Real-time monitoring system for the entire lifecycle of generator sets based on fog computing

Abstract

This invention discloses a real-time monitoring system for the entire lifecycle of a generator set based on fog computing, relating to the field of periodic monitoring. The method includes: acquiring monitoring network information of the monitoring network corresponding to the target generator set; filtering nodes based on the monitoring network information to determine potential node topologies and monitoring node topologies; calculating the node health entropy of each monitoring node and each potential node to obtain a bidirectional entropy set; making dynamic monitoring decisions based on a preset optimization algorithm, according to the bidirectional entropy set, the potential node topologies, and the monitoring node topologies, to generate a dynamic monitoring scheme; and applying the dynamic monitoring scheme to the monitoring network of the target generator set to perform real-time monitoring. This invention solves the problems of resource waste, low reliability, and poor maintenance in traditional static monitoring systems.

Description

Technical Field

[0001] This invention relates to the field of periodic real-time monitoring, specifically to a real-time monitoring system for the entire lifecycle of generator sets based on fog computing. Background Technology

[0002] Fog computing is a decentralized computing infrastructure that extends the capabilities of the cloud to the network edge. It brings data processing and storage closer to the data source, reducing latency and the need for continuous data transfer to the cloud. Fog computing facilitates predictive maintenance, reduces downtime, and improves overall operational efficiency.

[0003] Existing generator set monitoring relies on traditional monitoring technologies, which are prone to problems such as improper maintenance and management, untimely maintenance, poor responsiveness, and low efficiency in fault repair. Therefore, a system capable of timely analysis and monitoring of generator set equipment faults is needed to improve the efficiency of generator set equipment maintenance. Summary of the Invention

[0004] This application provides a real-time monitoring system for the entire lifecycle of generator sets based on fog computing, which is used to address the problems of poor equipment maintenance and low fault maintenance efficiency in the existing technology.

[0005] In view of the above problems, this application provides a real-time monitoring system for the entire life cycle of generator sets based on fog computing.

[0006] This application provides a real-time monitoring system for the entire lifecycle of generator sets based on fog computing, including:

[0007] The monitoring node topology construction module is used to obtain the monitoring network information of the monitoring network corresponding to the target generator set, and to perform node screening based on the monitoring network information to determine the potential node topology and the monitoring node topology. The monitoring network includes multiple monitoring nodes.

[0008] The bidirectional entropy acquisition module is used to acquire historical monitoring logs, parse the historical monitoring logs, calculate the node health entropy of each monitoring node and each potential node, and acquire a bidirectional entropy set, wherein the bidirectional entropy set includes a subset of the health entropy of the monitored object and a subset of the health entropy of the monitored subject.

[0009] The dynamic monitoring scheme generation module is used to generate a dynamic monitoring scheme based on a preset optimization algorithm, the bidirectional entropy set, the potential node topology, and the monitoring node topology, and to make dynamic monitoring decisions.

[0010] The dynamic monitoring scheme execution module is used to apply the dynamic monitoring scheme to the monitoring network of the target generator set and perform real-time monitoring.

[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0012] This application proposes a real-time monitoring system for the entire lifecycle of generator sets based on fog computing. Through machine learning algorithms, it improves the system's fault monitoring effectiveness and fault prediction and analysis capabilities. Compared to traditional manual monitoring methods, this invention, through the fog computing-based real-time monitoring system, assesses faulty components of generator sets, solving the problem of low maintenance efficiency. It avoids interference caused by improper human operation. Combined with the fog computing-based real-time monitoring system, it enables digital and transparent monitoring of generator set equipment, significantly improving the efficiency and effectiveness of equipment management, allowing for targeted maintenance and management, and effectively reducing maintenance costs. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of a real-time monitoring system for the entire lifecycle of a generator set based on fog computing.

[0015] Figure 2 This is a flowchart of the steps in the execution module of the dynamic monitoring scheme of a real-time monitoring system for the entire lifecycle of generator sets based on fog computing.

[0016] In the attached diagram, the labels represent the following: 11 Monitoring node topology construction module; 12 Bidirectional entropy acquisition module; 13 Dynamic monitoring scheme generation module; 14 Dynamic monitoring scheme execution module. Detailed Implementation

[0017] This application provides a real-time monitoring system for the entire lifecycle of generator sets based on fog computing. By using fog computing and machine learning algorithms, the system monitors generator sets throughout their entire lifecycle, thus solving the problem of low monitoring efficiency.

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0019] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0020] The present invention will now be described in detail with reference to the accompanying drawings.

[0021] like Figure 1 As shown, the real-time monitoring system for the entire lifecycle of a generator set based on fog computing in this application includes:

[0022] The monitoring node topology construction module 11 acquires the monitoring network information of the monitoring network corresponding to the target generator set, and performs node screening based on the monitoring network information to determine the potential node topology and the monitoring node topology. The monitoring network includes multiple monitoring nodes.

[0023] In this embodiment, monitoring devices with the potential to act as fog nodes are selected as potential nodes. Monitoring nodes can be divided into potential nodes and monitoring nodes. Potential nodes refer to devices with computing power that can be selected as fog nodes, such as embedded gateways. Monitoring nodes refer to devices that can directly monitor generator set components, such as temperature sensors and pressure sensors. Node selection is based on device type, computing power, location, etc., and ultimately forms a topology, i.e., network connection relationships.

[0024] For example, in a generator set, the monitoring network may include 100 monitoring nodes, of which 20 have strong computing capabilities, such as ARM industrial control computers, and are identified as potential nodes, forming a potential node topology; the remaining 80 are monitoring nodes, forming a monitoring node topology.

[0025] Compared with traditional technologies that use monitoring network information to screen nodes, classify the obtained monitoring nodes and construct the topology, the embodiments of this application avoid the technical problem of inaccurate monitoring effect of generator monitoring system due to inaccurate data source, improve data reliability and utilization, and provide accurate data analysis.

