A method and system for operation and maintenance control of a marine pasture

Abstract

The application discloses a kind of offshore pasture operation and maintenance management and control method and system, it is related to the field of intelligent operation and maintenance, including: by the preset agent cluster, the multiple monitoring data of the offshore pasture to be analyzed are analyzed respectively, obtain corresponding multiple monitoring features, and according to multiple monitoring features, look up preset monitoring feature hierarchical classification dictionary, determine environmental tendency;Multiple historical operation and maintenance strategies are clustered to obtain multiple operation and maintenance tendency clustering clusters, and according to the distance between environmental tendency and the cluster center of each operation and maintenance tendency clustering cluster, determine operation and maintenance tendency;Based on multiple monitoring features, multiple operation and maintenance knowledge nodes are screened from the preset knowledge base, and operation and maintenance strategy forest is constructed according to multiple operation and maintenance knowledge nodes;Operation and maintenance strategy forest is pruned in combination with operation and maintenance tendency to obtain target operation and maintenance strategy, and then offshore pasture is operated and maintained to control.The application can improve the accuracy of offshore pasture operation and maintenance control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent operation and maintenance, and in particular to a method and system for operation and maintenance management of a marine pasture. BACKGROUND

[0002] With the development and utilization of marine resources, the marine pasture, as a new marine resource development mode that can integrate ecological, resource and environmental advantages, is developing rapidly. However, as the scale of the marine pasture gradually expands, its management problems have gradually emerged. First, the whole process of monitoring, evaluation and early warning of the marine pasture operation is ignored, and the real-time and accurate monitoring of environmental parameters and biological states is still the main way of "looking at the sky and eating food". Second, the influence of the complex marine environment is ignored, which leads to the use of various existing operation modes without careful research on the actual operation of the marine pasture. The mismatch between the operation mode and the actual operation will cause errors in the allocation of human and material resources, and will also lead to management risks. The governance and elimination of these risks will also consume unnecessary human and material resources, resulting in resource waste.

[0003] At present, the general solution to the above management problems of the marine pasture is to install a high-precision marine pasture environment resource real-time monitoring system to monitor the environmental parameters and biological states in real time. This solution can solve some problems caused by monitoring, but it still lacks analysis of the running state of the marine pasture, which leads to the use of existing operation modes as the running mode, and cannot avoid the problem of resource waste caused by the mismatch between the operation mode and the actual operation. Therefore, how to accurately analyze the marine pasture to improve the accuracy of operation and maintenance management of the marine pasture is still a problem to be solved in the prior art. SUMMARY

[0004] The present application provides a method and system for operation and maintenance management of a marine pasture to solve the technical problem of inaccurate operation and maintenance management caused by the lack of accurate analysis of the marine pasture.

[0005] According to a first aspect of the embodiments of the present application, a method for operation and maintenance management of a marine pasture is provided, comprising:

[0006] The plurality of monitoring data of the marine pasture to be analyzed are analyzed by the preset agent cluster to obtain a plurality of corresponding monitoring features, and a preset monitoring feature hierarchical classification dictionary is searched according to the plurality of monitoring features to determine the environmental tendency of the marine pasture. The plurality of monitoring data include regional weather monitoring cloud maps, equipment operation data, regional ocean current monitoring data and microbial monitoring data.

[0007] cluster a plurality of historical operation strategies of the offshore ranch to represent a plurality of historical operation tendencies of the plurality of historical operation strategies, obtain a plurality of operation tendency clustering clusters, and determine an operation tendency of the offshore ranch according to a distance between the environment tendency and a cluster center of each operation tendency clustering cluster;

[0008] based on the plurality of monitoring features, filter a plurality of operation knowledge nodes corresponding from a preset knowledge base, and construct an operation strategy forest according to the plurality of operation knowledge nodes;

[0009] in combination with the operation tendency, prune the operation strategy forest to obtain a target operation strategy, and perform operation management and control on the offshore ranch according to the target operation strategy.

[0010] The present application first analyzes a plurality of monitoring data of an offshore ranch through an agent cluster, and the monitoring data includes four types of data, i.e. regional weather, equipment operation, regional ocean current and microorganism, thereby improving the comprehensiveness and accuracy of the data, and further determining the environment tendency according to a plurality of monitoring features, which is more in line with the actual operation of the current offshore ranch. In combination with a plurality of operation tendency clustering clusters obtained by clustering a plurality of historical operation strategies, the operation tendency can be determined, which can provide reference for the operation of the current offshore ranch through historical operation strategies and environment tendency, thereby reducing the error of operation management and control. Further, a plurality of operation knowledge nodes are filtered from the knowledge base through a plurality of monitoring features, and an operation strategy forest is constructed, which can provide a plurality of operation strategy references through the knowledge base. In combination with the operation tendency, the operation strategy forest is pruned to obtain a target operation strategy, which can combine the two references to improve the accuracy of the analysis of the offshore ranch, so that the operation management and control on the offshore ranch through the target operation strategy is more accurate.

[0011] In some embodiments of the present application, the agent cluster includes a weather analysis agent, an equipment analysis agent, an ocean current analysis agent and a microorganism analysis agent; the plurality of monitoring data of the offshore ranch to be analyzed are analyzed respectively through the preset agent cluster to obtain a plurality of monitoring features, which specifically includes:

[0012] based on a visual recognition model, identify the morphological feature changes of each cloud cluster in the regional weather monitoring cloud image, and through the weather analysis agent, perform correlation analysis on the rainfall change time sequence, the wind change time sequence of each block in the regional weather monitoring cloud image and the morphological feature changes of each cloud cluster, to obtain the weather time sequence monitoring feature corresponding to the regional weather monitoring cloud image;

[0013] The device analysis intelligent agent extracts features from the device operation data based on a wavelet transform method to obtain device operation features, and detects and verifies the device operation features by combining an isolated forest algorithm and a single classification algorithm to obtain device anomaly types and device anomaly probabilities, and obtains device operation monitoring features corresponding to the device operation data by comprehensively combining the device operation features, the device anomaly types, and the device anomaly probabilities.

[0014] Based on a fluid dynamics model, the regional ocean current monitoring data is visualized to obtain an ocean current monitoring graph, and the ocean current analysis intelligent agent is used to identify the morphological features of each ocean current in the ocean current monitoring graph and match the data features corresponding to the morphological features of each ocean current in the regional ocean current monitoring data, and the ocean current partition monitoring features corresponding to the regional ocean current monitoring data are obtained by comprehensively combining the morphological features and the data features of each ocean current.

