantly reduces the amount of data transmitted over the network, reduces the continuous occupation of edge network bandwidth, and also alleviates the data storage pressure on the collaborative management device.
[0066] See Figure 4 This is a grouped bar chart showing the load distribution of edge computing devices in a lithium bromide heat pump collaborative management system. It primarily displays the computing load composition and threshold matching of each node. The total load of all nodes is below the threshold of 0.5, with node 4 having the lowest total load, making it the optimal candidate for subtask coordination. The grouped bar chart clearly distinguishes the two dimensions of load composition, and the threshold line provides an intuitive way to filter low-load nodes. This type of chart is used in the collaborative task scheduling phase to help determine the subtask coordination nodes in the topology subgraph, ensuring balanced utilization of distributed computing power. This chart reflects the "threshold constraint" logic of computing load distribution: selecting only nodes with a total load below the threshold as coordination nodes avoids task delays caused by computing overload and improves the efficiency of distributed collaborative analysis.
[0067] In one embodiment of the present invention, in a specific implementation, the subtask coordination nodes of the topology subgraph Alpha are scheduled according to a predetermined subtask execution chain order. Before activating the edge computing device of the corresponding node in the topology subgraph Alpha to execute the computation stage, the subtask coordination node verifies the data integrity status of the target edge computing device. For example, when the subtask execution chain is about to execute the computation stage S3-A (corresponding to the lithium bromide heat pump device BHP-03 node), the subtask coordination node of the topology subgraph Alpha (the edge computing device corresponding to the lithium bromide heat pump device BHP-02) first sends a status query instruction to the edge computing device corresponding to the lithium bromide heat pump device BHP-03. After receiving the query instruction, the edge computing device corresponding to the lithium bromide heat pump device BHP-03 checks the data integrity status of its local status dataset and replies with a status report. The status report contains a flag indicating the data integrity status and a detailed description, such as the flag "complete" or "incomplete".
[0068] In some embodiments, the assessment of data integrity status can be calculated based on a quantified integrity ratio, which is the ratio of the number of valid data points actually collected within the most recent defined time window to the theoretical number of data points that should be collected according to the data collection template. The formula for calculating the data integrity ratio is as follows:
[0069]
[0070] in: Indicates the percentage of data integrity. This indicates the number of valid operational status data records actually collected and stored in the local status dataset by the edge computing device within the set evaluation time window. This represents the total number of operational status data records that should theoretically be collected within the same evaluation time window, based on the collection sequence specified in the data collection template. (When the integrity ratio...) When the data integrity status of the edge computing device is below a preset integrity threshold, it is determined to be "incomplete".
[0071] If the edge computing device corresponding to the lithium bromide heat pump equipment BHP-03 replies with an incomplete data integrity status, such as an integrity ratio... The integrity score is only 0.85, lower than the preset integrity threshold of 0.95. Therefore, the subtask coordination node of the Alpha topology subgraph will delay the activation of the S3-A computation phase. The subtask coordination node sends a response message containing a priority execution instruction to the edge computing device corresponding to the BHP-03 lithium bromide heat pump device, instructing the edge computing device to prioritize data integrity repair. Upon receiving the instruction, the edge computing device corresponding to the BHP-03, based on the acquisition sequence in the data acquisition template, initiates re-acquisition of data for missing time periods or executes an interpolation filling algorithm until the data integrity score is reached. Once the integrity threshold is reached or exceeded, the data integrity status is updated to "complete," and then the subtask coordination node of the topology subgraph Alpha is notified.
[0072] Understandably, after all subtask execution chains in all topology subgraphs have been successfully completed, and the collaborative management device has fused the results of each topology subgraph to generate the global analysis result, the collaborative management device maps the global analysis result back to the global device topology graph. The mapping process involves associating the conclusions or optimization indicators in the global analysis result with the specific affected lithium bromide heat pump device nodes in the global device topology graph. The collaborative management device generates a control parameter adjustment suggestion for each node. This suggestion is a specific and executable set of instructions. For example, for the lithium bromide heat pump device BHP-07 node, the control parameter adjustment suggestion might be "increase the generator operating temperature setpoint by 1.5 degrees Celsius" and "reduce the solution circulation pump frequency by 2 Hz".
[0073] The collaborative management device distributes the generated control parameter adjustment suggestions to the corresponding edge computing devices via the command distribution channel. For example, the collaborative management device sends the control parameter adjustment suggestions for the lithium bromide heat pump device BHP-07 node to the edge computing device corresponding to BHP-07. The edge computing device corresponding to BHP-07 parses the control parameter adjustment suggestions locally and drives the local lithium bromide heat pump device BHP-07 to perform specific parameter adjustment operations through its interface with the control system of the lithium bromide heat pump device BHP-07.
[0074] Optionally, in some embodiments, when the collaborative management device generates control parameter adjustment suggestions for a node, it will fine-tune the suggestions by combining the node's recent operating history and global analysis results to ensure the stability and security of the suggestions. After the edge computing device performs parameter adjustments, it will send the adjustment confirmation feedback and the initial operating status data under the new parameters back to the collaborative management device, which will then update the node's status in the global device topology diagram accordingly. At the data comparison level, if the mechanism of verifying the data integrity status of the subtask coordination node and delaying activation when incomplete is not adopted in this embodiment, the calculation phase may be directly executed based on the local state dataset with missing or inconsistent data. This will lead to deviations or even errors in the intermediate calculation results, which will propagate along the subtask execution chain and ultimately contaminate the topology subgraph results and the global analysis results. With this embodiment, data quality control is enforced before the calculation starts, ensuring the reliability of the local state dataset participating in the calculation. At the control level, if the collaborative management device only generates global analysis results without mapping them to specific, node-level control parameter adjustment suggestions, on-site operators need to manually interpret the global analysis results and manually adjust each device, which is slow and prone to errors. By adopting this implementation method, the closed loop from global analysis to specific device execution is fully automated. The collaborative management device generates accurate adjustment suggestions, which are directly driven by the edge computing device to execute. This realizes a closed loop from collaborative analysis to collaborative control, improving the automation level and response time of the overall system adjustment.