[0026] The bidirectional entropy acquisition module 12 acquires historical monitoring logs, parses the historical monitoring logs, calculates the node health entropy of each monitoring node and each potential node, and acquires a bidirectional entropy set, wherein the bidirectional entropy set includes a subset of the health entropy of the monitoring object and a subset of the health entropy of the monitoring subject.

[0027] In this embodiment, the first dimension is the monitoring node entropy, i.e., the entropy of the monitoring object corresponding to each monitoring node, which can represent the disorder of the health state of the generator subsystem or component. The second dimension is the node health entropy, i.e., the entropy of the monitoring subject corresponding to each potential node, which can represent the disorder of the health state of each candidate fog computing device. The higher the node health entropy value, the greater the uncertainty and instability of the system. High monitoring object health entropy indicates potential faults in generator components, such as turbine bearing failure; high monitoring subject health entropy indicates unstable performance of fog node devices, such as unstable edge servers.

[0028] In one embodiment, the bidirectional entropy acquisition module 12 is used for:

[0029] Acquire the intrinsic monitoring performance data of each monitoring node, and parse the historical monitoring logs to extract the historical monitoring time-series data of the monitoring object corresponding to each monitoring node;

[0030] Based on the intrinsic monitoring performance data of the nodes, the intrinsic health entropy of each monitoring node is calculated.

[0031] Based on the historical monitoring time series data, calculate the object health entropy for each monitoring node;

[0032] The intrinsic health entropy and the object health entropy are weighted to obtain the monitoring object health entropy of each monitoring node, and the output is a subset of the monitoring object health entropy.

[0033] Specifically, the intrinsic monitoring performance data of each monitoring node is acquired, such as monitoring accuracy and response time. Historical monitoring logs are then analyzed to extract historical monitoring time-series data, such as the time series of sensor readings. Entropy values ​​are calculated using the intrinsic monitoring performance data and historical monitoring time-series data to obtain intrinsic health entropy and object health entropy. These two values ​​are then weighted and fused to obtain the monitored object health entropy. Intrinsic health entropy reflects the health status of the monitoring node, such as sensor calibration status. Object health entropy reflects the health status of the monitored object, such as component data volatility. Through weighted fusion, the performance of the monitoring node and the health status of the corresponding monitored object can be comprehensively considered. The resulting monitored object health entropy reflects the health status of the monitored object corresponding to the monitoring node. A higher monitored object health entropy value indicates greater uncertainty in the monitored object's health status and a higher probability of failure. In addition, when performing weighted calculations, weights are set according to the influence of object health entropy and intrinsic health entropy. If object health is more important, the weight of object health entropy is set to be larger. The weight of intrinsic health entropy is set to: 1 - object health entropy, where, monitored object health entropy = [(object health entropy × object health entropy weight) + (intrinsic health entropy × intrinsic health entropy weight)], so that the calculated monitored object health entropy can better reflect the component status.

[0034] For example, the intrinsic performance data of a temperature sensor node includes: accuracy ±0.1℃ and mean time between failures (MTBF) of 1000 hours. The intrinsic health entropy of the node, calculated using the information entropy formula, is 0.68. The historical monitoring time series data is a temperature deviation time series, and the object health entropy, calculated using the approximate entropy formula, is 0.7, where a high entropy value indicates an abnormal temperature. If the object health entropy weight is set to 0.7, the weight of the intrinsic health entropy of the node is: (1-0.7)=0.3, and the monitored object health entropy = [(0.68×0.3)+(0.7×0.7)]=0.694.

[0035] In one embodiment, the bidirectional entropy acquisition module 12 is further configured to:

[0036] Analyze historical monitoring logs and extract historical node performance time-series data for each potential node.

[0037] Based on the historical node performance time-series data and combined with preset execution constraints, the typical computing power information and typical throughput information of each potential node are calculated and obtained.

[0038] Based on the historical node performance time-series data, the typical computing power information and typical throughput information of the nodes, the computing power fluctuation entropy and throughput fluctuation entropy of each potential node are calculated traversally.

[0039] The weighted computing power fluctuation entropy and throughput fluctuation entropy are used to output a subset of the health entropy of the monitored entity.

[0040] The set outputs the subset of health entropy of the monitored object and the subset of health entropy of the monitored subject, which together form the bidirectional entropy set.

[0041] Specifically, historical monitoring logs are parsed to extract historical node performance time-series data for potential nodes, such as CPU utilization and memory utilization. Based on historical data and execution constraints, such as maximum computing power, typical node computing power information, such as average computing power, and typical node throughput information, such as data processing rate, are calculated. Computing power fluctuation entropy and throughput fluctuation entropy are then calculated, and finally, through weighted fusion, the health entropy of the monitored entity is obtained. Computing power fluctuation entropy quantifies the stability of computing power; a higher value indicates greater fluctuations in CPU utilization. Throughput fluctuation entropy quantifies the stability of data transmission; a higher value indicates greater fluctuations in network throughput. During weighted fusion, appropriate weights are set according to the degree of influence of computing power fluctuation entropy and throughput fluctuation entropy. If typical node computing power information is more important, the weight of computing power fluctuation entropy is set higher. The weight of throughput fluctuation entropy is set as: 1 - computing power fluctuation entropy. The health entropy of the monitored entity = [(computing power fluctuation entropy × computing power fluctuation entropy weight) + (throughput fluctuation entropy × throughput fluctuation entropy weight)], which reflects the overall reliability of potential nodes. Finally, a bidirectional entropy set can be obtained by combining the subset of health entropy of the monitored object with the subset of health entropy of the monitored subject. The bidirectional entropy set is used to constrain the health of the monitored object and the health of the monitored subject to ensure that the calculation is within a reasonable range, for example, the maximum computing power of the node is 100 TFlops.