[0015] Based on the microorganism analysis intelligent agent, bioinformatics analysis is performed on the microorganism monitoring data to obtain regional species distribution and biological index features, and species community features are predicted from the microorganism monitoring data, and the microorganism monitoring features corresponding to the microorganism monitoring data are obtained by comprehensively combining the regional species distribution, the biological index features, and the species community features.

[0016] The weather analysis intelligent agent, the device analysis intelligent agent, the ocean current analysis intelligent agent, and the microorganism analysis intelligent agent are used to analyze the corresponding types of monitoring data, respectively, and through the classification analysis of different data, the accuracy of data analysis can be improved, and by increasing the types of data during data analysis, the comprehensiveness of data analysis can be improved, the accuracy of analyzing the actual operation of the marine pasture can be improved, and thus the accuracy of subsequent operation and maintenance control of the marine pasture can be improved.

[0017] In some embodiments of the present application, the construction process of the intelligent agent cluster is as follows:

[0018] The intelligent agent cluster is trained in multiple rounds, and each intelligent agent in the intelligent agent cluster uses preset construction parameters as initial training parameters at the beginning of training. Each intelligent agent uses the historical monitoring data of the corresponding type of the marine pasture as a training set for training, until each intelligent agent uses the weighted average of the training parameters of all intelligent agents in the current round as the training parameters of all intelligent agents in the next round at the end of each training in the shared number of rounds. After reaching the shared number of rounds, each intelligent agent performs iterative training by itself until the training parameters of each intelligent agent reach the corresponding threshold.

[0019] The application trains the agent cluster through multiple rounds of iteration, and each agent shares the weighted average of the training parameters of all agents before reaching the shared round, and iteratively trains after reaching the shared round. The shared parameters in the early stage can prevent the training exploration deviation of each agent, reduce the analysis deviation of each agent caused by completely independent training, and the independent training in the later stage can be targeted at different types of monitoring data corresponding to each agent, so that each agent learns the data characteristics corresponding to the type, thereby improving the analysis accuracy of each agent for the corresponding type of data.

[0020] In some embodiments of the application, the plurality of historical operation and maintenance strategies of the offshore ranch are clustered to represent a plurality of historical operation and maintenance tendencies of the plurality of historical operation and maintenance strategies, obtaining a plurality of operation and maintenance tendency clustering clusters, specifically comprising:

[0021] Based on the TF-IDF algorithm, the plurality of historical operation and maintenance strategies of the offshore ranch are respectively subjected to keyword extraction and vector feature extraction, obtaining a plurality of operation and maintenance strategy features;

[0022] The plurality of operation and maintenance strategy features are subjected to density-based clustering, obtaining a plurality of operation and maintenance tendency clustering clusters.

[0023] The application first extracts and vector features of a plurality of historical operation and maintenance strategies based on the TF-IDF algorithm, which can accurately extract keywords in each historical operation and maintenance strategy based on the word frequency method, and then performs density-based clustering on the plurality of operation and maintenance strategy features to obtain a plurality of operation and maintenance tendency clustering clusters. Through the clustering method without presetting the number of clustering clusters, the clustering error caused by artificially specifying the number of clusters is reduced, the clustering clusters obtained are more consistent with the actual multiple tendencies of the historical operation and maintenance strategies of the current offshore ranch, and the corresponding errors in subsequent analysis are avoided, improving the accuracy of subsequent analysis and operation and maintenance control based on the analysis.

[0024] In some embodiments of the application, the operation and maintenance tendency of the offshore ranch is determined according to the distance between the environmental tendency and the cluster center of each operation and maintenance tendency clustering cluster, specifically comprising:

[0025] The distance between the environmental tendency and the cluster center of each operation and maintenance tendency clustering cluster is calculated, and the dominant tendency clustering cluster in the operation and maintenance tendency clustering cluster is determined according to the minimum value of the distance, and the excluded tendency clustering cluster in the operation and maintenance tendency clustering cluster is determined according to the maximum value of the distance;

[0026] Based on the first large model, the operation and maintenance strategy features in the dominant tendency clustering cluster are summarized to obtain the corresponding dominant operation and maintenance tendency;

[0027] based on the first large model, summarize each operation and maintenance strategy feature of all operation and maintenance tendency clustering clusters except the dominant tendency clustering cluster and the exclusion tendency clustering cluster to obtain a corresponding secondary operation and maintenance tendency;

[0028] According to the dominant operation and maintenance tendency and the secondary operation and maintenance tendency, obtain the operation and maintenance tendency of the offshore pasture.

[0029] The application first determines the dominant tendency clustering cluster and the exclusion tendency clustering cluster according to the distance between the environmental tendency and the cluster center of each operation and maintenance tendency clustering cluster. The dominant tendency clustering cluster and the exclusion tendency clustering cluster can be simply and effectively determined by the physical meaning of clustering, and the operation and maintenance strategy features in the dominant tendency clustering cluster with the greatest impact are summarized to obtain the dominant operation and maintenance tendency. The operation and maintenance strategy features of all clustering clusters after excluding the dominant tendency clustering cluster and the exclusion tendency clustering cluster are summarized to obtain the secondary operation and maintenance tendency. The dominant operation and maintenance tendency with the highest correlation degree to the actual operation of the current offshore pasture and the secondary operation and maintenance tendency with the lowest correlation degree can be accurately analyzed and obtained. Therefore, the operation and maintenance tendency that is more in line with the actual operation and maintenance of the current offshore pasture can be obtained.

[0030] In some embodiments of the application, the plurality of monitoring features are filtered from a preset knowledge base to obtain a plurality of operation and maintenance knowledge nodes, and an operation and maintenance strategy forest is constructed according to the plurality of operation and maintenance knowledge nodes.

[0031] For each monitoring feature, a plurality of operation and maintenance knowledge nodes corresponding to each monitoring feature are filtered from a preset knowledge base, and a plurality of operation and maintenance knowledge paths corresponding to each monitoring feature are constructed with each monitoring feature as a root node according to the hierarchy of the operation and maintenance knowledge nodes.

[0032] According to the plurality of operation and maintenance knowledge paths corresponding to each monitoring feature, a full combination of the operation and maintenance knowledge paths corresponding to each monitoring feature is constructed to obtain an operation and maintenance strategy forest. When the operation and maintenance knowledge paths are combined, the order of the operation and maintenance knowledge nodes is rearranged according to the execution time of the operation and maintenance knowledge nodes in the operation and maintenance knowledge paths.