[0075] See Figure 5 This is a bar chart showing the data integrity ratio of a local status dataset for lithium bromide heat pump equipment. It primarily displays the data integrity status and threshold compliance status of each device. By comparing the bar height with the threshold line, it visually distinguishes between devices with complete data and those requiring repair. This type of chart is used in the local data processing stage to help identify devices with substandard data quality and guide edge computing devices to prioritize data integrity repair operations. This chart reflects the "quality constraint" logic of data integrity: only devices with compliant data integrity can participate in collaborative analysis, avoiding computational biases caused by incomplete data and ensuring the reliability of the overall analysis results.
[0076] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power, characterized in that, The system includes multiple edge computing devices and a collaborative management device. Each edge computing device is deployed near a lithium bromide heat pump device to collect the operating status data of the corresponding lithium bromide heat pump device. The collaborative management device establishes a global device topology graph. Nodes in the global device topology graph correspond to a lithium bromide heat pump device and its edge computing device. Edges in the global device topology graph represent the physical or logical connection relationship between two lithium bromide heat pump devices. The collaborative management device issues a data acquisition template to each edge computing device. The data acquisition template specifies the acquisition type and timing of the operating status data. The edge computing device acquires and stores the operating status data based on the data acquisition template to form a local status dataset. The collaborative management device initiates a collaborative analysis task based on the global device topology map. The collaborative analysis task includes multiple calculation stages. Each calculation stage is assigned to a node in the global device topology map. The edge computing device corresponding to the node executes the calculation stage. The calculation stage uses the local state dataset and intermediate data transmitted from adjacent nodes in the global device topology map to generate intermediate calculation results and transmits the intermediate calculation results to the next node specified by the collaborative management device. Before initiating a collaborative analysis task, the collaborative management device divides the global device topology graph to obtain multiple topology subgraphs. Each topology subgraph contains at least one node, and the nodes within the topology subgraphs are directly connected by edges. The collaborative management device assigns a set of computation stages to each topology subgraph, and the computation stages in the set of computation stages correspond one-to-one with the nodes in the topology subgraph. The collaborative management device sends the start command of the collaborative analysis task and the corresponding set of calculation stages to the node with the smallest load value in each topology subgraph. This node is designated as the subtask coordination node of the topology subgraph. After receiving the start command and the set of computation stages, the subtask coordination node determines the execution order of each computation stage in the set of computation stages according to the connection relationship between nodes in the topology subgraph, thus forming a subtask execution chain. The subtask coordination node activates the edge computing devices of the corresponding nodes in the topology subgraph to perform the calculation phase in sequence according to the subtask execution chain. When the edge computing device completes its calculation phase, the intermediate calculation results generated are passed along the subtask execution chain to the edge computing device corresponding to the next calculation phase. When the calculation phase requires intermediate data from other sub-topologies in the global device topology graph, the subtask coordination node sends a data request to the collaborative management device. The collaborative management device locates the subtask coordination nodes of other topological subgraphs that hold the required intermediate data according to the data request, and establishes a data channel between the current subtask coordination node and the target subtask coordination node. Data is transmitted through the data channel, and the receiving subtask coordination node injects the received intermediate data into the specified computation stage of the subtask execution chain it maintains. After each topology subgraph is calculated, its subtask coordination node will generate a topology subgraph result and upload the topology subgraph result to the collaborative management device. The collaborative management device collects all topology subgraph results and merges the results of each topology subgraph based on the global device topology graph to generate global analysis results.
2. The lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power according to claim 1, characterized in that, The collaborative management device calculates the load value of each node based on the historical computing power consumption data and real-time operating status data of the edge computing power device corresponding to each node. The node with the lowest load value is defined as one in which the sum of the average of its historical computing power consumption data and the current computing power utilization rate reflected by its real-time operating status data is lower than a preset load threshold.
3. The lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power according to claim 1, characterized in that, The collaborative management device determines the fusion weight between the results of each topology subgraph based on the connection attributes of the edges in the global device topology graph. The connection attributes include connection type and connection strength. The collaborative management device performs weighted calculations on the topology subgraph results according to the fusion weights, and then normalizes the weighted calculation results to obtain the global analysis results.
4. The lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power according to claim 1, characterized in that, The edge computing device periodically checks the data integrity of the local status dataset. When data is missing, the edge computing device re-acquires or interpolates to fill in the missing data according to the acquisition sequence in the data acquisition template. Once the data integrity requirements are met, the edge computing device compresses the local state dataset and synchronizes the metadata of the compressed local state dataset to the collaborative management device.
5. The lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power according to claim 1, characterized in that, Before activating the edge computing device of a node within its topology subgraph to perform the computation phase, the subtask coordination node will verify the data integrity status of the edge computing device. If the data integrity status is incomplete, the subtask coordination node will delay the activation of this computation phase and instruct the edge computing device to prioritize the data integrity repair operation.
6. The lithium bromide heat pump waste heat recovery collaborative management system based on distributed computing power according to claim 1, characterized in that, The collaborative management device maps the global analysis results back to the global device topology map and generates a control parameter adjustment suggestion for each node. The collaborative management device sends control parameter adjustment suggestions to the edge computing devices of the corresponding nodes, and the edge computing devices drive the local lithium bromide heat pump equipment to perform parameter adjustments.