[0042] For example, a potential node's historical CPU utilization fluctuates between 50% and 80%, with a typical average computing power of 80 TFlops. The computing power fluctuation entropy is 0.6. Historical network throughput fluctuates between 50 and 100 Mbps, with a typical average throughput of 75 Mbps. The throughput fluctuation entropy is 0.65. When weighting, if the node's typical computing power information is more critical, the computing power fluctuation entropy weight is set to 0.6, the throughput fluctuation entropy weight is (1-0.6)=0.4, and the monitoring entity's health entropy = [(0.6×0.6)+(0.65×0.4)]=0.62.

[0043] Compared to traditional technologies, this application's embodiments calculate intrinsic health entropy and object health entropy using historical node intrinsic monitoring performance data and historical monitoring time-series data. The weighted average of these data yields the monitored object health entropy for each monitoring node. Based on historical node performance time-series data, typical node computing power information, and typical node throughput information, the computing power fluctuation entropy and throughput fluctuation entropy for each potential node are calculated. This weighted average yields the monitored entity's health entropy. The calculated bidirectional entropy is used to simultaneously assess the health status of components and the reliability of nodes, achieving bidirectional assessment of the monitored object and its corresponding node. This avoids inaccurate analysis caused by incomplete monitoring data, which in turn leads to inaccurate real-time monitoring of equipment status. This improves equipment maintenance efficiency.

[0044] The dynamic monitoring scheme generation module 13 generates a dynamic monitoring scheme based on a preset optimization algorithm, by making dynamic monitoring decisions according to the bidirectional entropy set, the potential node topology, and the monitoring node topology.

[0045] In this embodiment, an optimization algorithm is used, based on the combined effect of bidirectional entropy sets, and a scheduling engine runs periodically. Combining the potential node topology with the monitoring node topology, fog nodes are dynamically selected for dynamic monitoring decisions, resulting in a reliable dynamic monitoring scheme. The scheme includes the selection of fog nodes and the monitoring mapping relationship between fog nodes and monitoring nodes.

[0046] In one embodiment, the dynamic monitoring scheme generation module 13 is used for:

[0047] Based on the bidirectional entropy set, initialize the by-the-round probability of multiple potential nodes and the capacity correction coefficient of multiple monitoring nodes;

[0048] By combining the by-by probability and the capacity correction coefficient, the potential node topology is iteratively and randomly selected to obtain a set of candidate fog node schemes;

[0049] The candidate fog node scheme set is iteratively optimized based on the preset fog node evaluation function, and the optimized result is the target fog node scheme. The fog node evaluation function includes a distribution uniformity factor, a capacity matching factor, a scheme average entropy factor, and a scheme total entropy factor.

[0050] Based on the target fog node scheme and the preset monitoring matching rules, a monitoring association decision is made to establish a monitoring mapping relationship between multiple target fog nodes and multiple monitoring nodes, and the monitoring mapping relationship and the target fog node scheme are associated and output as the dynamic monitoring scheme.

[0051] In this embodiment, based on the bidirectional entropy set, we can obtain the idle probability of multiple potential nodes and the capacity correction coefficient of multiple monitoring nodes. By randomly selecting potential nodes, a set of candidate fog node schemes is generated.

[0052] When optimizing alternative solutions, a fog node evaluation function is used to help select the best solution or path. This function includes a distribution uniformity factor, a capacity matching factor, a solution average entropy factor, and a solution total entropy factor. The evaluation function is used to quantify the merits of a state or decision solution. In artificial intelligence and machine learning, evaluation functions are commonly used in heuristic search, optimization problems, and decision analysis to ultimately output a target fog node solution. A monitoring mapping relationship is established based on monitoring matching rules. After the target fog node solution is determined, a mapping relationship is established from high-entropy monitoring nodes to high-capacity fog nodes. Subsequently, monitoring association decisions are made based on the target fog node solution and preset monitoring matching rules, establishing multiple monitoring mapping relationships between multiple target fog nodes and multiple monitoring nodes. The output monitoring mapping relationship and the target fog node solution constitute a dynamic monitoring scheme. The optimization algorithm minimizes the overall weighted quadratic entropy of the mapping relationship. The weighted quadratic entropy of each fog node can be defined as the weighted sum of the by-the-wheel probability and the capacity correction coefficient associated with the fog node.

[0053] The step of initializing the skip probability of multiple potential nodes and the capacity correction coefficient of multiple monitoring nodes based on the bidirectional entropy set includes:

[0054] Perform Z-score normalization on the health entropy subset of the monitored objects to obtain the capacity correction coefficients of multiple monitored nodes;

[0055] Perform min-max normalization on the subset of health entropy of the monitored entity to obtain the skip probability of multiple potential nodes.

[0056] Specifically, initialization is performed using Z-score normalization. Z-score (standard score) is used to assess the distance of a sample point from the population mean. Z-score primarily measures the number of standard deviations between the original data and the population mean. Z-score allows comparison of test results with normal results. Z-score=1 means the sample value exceeds the mean by 1 standard deviation; Z-score=2 means the sample value exceeds the mean by 2 standard deviations. Z-score indicates the position of the sample value on the normal distribution curve. Z-score=0 means the sample is exactly at the mean; Z-score=3 means the sample value is significantly higher than the mean. Normalizing the subsets of health entropy of the monitored objects and the subsets of health entropy of the monitored subjects respectively yields the capacity correction coefficient of the monitored nodes and the probability of potential nodes being skipped. The capacity correction coefficient is a coefficient C between 0 and 1. correction This is used to adjust the nominal computational / connectivity of nodes. And C correction With the health entropy H of the node nodeNegatively correlated, the higher the health entropy value of the node, the smaller the capacity correction coefficient. For example, for a monitoring node with a required computing power of 8 TFlops, the corresponding H node is relatively high, resulting in C correction = 0.7. Then, when scheduling and allocating tasks, it is only regarded as a node with a required computing power of 8÷0.7≈11.43 TFlops (rounded up) for processing. In this way, it is ensured that the allocated fog nodes have sufficient computing power redundancy, preventing node failures caused by computing overload, achieving refined management and load balancing, and further ensuring the stability of the system.