[0033] The application can provide multiple operation and maintenance strategy references for each monitoring feature through the knowledge base by filtering a plurality of operation and maintenance knowledge nodes corresponding to each monitoring feature from the knowledge base and constructing a plurality of operation and maintenance knowledge paths for each monitoring feature according to the hierarchy. The full combination of the operation and maintenance knowledge paths for each monitoring feature can provide as many reference strategies as possible for the current offshore pasture, improve the comprehensiveness of the reference strategies, and prevent omission of operation and maintenance strategies.

[0034] In some embodiments of the application, the operation and maintenance strategy forest is pruned in combination with the operation and maintenance tendency to obtain a target operation and maintenance strategy, specifically including:

[0035] prune the operation and maintenance policy forest to remove operation and maintenance policy trees with contradictions in nodes, and based on the operation and maintenance tendency, perform secondary pruning on the operation and maintenance policy forest based on a preset second large model to screen a target operation and maintenance policy.

[0036] The operation and maintenance policy forest is pruned multiple times, which can avoid selecting a wrong operation and maintenance policy through one-time pruning, and then accurately determine a target operation and maintenance policy that is consistent with the operation and maintenance tendency of the current offshore pasture through secondary pruning, thereby improving the analysis accuracy of the offshore pasture and improving the subsequent operation and maintenance control of the current offshore pasture through the target operation and maintenance policy.

[0037] According to a second aspect of the embodiments of the present application, an offshore pasture operation and maintenance control system is provided, which includes a data tendency analysis module, a strategy clustering analysis module, a strategy forest construction module, and a strategy forest pruning module.

[0038] The data tendency analysis module is configured to analyze a plurality of monitoring data of the offshore pasture to be analyzed through a preset agent cluster to obtain corresponding monitoring features, and search a preset monitoring feature hierarchical classification dictionary according to the monitoring features to determine the environmental tendency of the offshore pasture; wherein the plurality of monitoring data includes regional weather monitoring cloud images, equipment operation data, regional ocean current monitoring data, and microbial monitoring data.

[0039] The strategy clustering analysis module is configured to cluster a plurality of historical operation and maintenance strategies of the offshore pasture to represent a plurality of historical operation and maintenance tendencies of the plurality of historical operation and maintenance strategies, obtain a plurality of operation and maintenance tendency clustering clusters, and determine the operation and maintenance tendency of the offshore pasture according to the distance between the environmental tendency and the cluster center of each operation and maintenance tendency clustering cluster.

[0040] The strategy forest construction module is configured to screen a plurality of operation and maintenance knowledge nodes corresponding to the plurality of monitoring features from a preset knowledge base, and construct an operation and maintenance policy forest according to the plurality of operation and maintenance knowledge nodes.

[0041] The strategy forest pruning module is configured to prune the operation and maintenance policy forest in combination with the operation and maintenance tendency to obtain a target operation and maintenance policy, and perform operation and maintenance control on the offshore pasture according to the target operation and maintenance policy.

[0042] In some embodiments of the present application, the agent cluster includes a weather analysis agent, an equipment analysis agent, an ocean current analysis agent, and a microbial analysis agent; the data tendency analysis module includes a feature analysis submodule; and the feature analysis submodule includes a weather analysis unit, an equipment analysis unit, an ocean current analysis unit, and a microbial analysis unit.

[0043] The weather analysis unit is configured to identify morphological feature changes of each cloud cluster in the regional weather monitoring cloud image based on a visual recognition model, and to perform correlation analysis on the rainfall variation time sequence, the wind variation time sequence of each block in the regional weather monitoring cloud image, and the morphological feature changes of each cloud cluster by the weather analysis agent to obtain weather time sequence monitoring features corresponding to the regional weather monitoring cloud image.

[0044] The device analysis unit is configured to perform feature extraction on the device operation data based on a wavelet transform method by the device analysis agent to obtain device operation features, and to detect and verify the device operation features by combining an isolation forest algorithm and a single classification algorithm to obtain device abnormality types and device abnormality probabilities, and to obtain device operation monitoring features corresponding to the device operation data by comprehensively combining the device operation features, the device abnormality types, and the device abnormality probabilities.

[0045] The ocean current analysis unit is configured to visualize the regional ocean current monitoring data based on a fluid dynamics model to obtain an ocean current monitoring image, to identify morphological features of each ocean current in the ocean current monitoring image by the ocean current analysis agent, and to match data features corresponding to the morphological features of each ocean current in the regional ocean current monitoring data, and to obtain ocean current partition monitoring features corresponding to the regional ocean current monitoring data by comprehensively combining the morphological features and the data features of each ocean current.

[0046] The microorganism analysis unit is configured to perform bioinformatics analysis on the microorganism monitoring data based on the microorganism analysis agent to obtain regional species distribution and biological index features, and to predict species community features from the microorganism monitoring data, and to obtain microorganism monitoring features corresponding to the microorganism monitoring data by comprehensively combining the regional species distribution, the biological index features, and the species community features.

[0047] In some embodiments of the present application, the construction process of the agent cluster is specifically as follows:

[0048] The agent cluster is subjected to multi-round iterative training. At the beginning of the training, each agent in the agent cluster uses preset construction parameters as initial training parameters, and each agent uses historical monitoring data of the corresponding type of the marine pasture as a training set for training, until each agent uses a weighted average of the training parameters of all agents in the current round as the training parameters of all agents in the next round at the end of each training in the shared number of rounds, and each agent performs iterative training by itself after the shared number of rounds is reached, until the training parameters of each agent reach the corresponding threshold.

[0049] In some embodiments of this application, the strategy clustering analysis module includes a strategy clustering submodule; the strategy clustering submodule includes a feature extraction unit and a feature clustering unit;

[0050] The feature extraction unit is used to extract keywords and vector features from multiple historical operation and maintenance strategies of the marine ranch based on the TF-IDF algorithm, so as to obtain the corresponding multiple operation and maintenance strategy features.

[0051] The feature clustering unit is used to perform density-based clustering on the multiple operation and maintenance strategy features to obtain multiple operation and maintenance tendency clusters.