[0057] The idle probability (P exclude ) refers to the probability that each alternative fog node is excluded from the candidate list due to its own unreliability when being assigned a new task. It is positively correlated with the health entropy of the monitoring entity. The higher the health entropy value of the monitoring entity, the relatively lower the stability of the corresponding node can be considered, and the higher the idle probability, that is, P exclude is positively correlated with the health entropy H node of the node. Specifically, it can be mapped through a monotonically increasing function. For example: P exclude = 0 (when H node is less than the threshold T1, indicating that the node is very healthy); P exclude =(H node - T1) / (T2 - T1) (when T1 <= H node < T2, linear or non-linear growth); P exclude = 1 (when H node >= T2, the node is considered unavailable). Among them, the poorer the self-state of the node, the lower the probability of being selected, avoiding the possibility of assigning critical tasks to unreliable nodes from the source, and improving the overall stability of the system.

[0058] Among them, iteratively and randomly selecting the potential node topology by combining the idle probability and the capacity correction coefficient to obtain an alternative fog node solution set includes:

[0059] Randomly extracting in the potential node topology to obtain a first potential node;

[0060] Synchronously activating a random number generator to obtain a first random number N, where N is greater than or equal to 0 and less than or equal to 1;

[0061] Comparing the idle probability corresponding to the first potential node with the first random number N. If the first random number N is larger, adding the first potential node to the first alternative fog node solution;

[0062] Iteratively updating and selecting the first potential node until the node capacities of all the potential nodes in the first alternative fog node solution meet the monitoring demand capacity of the monitoring network;

[0063] The process involves iterative random sampling to obtain multiple candidate fog node schemes, which are then merged with the first candidate fog node scheme to output the candidate fog node scheme set.

[0064] Specifically, a first potential node is selected from the potential node topology using a random sampling method. A first random number N is generated in a random number generator. If the obtained first random number N is greater than the bye probability corresponding to the first potential node, the first potential node can be used as a candidate for a fog node.

[0065] For example, the probability of randomly selecting node A being left out is 0.3, and the generated first random number N = 0.6. Since N > 0.3, the corresponding first potential node can be added to the first alternative fog node scheme.

[0066] Furthermore, the first potential node is selected through iterative updates multiple times until the node capacity of all the potential nodes in the first candidate fog node scheme meets the monitoring requirement capacity of the monitoring network, i.e., the number of interfaces required by the monitoring node. When the monitoring requirement capacity of the monitoring network is reached, the iteration can be stopped.

[0067] For example, if the monitoring capacity requirement is 200Mbps, node A is randomly selected with a 0.3% chance of being left unselected, and node B has a 0.7% chance of being left unselected. When N=0.6, node A is selected (0.6>0.3), and node B is not selected (0.6<0.7). Nodes C and D are then selected until the total throughput is ≥500Mbps. This process is repeated multiple times until the number of selected potential nodes reaches 200Mbps, at which point selection stops.

[0068] The step of iteratively optimizing the candidate fog node scheme set based on a preset fog node evaluation function, and outputting the optimized result as the target fog node scheme, includes:

[0069] By combining the preset neighborhood constraints, the potential node topology and the monitoring node topology, the number of neighborhood monitoring nodes of multiple potential nodes in each candidate fog node scheme is obtained, and the mean square error value is calculated accordingly, and the output is the distribution uniformity factor.

[0070] Calculate the relative deviation between the sum of the typical throughput information of multiple potential nodes in each of the candidate fog node schemes and the monitoring demand capacity, and obtain the capacity matching factor of each of the candidate fog node schemes.

[0071] Sum the monitoring subject health entropy of multiple potential nodes in each candidate fog node scheme, and take the reciprocal to obtain the scheme total entropy factor of each candidate fog node scheme;

[0072] Calculate the ratio of the total entropy factor of the scheme to the sum of the typical throughput information of the nodes of the multiple potential nodes in each of the candidate fog node schemes, and obtain the average entropy factor of each of the candidate fog node schemes.

[0073] Multiple fog node evaluation values ​​are obtained by weighting the distribution uniformity factor, capacity matching factor, average entropy factor, and total entropy factor of each candidate fog node scheme;

[0074] The candidate fog node scheme set is iteratively updated and the corresponding fog node evaluation value is calculated. The fog node scheme with the largest corresponding fog node evaluation value is selected as the target fog node scheme.

[0075] Specifically, based on preset neighborhood constraints such as geographical distance, potential node topology and monitoring node topology, the number of neighboring monitoring nodes for each fog node is calculated, the mean square error is calculated, and the output is the distribution uniformity factor. A smaller mean square error indicates a more uniform distribution.

[0076] For example, a standard deviation of 0.05 indicates that the fog nodes are evenly distributed; if the standard deviation is 0.2, the fog nodes are not evenly distributed.

[0077] Furthermore, the relative deviation between the sum of typical throughput information of multiple potential nodes in each candidate fog node scheme and the monitoring demand capacity is calculated to obtain the capacity matching factor of each candidate fog node scheme. The formula for calculating the capacity matching factor is: (|sum of typical node throughput information - monitoring demand capacity|) / monitoring demand capacity. The smaller the value of the capacity matching factor, the smaller the relative deviation, and the better the match between the sum of typical node throughput information and the monitoring demand capacity.

[0078] For example, if the sum of the typical throughput information of a node is 360Mbps and the monitoring capacity requirement is 200Mbps, the capacity matching factor is (360-200) / 200=0.8.

[0079] Furthermore, the health entropy of the monitoring subject of multiple potential nodes in each candidate fog node scheme is summed, and the reciprocal is taken to obtain the total entropy factor of each candidate fog node scheme. Among them, the health entropy of the monitoring subject is negatively correlated with the total entropy factor of the scheme. The lower the health entropy value of the monitoring subject, the higher the total entropy factor value of the scheme.