[0052] In some embodiments of this application, the strategy clustering analysis module includes a tendency analysis submodule; the tendency analysis submodule includes a cluster determination unit, a dominant tendency determination unit, a secondary tendency determination unit, and an operation and maintenance tendency determination unit;

[0053] The cluster determination unit is used to calculate the distance between the environmental tendency and the cluster center of each operation and maintenance tendency cluster, and to determine the dominant tendency cluster in the operation and maintenance tendency cluster based on the minimum distance, and to determine the exclusion tendency cluster in the operation and maintenance tendency cluster based on the maximum distance.

[0054] The dominant tendency determination unit is used to summarize the characteristics of each operation and maintenance strategy in the dominant tendency cluster based on a preset first major model, and obtain the corresponding dominant operation and maintenance tendency.

[0055] The secondary tendency determination unit is used to summarize the operation and maintenance strategy characteristics of all operation and maintenance tendency clusters other than the dominant tendency cluster and the excluded tendency cluster based on the first large model, so as to obtain the corresponding secondary operation and maintenance tendency.

[0056] The operation and maintenance tendency determination unit is used to obtain the operation and maintenance tendency of the marine ranch based on the dominant operation and maintenance tendency and the secondary operation and maintenance tendency.

[0057] In some embodiments of this application, the strategy forest construction module includes a knowledge node filtering submodule and a strategy forest construction submodule;

[0058] The knowledge node filtering submodule is used to filter out multiple operation and maintenance knowledge nodes corresponding to each monitoring feature from a preset knowledge base, and construct multiple operation and maintenance knowledge paths corresponding to each monitoring feature with each monitoring feature as the root node according to the hierarchy of the operation and maintenance knowledge nodes.

[0059] The policy forest construction submodule is configured to construct a full combination of the operation and maintenance knowledge path corresponding to each monitoring feature according to a plurality of operation and maintenance knowledge paths corresponding to each monitoring feature, to obtain an operation and maintenance policy forest; when the operation and maintenance knowledge paths are combined, the order of the operation and maintenance knowledge nodes is rearranged according to the execution time of the operation and maintenance knowledge nodes in the operation and maintenance knowledge paths.

[0060] In some embodiments of the present application, the policy forest pruning module includes a policy pruning submodule;

[0061] The policy pruning submodule is configured to prune the operation and maintenance policy forest once, and eliminate operation and maintenance policy trees with contradictory nodes in the operation and maintenance policy forest; and based on a preset second large model, the operation and maintenance policy forest is pruned twice according to the operation and maintenance tendency, to obtain a target operation and maintenance policy.

[0062] The present application first analyzes a plurality of monitoring data of the offshore ranch through the agent cluster, and the monitoring data includes four types of data, i.e., regional weather, equipment operation, regional ocean current and microorganism, thereby improving the comprehensiveness and accuracy of the data, and further determining the environment tendency according to a plurality of monitoring features, which is more in line with the actual operation of the current offshore ranch. In addition, the operation and maintenance tendency is determined in combination with a plurality of operation and maintenance tendency clustering clusters obtained by clustering a plurality of historical operation and maintenance strategies, which can provide a reference for the operation and maintenance of the current offshore ranch through the historical operation and maintenance strategies and the environment tendency, and reduce the error of the operation and maintenance control. Further, a plurality of operation and maintenance knowledge nodes are screened from the knowledge base through a plurality of monitoring features, and an operation and maintenance policy forest is constructed, which can provide a plurality of operation and maintenance strategy references through the knowledge base, and the operation and maintenance policy forest is pruned in combination with the operation and maintenance tendency to obtain a target operation and maintenance policy, which can combine the two references to improve the accuracy of the analysis of the offshore ranch, so that the operation and maintenance control of the offshore ranch through the target operation and maintenance policy is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 : a flowchart of an offshore ranch operation and maintenance control method shown in some embodiments of the present application;

[0064] Figure 2 : a module structure diagram of an offshore ranch operation and maintenance control system shown in some embodiments of the present application. DETAILED DESCRIPTION

[0065] Embodiments of the present application are described below in detail with reference to the accompanying drawings, wherein the same or similar numerals represent the same or similar elements or elements having the same or similar functions throughout the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain some embodiments of the present application, and cannot be understood as limiting the embodiments of the present application. Based on the embodiments shown in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0066] In the description of the present application, it should be understood that the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can be explicitly or implicitly included at least one of the features. In the description of the present application, unless otherwise explicitly and specifically limited, the meaning of "a plurality of" "several" is two or more.

[0067] The general solution for ignoring the whole process monitoring and ignoring the influence of the complex marine environment of the current offshore ranch is to install a high-precision marine ranch environment resource real-time monitoring system to monitor the environmental parameters and biological state in real time. This scheme still lacks analysis of the running state of the offshore ranch, resulting in the use of existing operation modes, and unable to avoid the problem of resource waste caused by mismatched operation modes. Therefore, how to accurately analyze the offshore ranch and improve the accuracy of the operation and maintenance control of the offshore ranch is still a problem to be solved in the prior art.

[0068] Based on the above technical background, please refer to Figure 1 The embodiments of the present application provide an offshore ranch operation and maintenance control method, comprising steps S101 to S104, and each step is specifically as follows:

[0069] Step S101: analyzing a plurality of monitoring data of the offshore ranch to be analyzed by a preset agent cluster to obtain a plurality of corresponding monitoring features, and searching a preset monitoring feature hierarchical classification dictionary according to the plurality of monitoring features to determine the environmental tendency of the offshore ranch; wherein the plurality of monitoring data comprises regional weather monitoring cloud images, equipment operation data, regional ocean current monitoring data and microbial monitoring data.

[0070] In some embodiments of the present application, the agent cluster comprises a weather analysis agent, an equipment analysis agent, an ocean current analysis agent and a microbial analysis agent; the plurality of monitoring data of the offshore ranch to be analyzed are analyzed by the preset agent cluster to obtain a plurality of corresponding monitoring features, which specifically comprises:

[0071] Based on the visual recognition model, the morphological feature changes of each cloud cluster in the regional weather monitoring cloud image are identified, and through the weather analysis agent, the rainfall change time sequence, the wind change time sequence of each block in the regional weather monitoring cloud image, and the morphological feature changes of each cloud cluster are associated and analyzed to obtain the weather time sequence monitoring features corresponding to the regional weather monitoring cloud image.

[0072] Through the equipment analysis agent, the equipment operation features are obtained by performing feature extraction on the equipment operation data based on a wavelet transform method, and the equipment anomaly type and the equipment anomaly probability are obtained by detecting and verifying the equipment operation features based on an isolation forest algorithm and a single classification algorithm. The equipment operation monitoring features corresponding to the equipment operation data are obtained by comprehensively considering the equipment operation features, the equipment anomaly type, and the equipment anomaly probability.