[0080] For example, if the health entropy of the monitoring subject in the alternative fog node scheme is 0.6, the corresponding total entropy factor of the scheme is 1 / 0.62≈1.61.

[0081] Furthermore, the ratio of the total entropy factor of the scheme to the sum of the typical throughput information of multiple potential nodes in each candidate fog node scheme is calculated to obtain the scheme average entropy factor of each candidate fog node scheme. The calculation formula is: (total entropy factor of the scheme / sum of typical throughput information of nodes). The larger the ratio, the higher the credibility of the typical throughput information of the potential node per unit node.

[0082] For example, if the sum of the typical throughput information of a node is 50, the average entropy factor of the scheme is 1.61 / 50 = 0.03.

[0083] Furthermore, based on the influence of each candidate fog node scheme's distribution uniformity factor, capacity matching factor, average entropy factor, and total entropy factor on its evaluation as a fog node, corresponding weights are assigned, and a weighted fusion calculation is performed to obtain multiple fog node evaluation values. Fog node evaluation value = [(Distribution uniformity factor × Distribution uniformity factor weight) + (Capacity matching factor × Capacity matching factor weight) + (Scheme average entropy factor × Scheme average entropy factor weight) + (Scheme total entropy factor × Scheme total entropy factor weight)]. The greater the influence on the evaluated fog node, the larger its weight percentage, the higher the weighted fog node evaluation value, and the greater the probability that the corresponding potential node will become a fog node.

[0084] For example, the weights of the distribution uniformity factor, capacity matching factor, scheme average entropy factor, and scheme total entropy factor are set to 0.3, 0.15, 0.25, and 0.3, respectively. The fog node evaluation value is obtained as [(0.12×0.3)+(0.8×0.15)+(0.03×0.25)+(1.61×0.3)]≈0.35.

[0085] Furthermore, the set of candidate fog node schemes is iteratively updated and the corresponding fog node evaluation values ​​are calculated. The fog node scheme with the largest corresponding fog node evaluation value is selected as the optimal target fog node scheme.

[0086] For example, Scheme A: Fog nodes are unevenly distributed with a standard deviation of 0.2, a distribution uniformity factor of 0.12, a capacity matching factor of 0.15, a scheme average entropy factor of 0.03, and a scheme total entropy factor of 1.61, resulting in a fog node evaluation value of 0.35. Scheme B: Fog nodes are evenly distributed with a standard deviation of 0.05, a capacity matching degree of 0.95, a total entropy factor of 0.9, an average entropy factor of 0.85, and a fog node evaluation value of 0.9. Since 0.9 > 0.35, Scheme B is selected as the target scheme.

[0087] In one embodiment, the dynamic monitoring scheme generation module 13 is further configured to:

[0088] Real-time monitoring of network fluctuations in the monitoring network, and fluctuation identification, including:

[0089] If the fluctuation of the monitoring network is a fluctuation of the monitoring node, then based on the real-time target fog node scheme, verify whether the typical throughput information of the corresponding fog node meets the requirements of the monitoring node after the fluctuation.

[0090] If satisfied, the target fog node scheme is maintained in real time;

[0091] If not satisfied, then by combining the preset local adjustment constraints and the health entropy of the monitored object of the monitored node after fluctuation, local iterative optimization of the fog node is performed, and the target fog node scheme is updated based on the local iterative optimization results.

[0092] If the fluctuation of the monitoring network is a fog node fluctuation, then based on the health entropy of the monitoring subject of the fog node after the fluctuation, combined with the preset fog node evaluation function, global iterative optimization of the fog node is performed.

[0093] Specifically, if the network fluctuation is a monitoring node fluctuation, then based on the real-time target fog node scheme, the typical throughput information of the corresponding fog node is verified to ensure it meets the requirements of the monitoring node after the fluctuation. Monitoring node fluctuations can be caused by factors such as sensor failure or sensor loss. If the monitoring node meets the requirements after the fluctuation, the real-time target fog node scheme is maintained, and the resulting scheme is the optimal one, requiring no adjustment. If it does not meet the requirements, then by combining preset local adjustment constraints and the health entropy of the monitored object of the monitoring node after the fluctuation, local iterative optimization of the fog node is performed by adjusting a portion of the mapping, and the target fog node scheme is updated based on the results of the local iterative optimization.

[0094] For example, if a temperature sensor fails, the system checks if the fog nodes responsible for that area have redundant capacity. If so, the existing configuration is maintained; otherwise, the monitoring node is reassigned to another fog node within the corresponding area.

[0095] Furthermore, if the monitoring network fluctuation is a fog node fluctuation, then based on the monitoring subject's health entropy of the fog node after the fluctuation, combined with the preset fog node evaluation function, a global iterative optimization of the fog node is performed. If the fluctuation is a fog node fluctuation, such as a fog node crash, a global iterative optimization can be performed based on the updated monitoring subject's health entropy, and a new solution can be generated.

[0096] For example, if a fog node fails, the system recalculates the health entropy of all potential nodes and runs an optimization algorithm to generate a new solution.

[0097] Compared to traditional technologies, this embodiment obtains distribution uniformity factors, capacity matching factors, average entropy factors, and total entropy factors through evaluation functions. These factors are then weighted and fused to calculate the fog node evaluation value. By comparing and analyzing multiple evaluation functions from various aspects to balance distribution, capacity, and reliability, the obtained fog node evaluation value can be used to assess the value of the solution. A higher fog node evaluation value indicates a more suitable fog node solution. Simultaneously, by verifying the fog node solution, adjustments are made to enable the system to adapt. Local optimization reduces computational power consumption, thereby achieving global optimization and ensuring the generation of the optimal fog node solution. Thus, through precise computational power calculation and comparison, the target fog node solution exhibits high adaptability and accuracy, achieving precise management of the monitoring system and improving its reliability.

[0098] The dynamic monitoring scheme execution module 14 applies the dynamic monitoring scheme to the monitoring network of the target generator set to perform real-time monitoring.