[0073] Based on the fluid dynamics model, the ocean current monitoring data of the region is visualized to obtain an ocean current monitoring image, and through the ocean current analysis agent, the morphological features of each ocean current in the ocean current monitoring image are identified, and the data features corresponding to the morphological features of each ocean current are matched in the regional ocean current monitoring data. The ocean current partition monitoring features corresponding to the regional ocean current monitoring data are obtained by comprehensively considering the morphological features and the data features of each ocean current.

[0074] Based on the microorganism analysis agent, the regional species distribution and the biological index features are obtained by performing bioinformatics analysis on the microorganism monitoring data, and the species community features are predicted according to the microorganism monitoring data. The microorganism monitoring features corresponding to the microorganism monitoring data are obtained by comprehensively considering the regional species distribution, the biological index features, and the species community features.

[0075] In some embodiments of the present application, when identifying the morphological feature changes of each cloud cluster in the regional weather monitoring cloud image, the visual recognition model includes but is not limited to a YOLO series model, an SSD model, or an R-CNN model and an improved Fast R-CNN model thereof.

[0076] In the present application, the weather analysis agent, the equipment analysis agent, the ocean current analysis agent, and the microorganism analysis agent are respectively used to analyze the corresponding types of monitoring data. Through the classification analysis of different data, the accuracy of data analysis can be improved, and through the increase of the types of data in data analysis, the comprehensiveness of data analysis can be improved, and the accuracy of the analysis of the actual operation of the marine pasture can be improved, thereby improving the accuracy of subsequent operation and maintenance control of the marine pasture.

[0077] In some embodiments of the present application, the construction process of the agent cluster is as follows:

[0078] The agent cluster is trained in multiple rounds of iterations, each agent in the agent cluster uses preset construction parameters as initial training parameters at the beginning of training, each agent uses historical monitoring data of a corresponding type of the offshore pasture as a training set for training, until each agent uses a weighted average of training parameters of all agents in the current round as training parameters of all agents in the next round at the end of each round of training before the shared round, each agent performs iterative training individually after the shared round, until the training parameters of each agent reach a corresponding threshold.

[0079] In some embodiments of the present application, one way of obtaining a weighted average of training parameters of all agents before the shared round can be arithmetic average, or the corresponding weight can be determined according to the importance of the data type corresponding to the agent in the offshore pasture, which can be obtained by examining the environmental conditions corresponding to each type of data before the construction of the offshore pasture.

[0080] The present application trains the agent cluster in multiple rounds of iterations, and each agent shares a weighted average of training parameters of all agents in each round before the shared round, and each agent performs iterative training individually after the shared round, which can prevent the training exploration of each agent from deviating, reduce the analysis deviation of each agent caused by completely independent training, and enable each agent to learn the data characteristics of the corresponding type through independent training in the later stage, thereby improving the analysis accuracy of each agent for the corresponding type of data.

[0081] Step S102: clustering a plurality of historical operation and maintenance strategies of the offshore pasture to represent a plurality of historical operation and maintenance tendencies of the plurality of historical operation and maintenance strategies, obtaining a plurality of operation and maintenance tendency clustering clusters, and determining an operation and maintenance tendency of the offshore pasture according to a distance between the environmental tendency and a cluster center of each operation and maintenance tendency clustering cluster.

[0082] In some embodiments of the present application, the clustering of the plurality of historical operation and maintenance strategies of the offshore pasture to represent a plurality of historical operation and maintenance tendencies of the plurality of historical operation and maintenance strategies, obtaining a plurality of operation and maintenance tendency clustering clusters, specifically includes:

[0083] Based on the TF-IDF algorithm, keyword extraction and vector feature extraction are performed on the plurality of historical operation and maintenance strategies of the offshore pasture respectively, obtaining a plurality of corresponding operation and maintenance strategy features;

[0084] The plurality of operation and maintenance strategy features are subjected to density-based clustering to obtain a plurality of operation and maintenance tendency clustering clusters.

[0085] Specifically, the TF-IDF algorithm (Term Frequency-Inverse Document Frequency) is used in some embodiments of the present application to extract keywords and vectorize features of the multiple historical operation and maintenance strategies of the marine pasture, respectively, to obtain corresponding multiple operation and maintenance strategy features, specifically:

[0086] The segmentation feature of each historical operation and maintenance strategy is extracted, and the word frequency of each word in the same historical operation and maintenance strategy is counted during the segmentation process to obtain the corresponding word group and the word frequency of each word in the word group.

[0087] The total word group is obtained by merging and removing duplicates of the word group corresponding to each historical operation and maintenance strategy, and the number of historical operation and maintenance strategies containing the current word is counted as the inverse file number of the current word.

[0088] The logarithm of the quotient of the number of historical operation and maintenance strategies and the inverse file number of each word in the total word group is taken as the inverse word frequency of each word in the total word group.

[0089] According to the product of the word frequency and inverse word frequency of each word in the word group corresponding to each historical operation and maintenance strategy, the word group sorting sequence corresponding to each historical operation and maintenance strategy is obtained, the word group sorting sequence is sorted from large to small according to the product of the word frequency and inverse word frequency of each word, and the words in the last N positions of the word group sorting sequence of each historical operation and maintenance strategy are removed to obtain the operation and maintenance strategy feature corresponding to each historical operation and maintenance strategy.

[0090] The density-based clustering of multiple operation and maintenance strategy features is considered because the common clustering method knows the type of classification, i.e., the number of clusters to be classified, so the number of clustering clusters can be directly limited to reduce the corresponding operation process. However, the clustering of operation and maintenance strategy features is unknown classification type, and limiting the number of clustering clusters will classify one or more classifications that should not be classified into the same cluster into the same cluster, resulting in uneven and inaccurate actual classification. Therefore, by using the density-based clustering method without presetting the number of clustering clusters, the classification error caused by artificially limiting the number of clusters can be directly avoided, thereby obtaining more accurate clustering clusters, and subsequent use based on the clustering clusters is more consistent with the actual situation.

[0091] The application first extracts keywords from multiple historical operation and maintenance strategies and vectorizes the keywords to obtain multiple operation and maintenance strategy features based on the TF-IDF algorithm. The keywords in each historical operation and maintenance strategy can be accurately extracted based on the word frequency. The multiple operation and maintenance strategy features are clustered based on density to obtain multiple operation and maintenance tendency clustering clusters. The clustering method without presetting the number of clustering clusters can reduce the clustering error caused by manually specifying the number of clusters, so that the obtained clustering clusters are more consistent with the actual multiple tendencies of the historical operation and maintenance strategies of the current offshore pasture. This can avoid errors in subsequent analysis and improve the accuracy of subsequent analysis and operation and maintenance control based on the analysis.

[0092] In some embodiments of the application, the operation and maintenance tendency of the offshore pasture is determined according to the distance between the environment tendency and the cluster center of each operation and maintenance tendency clustering cluster, specifically comprising:

[0093] The distance between the environment tendency and the cluster center of each operation and maintenance tendency clustering cluster is calculated, and the dominant tendency clustering cluster in the operation and maintenance tendency clustering cluster is determined according to the minimum value of the distance, and the excluded tendency clustering cluster in the operation and maintenance tendency clustering cluster is determined according to the maximum value of the distance;

[0094] Based on the first large model, the operation and maintenance strategy features in the dominant tendency clustering cluster are summarized to obtain the corresponding dominant operation and maintenance tendency;

[0095] Based on the first large model, the operation and maintenance strategy features of all operation and maintenance tendency clustering clusters except the dominant tendency clustering cluster and the excluded tendency clustering cluster are summarized to obtain the corresponding secondary operation and maintenance tendency;

[0096] The operation and maintenance tendency of the offshore pasture is obtained according to the dominant operation and maintenance tendency and the secondary operation and maintenance tendency.

[0097] In some embodiments of the application, the first large model can be WenXin YiYan, TongYi QianWen, XunFei XingHuo, Deepseek, Gemini, GPT-4 or Kimi, or a large model obtained by adjusting a domain large model, especially a text-based large model, by Sora.

[0098] Specifically, when summarizing according to the first large model, the corresponding operation and maintenance strategy features can be directly used as input, and a preset prompt word for summarization can be used to ask questions to obtain the corresponding dominant / secondary operation and maintenance tendency.

[0099] The application first determines a dominant tendency cluster and an excluded tendency cluster according to the distance between the cluster center of the environment tendency and each operation and maintenance tendency, can simply and effectively determine the cluster with the greatest and smallest influence on the operation and maintenance tendency through the physical meaning of clustering, and then summarizes each operation and maintenance strategy feature in the dominant tendency cluster with the greatest influence to obtain a dominant operation and maintenance tendency, and summarizes each operation and maintenance strategy feature of all clusters after excluding the dominant tendency cluster and the excluded tendency cluster to obtain a secondary operation and maintenance tendency, can accurately analyze the dominant operation and maintenance tendency with the highest correlation degree to the actual operation of the current offshore pasture and the secondary operation and maintenance tendency with the lowest correlation degree, and thus obtain an operation and maintenance tendency more in line with the actual operation and maintenance of the current offshore pasture.

[0100] Step S103: based on the plurality of monitoring features, filtering a plurality of operation and maintenance knowledge nodes corresponding from a preset knowledge base, and constructing an operation and maintenance strategy forest according to the plurality of operation and maintenance knowledge nodes.

[0101] In some embodiments of the application, the operation and maintenance strategy forest is constructed based on the plurality of monitoring features, filtering a plurality of operation and maintenance knowledge nodes corresponding from a preset knowledge base, and according to the plurality of operation and maintenance knowledge nodes, specifically comprising:

[0102] For each monitoring feature, a plurality of operation and maintenance knowledge nodes corresponding to each monitoring feature are filtered from a preset knowledge base, and according to the hierarchy of the operation and maintenance knowledge nodes, each monitoring feature is taken as a root node to construct a plurality of operation and maintenance knowledge paths corresponding to each monitoring feature;

[0103] According to the plurality of operation and maintenance knowledge paths corresponding to each monitoring feature, the full combination of the operation and maintenance knowledge paths corresponding to each monitoring feature is constructed to obtain an operation and maintenance strategy forest; wherein when the operation and maintenance knowledge paths are combined, the order of the operation and maintenance knowledge nodes is rearranged according to the execution time of the operation and maintenance knowledge nodes in the operation and maintenance knowledge paths.

[0104] Specifically, the storage structure of the knowledge base is a graph structure, which is a directed acyclic graph. Specifically, when a certain monitoring feature is filtered, there is a possibility of filtering a plurality of starting nodes from the knowledge base, and when each intermediate node is connected to the end node through the direction of the starting node, there is a possibility that the successor of one starting node or one intermediate node has multiple intermediate nodes, so there are multiple operation and maintenance knowledge paths corresponding to each monitoring feature. Generally, when the current monitoring feature is filtered, the next node of the root node is the next node of the starting node, and the end node is a node without out-degree in the knowledge base.

[0105] For example, one example of constructing the full combination of the operation and maintenance knowledge paths corresponding to each monitoring feature can be: the operation and maintenance knowledge path corresponding to the monitoring feature A is , and the operation and maintenance knowledge path corresponding to the monitoring feature B is monitoring feature C corresponds to an operation and maintenance knowledge path The corresponding full combination includes the following 12 cases: That is, 12 operation and maintenance strategy trees in the constructed operation and maintenance strategy forest. Exemplarily, when the combination obtains an operation and maintenance strategy tree , the corresponding operation and maintenance knowledge path is combined end to end, and the order of the operation and maintenance knowledge nodes in the operation and maintenance strategy tree is rearranged according to the execution time of each operation and maintenance knowledge node in the operation and maintenance knowledge path.

[0106] The present application can provide multiple operation and maintenance strategy references for each monitoring feature through the knowledge base by screening multiple operation and maintenance knowledge nodes corresponding to each monitoring feature from the knowledge base, combining the hierarchy to construct multiple operation and maintenance knowledge paths for each monitoring feature, and constructing the full combination of the operation and maintenance knowledge paths for each monitoring feature, so as to provide as many reference strategies as possible for the current offshore pasture, improve the comprehensiveness of the reference strategies, and prevent omission of operation and maintenance strategies.

[0107] Step S104: combining the operation and maintenance tendency, pruning the operation and maintenance strategy forest to obtain a target operation and maintenance strategy, and performing operation and maintenance management and control on the offshore pasture according to the target operation and maintenance strategy.