[0099] In this embodiment of the application, the obtained dynamic monitoring scheme is applied to the monitoring network of the target generator set, which enables real-time monitoring and real-time monitoring of the entire life cycle of the target generator. The real-time monitoring system is used to achieve dynamic adaptation and efficient real-time data acquisition and processing.

[0100] In one embodiment, such as Figure 2 As shown, the dynamic monitoring scheme execution module 14 is used for:

[0101] The monitoring network of the target generator set is continuously monitored, and the health entropy is updated and calculated to identify fluctuations in the monitoring network.

[0102] If the monitoring network fluctuation is due to the monitoring device fluctuation, then a quick decision is first made under the original fog node, such as mixed integer linear programming, and the result of the quick decision is verified. If the verification is successful, the fog computing architecture is updated based on the result of the quick decision.

[0103] If the verification fails, the device health entropy and node health entropy within the local area are updated in combination with the preset local adjustment constraints, and a region-iterative decision analysis based on the optimization algorithm is performed.

[0104] If the monitored network fluctuations are fog node fluctuations, including active fog nodes and inactive / alternate fog nodes, then update the device health entropy and node health entropy and perform a global-iterative decision analysis based on an optimization algorithm.

[0105] Specifically, the monitoring network of the target generator set is continuously monitored using monitoring equipment such as vibration sensors, temperature sensors, and data sensors. The health entropy is then updated and calculated to identify fluctuations in the monitoring network. If the fluctuation is attributed to monitoring equipment fluctuations, a rapid decision-making process, such as mixed-integer linear programming (MILP), is first performed at the original fog node level. The results of this rapid decision are then validated. If the validation is successful, the fog computing architecture is updated based on the results. Monitoring equipment fluctuations refer to the addition, removal, or drastic performance changes of monitoring nodes, such as sensors and data acquisition devices, which can interfere with the data sources of generator set components. Mixed-integer linear programming (MILP) is an accurate mathematical optimization method that can quickly find the optimal task redistribution scheme and rapidly generate local repair strategies under the constraint of fixed computing resources, i.e., the original fog node set.

[0106] If the verification fails, the device health entropy and node health entropy within the local area are updated in conjunction with the preset local adjustment constraints. A region-iterative decision analysis based on the optimization algorithm is then performed. Within the local area, potential candidate nodes are reselected and activated as fog nodes, and the mapping relationship is re-established. If the monitored network fluctuations are fog node fluctuations, including both activated fog nodes and inactive / candidate fog nodes, it indicates a change in the data resources themselves. In this case, the optimization algorithm must be rerun from a global perspective, updating the device health entropy and node health entropy, performing a global-iterative decision analysis based on the optimization algorithm, and regenerating the dynamic monitoring scheme.

[0107] For example, the original monitoring scheme has 50 monitoring points: S1 (turbine temperature), S2 (bearing vibration), S3 (lubricating oil pressure), and S4 (exhaust temperature). The fog nodes are P1 (control cabinet industrial computer), P2 (field server), and P3 (gateway device). S1 and S2 are mapped to P1; S3 and S4 are mapped to P2. The required monitoring capacity is 200Mbps.

[0108] The bearing vibration sensor at monitoring node S2 experienced increased data noise due to aging wiring, causing the health entropy of the monitored object to rise from 0.3 to 0.8. Simultaneously, a new high-frequency acoustic monitoring sensor, S51, was added for precise bearing condition diagnosis. During fluctuation detection, the system identified this as a fluctuation in the monitoring equipment, with one node experiencing a drastic entropy change and another node being added. The system updated the health entropy, setting the health entropy of the monitored object at S2 to 0.8 and initializing a higher entropy value (e.g., 0.7) for S51. Further rapid decision-making was performed using a mixed-integer linear programming (MILP) model to reallocate the data analysis tasks of S2 and S51 while satisfying the existing capacity constraints of P1, P2, and P3. The model calculation revealed that since both S2 and S51 are high-entropy monitoring points, the remaining capacity of any single fog node in the original scheme could not simultaneously handle both tasks. The validation failed. Therefore, further region-iterative decision analysis was performed, and the system initiated local optimization. The original fog nodes P1 and P2, and the unselected potential candidate node P4, were updated. Based on the entropy values ​​of local monitoring nodes such as S1 and S51, an optimization algorithm is run on the potential node topology of [P1, P2, P4] and the monitoring node topology of [S2, S51], as well as other neighboring monitoring points, to output a new local target fog node scheme. For example, P4 is activated, the analysis task of S2 is migrated from P1 to P4, and S51 is assigned to P2. The system only updates the mapping relationship of this local area, while the mapping of other parts of the global system remains unchanged.

[0109] If fog node P2 experiences a heat dissipation failure, causing continuous CPU temperature alarms and a rapid increase in the monitored entity's health entropy from 0.4 to 0.9, maintenance personnel deploy fog node P5. The system identifies this as fog node fluctuation, indicating a node performance degradation and the addition of a backup node. The monitoring system updates the health entropy, setting P2's monitored entity health entropy to 0.9, indicating a very unhealthy device status. P5's monitored entity health entropy is initialized to 0.2, indicating a very healthy device status. Further, a global-iterative decision analysis is performed. The system identifies changes in computing resources, a major node nearing failure, and the addition of a new node, initiating a global optimization process. The potential node topology with the currently updated health entropy is obtained, including P1, P2, P3, P4, P5, and the monitoring node topology, encompassing all 50 monitoring points. Based on the latest bidirectional entropy set, the optimization algorithm is fully executed, initializing the idle probability, generating alternative solutions, iteratively optimizing the evaluation function, and outputting a new global dynamic monitoring scheme. The new scheme marks P2 as idle and activates P5. All monitoring nodes originally managed by P2 (such as S3 and S4) and other nodes that need adjustment are reassigned to P1, P3, P4, and P5, forming a completely new and globally optimal monitoring mapping relationship.

[0110] Compared to traditional technologies, this application's embodiments utilize a monitoring network that continuously monitors the target generator set. Under the original fog node conditions, rapid decision-making is achieved, and the decision results are verified using mixed-integer linear programming. Adjustments are made based on the verification results. When the monitoring network fluctuation is attributed to monitoring equipment fluctuation, a region-iterative decision analysis based on an optimization algorithm is performed. When the monitoring network fluctuation is attributed to fog node fluctuation, a global-iterative decision analysis based on an optimization algorithm is performed. Through bidirectional health entropy and optimization algorithms, dynamic optimization of the monitoring system is achieved, overcoming the shortcomings of traditional static monitoring systems such as resource waste and low reliability.

[0111] In summary, compared to traditional technologies, this embodiment uses a monitoring node topology construction module 11 to filter monitoring network information, classify the obtained monitoring nodes, and construct a topology. This avoids inaccurate monitoring results caused by inaccurate data sources, improves data reliability and utilization, and provides accurate data analysis. Using a bidirectional entropy acquisition module 12, intrinsic health entropy and object health entropy are calculated from historical node intrinsic monitoring performance data and historical monitoring time-series data. The weighted average of these data yields the monitoring object health entropy for each monitoring node. Based on historical node performance time-series data, typical node computing power information, and typical node throughput information, the computing power fluctuation entropy and throughput fluctuation entropy for each potential node are calculated. The weighted average of these data yields the monitoring subject health entropy. The calculated bidirectional entropy is used to simultaneously evaluate the health status of components and the reliability of nodes, achieving bidirectional evaluation of the monitored object and its corresponding node. This avoids inaccurate analysis caused by incomplete monitoring data analysis, which in turn leads to inaccurate real-time monitoring of equipment status. This improves equipment maintenance efficiency. Using the dynamic monitoring scheme generation module 13, the distribution uniformity factor, capacity matching factor, scheme average entropy factor, and scheme total entropy factor are obtained through evaluation functions. These factors are then weighted and fused to obtain the fog node evaluation value. Through multi-faceted comparison and analysis using multiple evaluation functions, distribution, capacity, and reliability are balanced. The obtained fog node evaluation value allows for scheme value assessment; a higher fog node evaluation value indicates a more suitable fog node scheme. Simultaneously, by verifying the fog node scheme, adjustments are made to enable the system to adapt. Local optimization reduces computational power consumption, thereby achieving global optimization and ensuring the generation of the optimal fog node scheme. Thus, through precise computational power calculation and comparison, the target fog node scheme demonstrates high adaptability and accuracy, achieving precise management of the monitoring system and improving its reliability. Using the dynamic monitoring scheme execution module 14, the monitoring network of the target generator unit is continuously monitored. Rapid decision-making is performed under the original fog node conditions, and mixed-integer linear programming is used to verify the decision results. Adjustments are made based on the verification results. When monitoring network fluctuations correspond to monitoring equipment fluctuations, a region-iterative decision analysis based on optimization algorithms is performed. When monitoring network fluctuations manifest as fog node fluctuations, a global-iterative decision analysis based on an optimization algorithm is performed. Dynamic optimization of the monitoring system is achieved through bidirectional health entropy and the optimization algorithm.

[0112] Compared to traditional technologies, the embodiments of this application avoid inaccurate data sources that could lead to inaccurate monitoring results in the generator set monitoring system, improving data reliability and utilization, providing accurate data analysis, and overcoming the shortcomings of traditional static monitoring systems such as resource waste and low reliability. It also avoids inaccurate analysis caused by incomplete monitoring data, which in turn leads to inaccurate real-time equipment status monitoring. This improves equipment maintenance efficiency. Through precise computing power calculation and comparison, the target fog node scheme is found to have high adaptability and accuracy, achieving precise management of the monitoring system and improving its reliability. Furthermore, through bidirectional health entropy and optimization algorithms, dynamic optimization and refined management of the monitoring system are achieved, reducing the latency of dynamic optimization.

[0113] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0114] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A real-time monitoring system for the entire lifecycle of generator sets based on fog computing, characterized in that, include: The monitoring node topology construction module is used to obtain the monitoring network information of the monitoring network corresponding to the target generator set, and to perform node screening based on the monitoring network information to determine the potential node topology and the monitoring node topology. The monitoring network includes multiple monitoring nodes, which are divided into potential nodes and monitoring nodes. The potential nodes refer to devices with computing capabilities that can be selected as fog nodes, and the monitoring nodes refer to devices that directly monitor the generator set components. A bidirectional entropy acquisition module is used to acquire historical monitoring logs, parse the historical monitoring logs, calculate the node health entropy of each monitoring node and each potential node, and obtain a bidirectional entropy set, wherein the bidirectional entropy set includes a subset of the health entropy of the monitored object and a subset of the health entropy of the monitored subject; using intrinsic monitoring performance data and historical monitoring time series data, the intrinsic health entropy and object health entropy of each monitoring node are obtained respectively, and the weighted intrinsic health entropy and object health entropy are used to obtain a subset of the health entropy of the monitored object; based on historical node performance time series data, typical node computing power information and typical node throughput information, the computing power fluctuation entropy and throughput fluctuation entropy of each potential node are obtained; the weighted computing power fluctuation entropy and throughput fluctuation entropy are used to obtain a subset of the health entropy of the monitored subject; The dynamic monitoring scheme generation module is used to generate a dynamic monitoring scheme based on a preset optimization algorithm, the bidirectional entropy set, the potential node topology, and the monitoring node topology, and to make dynamic monitoring decisions. Based on the bidirectional entropy set, the idle probability of multiple potential nodes and the capacity correction coefficient of multiple monitoring nodes are initialized; the potential node topology is iteratively and randomly selected by combining the idle probability and the capacity correction coefficient to obtain a set of candidate fog node schemes; the set of candidate fog node schemes is iteratively optimized based on a preset fog node evaluation function, and the optimized result is the target fog node scheme, wherein the fog node evaluation function includes a distribution uniformity factor, a capacity matching factor, a scheme average entropy factor, and a scheme total entropy factor; monitoring association decision is made according to the target fog node scheme and a preset monitoring matching rule, a monitoring mapping relationship between multiple target fog nodes and multiple monitoring nodes is established, and the monitoring mapping relationship and the target fog node scheme are associated and output as the dynamic monitoring scheme; The dynamic monitoring scheme execution module is used to apply the dynamic monitoring scheme to the monitoring network of the target generator set and perform real-time monitoring.