[0108] In some embodiments of the present application, the combination of the operation and maintenance tendency and the pruning of the operation and maintenance strategy forest to obtain a target operation and maintenance strategy specifically includes:

[0109] Pruning the operation and maintenance strategy forest once, eliminating operation and maintenance strategy trees with contradictory nodes in the operation and maintenance strategy forest, and performing secondary pruning on the operation and maintenance strategy forest based on a preset second large model according to the operation and maintenance tendency to obtain a target operation and maintenance strategy.

[0110] Exemplarily, in the process of pruning once, if there is contradiction between certain operation and maintenance knowledge nodes in the operation and maintenance strategy tree , , it is considered that the operation and maintenance strategy tree corresponds to an unachievable operation and maintenance strategy, and the operation and maintenance strategy tree should be pruned and eliminated from the operation and maintenance strategy forest.

[0111] In some embodiments of the present application, the second large model can be WenXin YiYan, TongYi QianWen, XunFei XingHuo, Deepseek, Gemini, GPT-4 or Kimi, or a large model obtained by adjusting a domain large model, especially a text-based large model, through Sora.

[0112] Specifically, when the operation and maintenance strategy forest is pruned again according to the operation and maintenance tendency and the second largest model, the prompt word can be directly constructed according to the operation and maintenance tendency, and the corresponding question can be asked to the second largest model to directly screen the operation and maintenance strategy tree in the operation and maintenance strategy forest.

[0113] Through multiple pruning of the operation and maintenance strategy forest, the application can avoid selecting the wrong generated contradictory operation and maintenance strategy through one-time pruning, and then accurately determine the target operation and maintenance strategy that conforms to the operation and maintenance tendency of the current offshore pasture through secondary pruning, thereby improving the analysis accuracy of the offshore pasture, and then improving the more accurate operation and maintenance control of the current offshore pasture through the target operation and maintenance strategy.

[0114] Compared with the prior art, the application first analyzes the various monitoring data of the offshore pasture through the agent cluster, and the monitoring data includes four types of data, i.e., regional weather, equipment operation, regional ocean current and microorganism, thereby improving the comprehensiveness and accuracy of the data, and then determining the environmental tendency according to the various monitoring features is more in line with the actual operation of the current offshore pasture, and then determining the operation and maintenance tendency in combination with the multiple operation and maintenance tendency clustering clusters obtained by clustering the multiple historical operation and maintenance strategies can provide a reference for the operation and maintenance of the current offshore pasture through the historical operation and maintenance strategies and the environmental tendency, thereby reducing the error of operation and maintenance control; and then screening multiple operation and maintenance knowledge nodes from the knowledge base through various monitoring features and constructing an operation and maintenance strategy forest, which can provide various operation and maintenance strategy references through the knowledge base, and the target operation and maintenance strategy can be obtained by pruning the operation and maintenance strategy forest in combination with the operation and maintenance tendency, thereby combining the two references, improving the accuracy of the analysis of the offshore pasture, and thus being more accurate when the offshore pasture is operated and controlled through the target operation and maintenance strategy.

[0115] Corresponding to the foregoing method, please refer to Figure 2 The embodiment of the application provides an offshore pasture operation and maintenance control system, which comprises a data tendency analysis module 210, a strategy clustering analysis module 220, a strategy forest construction module 230 and a strategy forest pruning module 240.

[0116] The data tendency analysis module 210 is used for analyzing various monitoring data of an offshore pasture to be analyzed through a preset agent cluster, obtaining corresponding various monitoring features, and searching a preset monitoring feature hierarchical classification dictionary according to the various monitoring features to determine the environmental tendency of the offshore pasture; wherein the various monitoring data comprises regional weather monitoring cloud images, equipment operation data, regional ocean current monitoring data and microorganism monitoring data.

[0117] The strategy clustering analysis module 220 is configured to cluster a plurality of historical operation and maintenance strategies of the offshore pasture to represent a plurality of historical operation and maintenance tendencies of the plurality of historical operation and maintenance strategies, obtain a plurality of operation and maintenance tendency clustering clusters, and determine an operation and maintenance tendency of the offshore pasture according to a distance between the environmental tendency and a cluster center of each operation and maintenance tendency clustering cluster.

[0118] The strategy forest construction module 230 is configured to filter a plurality of operation and maintenance knowledge nodes corresponding to the plurality of monitoring features from a preset knowledge base, and construct an operation and maintenance strategy forest according to the plurality of operation and maintenance knowledge nodes.

[0119] The strategy forest pruning module 240 is configured to prune the operation and maintenance strategy forest in combination with the operation and maintenance tendency to obtain a target operation and maintenance strategy, and perform operation and maintenance management and control on the offshore pasture according to the target operation and maintenance strategy.

[0120] In some embodiments of the present application, the agent cluster includes a weather analysis agent, a device analysis agent, a current analysis agent, and a microorganism analysis agent; the data tendency analysis module 210 includes a feature analysis sub-module; and the feature analysis sub-module includes a weather analysis unit, a device analysis unit, a current analysis unit, and a microorganism analysis unit.

[0121] The weather analysis unit is configured to identify morphological feature changes of each cloud cluster in the regional weather monitoring cloud image based on a visual recognition model, and perform correlation analysis on the rainfall change time sequence, the wind change time sequence of each block in the regional weather monitoring cloud image, and the morphological feature changes of each cloud cluster in combination with the weather analysis agent to obtain weather time sequence monitoring features corresponding to the regional weather monitoring cloud image.

[0122] The device analysis unit is configured to perform feature extraction on the device operation data based on a wavelet transform method through the device analysis agent to obtain device operation features, and perform detection and verification on the device operation features in combination with an isolation forest algorithm and a single classification algorithm to obtain a device abnormality type and a device abnormality probability, and obtain device operation monitoring features corresponding to the device operation data in combination with the device operation features, the device abnormality type, and the device abnormality probability.

[0123] The current analysis unit is configured to visualize the regional current monitoring data based on a fluid dynamics model to obtain a current monitoring graph, identify morphological features of each current in the current monitoring graph through the current analysis agent, match data features corresponding to the morphological features of each current in the regional current monitoring data, and obtain current partition monitoring features corresponding to the regional current monitoring data in combination with the morphological features and the data features of each current.

[0124] The microorganism analysis unit is configured to perform bioinformatics analysis on the microorganism monitoring data based on the microorganism analysis agent to obtain regional species distribution and biological index characteristics, and to predict species community characteristics based on the microorganism monitoring data, and to obtain microorganism monitoring characteristics corresponding to the microorganism monitoring data by comprehensively combining the regional species distribution, the biological index characteristics, and the species community characteristics.