2. The real-time monitoring system for the entire lifecycle of a generator set based on fog computing as described in claim 1, characterized in that, Obtain historical monitoring logs, parse the logs, calculate the node health entropy for each monitored node and each potential node, and obtain a bidirectional entropy set, including: Acquire the intrinsic monitoring performance data of each monitoring node, and parse the historical monitoring logs to extract the historical monitoring time-series data of the monitoring object corresponding to each monitoring node; Based on the intrinsic monitoring performance data of the nodes, the intrinsic health entropy of each monitoring node is calculated. Based on the historical monitoring time series data, calculate the object health entropy for each monitoring node; The intrinsic health entropy and the object health entropy are weighted to obtain the monitoring object health entropy of each monitoring node, and the output is a subset of the monitoring object health entropy.

3. The real-time monitoring system for the entire lifecycle of a generator set based on fog computing as described in claim 2, characterized in that, The process includes obtaining historical monitoring logs, parsing these logs, calculating the node health entropy for each monitored node and each potential node, obtaining a bidirectional entropy set, and further including: Analyze historical monitoring logs and extract historical node performance time-series data for each potential node. Based on the historical node performance time-series data and combined with preset execution constraints, the typical computing power information and typical throughput information of each potential node are calculated and obtained. Based on the historical node performance time-series data, the typical computing power information and typical throughput information of the nodes, the computing power fluctuation entropy and throughput fluctuation entropy of each potential node are calculated traversally. The weighted computing power fluctuation entropy and throughput fluctuation entropy are used to output a subset of the health entropy of the monitored entity. The set outputs the subset of health entropy of the monitored object and the subset of health entropy of the monitored subject, which together form the bidirectional entropy set.

4. The real-time monitoring system for the entire lifecycle of a generator set based on fog computing as described in claim 1, characterized in that, Based on the bidirectional entropy set, the skip probability of multiple potential nodes and the capacity correction coefficient of multiple monitoring nodes are initialized, including: Perform Z-score normalization on the health entropy subset of the monitored objects to obtain the capacity correction coefficients of multiple monitored nodes; Perform min-max normalization on the subset of health entropy of the monitored entity to obtain the skip probability of multiple potential nodes.

5. The real-time monitoring system for the entire lifecycle of a generator set based on fog computing as described in claim 1, characterized in that, By combining the by-the-by probability and the capacity correction coefficient, the potential node topology is iteratively and randomly selected to obtain a set of candidate fog node schemes, including: Randomly select from the potential node topology to obtain the first potential node; Synchronously activate the random number generator to obtain the first random number N, where N is greater than or equal to 0 and less than or equal to 1; Compare the by-the-wheel probability corresponding to the first potential node with the first random number N. If the first random number N is larger, then add the first potential node to the first alternative fog node scheme. The first potential node is selected through iterative updates until the node capacity of all potential nodes in the first candidate fog node scheme meets the monitoring requirements of the monitoring network. The process involves iterative random sampling to obtain multiple candidate fog node schemes, which are then merged with the first candidate fog node scheme to output the candidate fog node scheme set.

6. The real-time monitoring system for the entire lifecycle of a generator set based on fog computing as described in claim 1, characterized in that, The candidate fog node scheme set is iteratively optimized based on a preset fog node evaluation function, and the optimized result is the target fog node scheme, including: By combining the preset neighborhood constraints, the potential node topology and the monitoring node topology, the number of neighborhood monitoring nodes of multiple potential nodes in each candidate fog node scheme is obtained, and the mean square error value is calculated accordingly, and the output is the distribution uniformity factor. Calculate the relative deviation between the sum of the typical throughput information of multiple potential nodes in each of the candidate fog node schemes and the monitoring demand capacity, and obtain the capacity matching factor of each of the candidate fog node schemes. Sum the monitoring subject health entropy of multiple potential nodes in each candidate fog node scheme, and take the reciprocal to obtain the scheme total entropy factor of each candidate fog node scheme; Calculate the ratio of the total entropy factor of the scheme to the sum of the typical throughput information of the nodes of the multiple potential nodes in each of the candidate fog node schemes, and obtain the average entropy factor of each of the candidate fog node schemes. Multiple fog node evaluation values ​​are obtained by weighting the distribution uniformity factor, capacity matching factor, average entropy factor, and total entropy factor of each candidate fog node scheme. The candidate fog node scheme set is iteratively updated and the corresponding fog node evaluation value is calculated. The fog node scheme with the largest corresponding fog node evaluation value is selected as the target fog node scheme.

7. The real-time monitoring system for the entire lifecycle of a generator set based on fog computing as described in claim 1, characterized in that, Also includes: Real-time monitoring of network fluctuations in the monitoring network, and fluctuation identification, including: If the fluctuation of the monitoring network is a fluctuation of the monitoring node, then based on the real-time target fog node scheme, verify whether the typical throughput information of the corresponding fog node meets the requirements of the monitoring node after the fluctuation. If satisfied, the target fog node scheme is maintained in real time; If not satisfied, then by combining the preset local adjustment constraints and the health entropy of the monitored object of the monitored node after fluctuation, local iterative optimization of the fog node is performed, and the target fog node scheme is updated based on the local iterative optimization results. If the fluctuation of the monitoring network is a fog node fluctuation, then based on the health entropy of the monitoring subject of the fog node after the fluctuation, combined with the preset fog node evaluation function, global iterative optimization of the fog node is performed.

Images