[0125] In some embodiments of the present application, the construction process of the agent cluster comprises:

[0126] The agent cluster is subjected to multi-round iterative training. At the beginning of the training, each agent in the agent cluster uses preset construction parameters as initial training parameters. Each agent uses historical monitoring data of a corresponding type of the offshore pasture as a training set for training. Until the shared number of rounds is reached, each agent uses a weighted average of training parameters of all agents in the current round as training parameters of all agents in the next round at the end of each training in the shared round. After the shared number of rounds is reached, each agent is individually subjected to iterative training until the training parameters of each agent reach a corresponding threshold.

[0127] In some embodiments of the present application, the strategy clustering analysis module 220 comprises a strategy clustering sub-module; the strategy clustering sub-module comprises a feature extraction unit and a feature clustering unit;

[0128] The feature extraction unit is configured to perform keyword extraction and vector feature extraction on a plurality of historical operation and maintenance strategies of the offshore pasture based on a TF-IDF algorithm to obtain a plurality of operation and maintenance strategy features corresponding thereto.

[0129] The feature clustering unit is configured to perform density-based clustering on the plurality of operation and maintenance strategy features to obtain a plurality of operation and maintenance tendency clustering clusters.

[0130] In some embodiments of the present application, the strategy clustering analysis module 220 comprises a tendency analysis sub-module; the tendency analysis sub-module comprises a clustering cluster determination unit, a dominant tendency determination unit, a secondary tendency determination unit, and an operation and maintenance tendency determination unit.

[0131] The clustering cluster determination unit is configured to calculate distances between the environmental tendency and cluster centers of each operation and maintenance tendency clustering cluster, and to determine a dominant tendency clustering cluster in the operation and maintenance tendency clustering cluster according to a minimum value of the distances, and to determine an excluded tendency clustering cluster in the operation and maintenance tendency clustering cluster according to a maximum value of the distances.

[0132] The dominant tendency determination unit is configured to summarize each operation and maintenance strategy feature in the dominant tendency clustering cluster based on a preset first large model to obtain a corresponding dominant operation and maintenance tendency.

[0133] The secondary tendency determination unit is configured to summarize each operation and maintenance strategy feature of all operation and maintenance tendency clustering clusters except the dominant tendency clustering cluster and the exclusion tendency clustering cluster based on the first large model to obtain a corresponding secondary operation and maintenance tendency.

[0134] The operation and maintenance tendency determination unit is configured to obtain an operation and maintenance tendency of the offshore ranch according to the dominant operation and maintenance tendency and the secondary operation and maintenance tendency.

[0135] In some embodiments of the present application, the policy forest construction module 230 comprises a knowledge node screening submodule and a policy forest construction submodule.

[0136] The knowledge node screening submodule is configured to, for each monitoring feature, screen a plurality of operation and maintenance knowledge nodes corresponding to each monitoring feature from a preset knowledge base, and construct, as a root node of each monitoring feature, a plurality of operation and maintenance knowledge paths corresponding to each monitoring feature according to a hierarchy of the operation and maintenance knowledge nodes.

[0137] The policy forest construction submodule is configured to construct a full combination of the operation and maintenance knowledge paths corresponding to each monitoring feature according to the plurality of operation and maintenance knowledge paths corresponding to each monitoring feature to obtain an operation and maintenance strategy forest; and when the operation and maintenance knowledge paths are combined, the order of the operation and maintenance knowledge nodes is rearranged according to the execution time of the operation and maintenance knowledge nodes in the operation and maintenance knowledge paths.

[0138] In some embodiments of the present application, the policy forest pruning module 240 comprises a policy pruning submodule.

[0139] The policy pruning submodule is configured to perform a first pruning on the operation and maintenance strategy forest, to eliminate, in the pruning, operation and maintenance strategy trees with contradictory nodes in the operation and maintenance strategy forest, and to perform a second pruning on the operation and maintenance strategy forest based on a preset second large model according to the operation and maintenance tendency, to screen a target operation and maintenance strategy.

[0140] The application first analyzes various monitoring data of the sea ranch through the agent cluster, and the monitoring data includes four types of data of regional weather, equipment operation, regional ocean current and microorganism, improves the comprehensiveness and accuracy of the data, and then determines the environment tendency according to various monitoring characteristics, which is more in line with the actual operation of the current sea ranch, and determines the operation and maintenance tendency by combining the operation and maintenance tendency clustering obtained by clustering of multiple historical operation and maintenance strategies, which can provide reference for the operation and maintenance of the current sea ranch through the historical operation and maintenance strategy and the environment tendency, and reduce the error of operation and maintenance control; and then a plurality of operation and maintenance knowledge nodes are screened from the knowledge base through various monitoring characteristics, and an operation and maintenance strategy forest is constructed, which can provide various operation and maintenance strategy references through the knowledge base, and the target operation and maintenance strategy is obtained by pruning the operation and maintenance strategy forest combined with the operation and maintenance tendency, which can combine the two references to improve the accuracy of the analysis of the sea ranch, so that the operation and maintenance of the sea ranch through the target operation and maintenance strategy is more accurate.

[0141] It should be understood that the system provided by the embodiments of the application corresponds to the foregoing method, and the sea ranch operation and maintenance control system provided by the embodiments of the application can implement the sea ranch operation and maintenance control method provided by any one of the embodiments of the application.

[0142] Adaptively, the embodiments of the application further provide a computer device and a computer readable storage medium.

[0143] The computer device comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor;

[0144] When the processor executes the computer program, the sea ranch operation and maintenance control method of the application is implemented.

[0145] The computer readable storage medium stores a plurality of instructions, which are suitable for being loaded by the processor to execute the sea ranch operation and maintenance control method of the application.

[0146] The above is part of the embodiments of the application, which further details the purpose, technical solutions and beneficial effects of the application. It should be clear that the above part of the embodiments of the application cannot be understood as a limitation of the application. In particular, for those skilled in the art, any changes, modifications, equivalent replacements and variations, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. A method for the operation and maintenance management of marine ranches, characterized in that, include: Through a pre-defined cluster of intelligent agents, various monitoring data of the marine ranch to be analyzed are analyzed to obtain corresponding monitoring features. Based on these features, a pre-defined hierarchical classification dictionary is searched to determine the environmental characteristics of the marine ranch. The various monitoring data include regional weather monitoring cloud maps, equipment operation data, regional ocean current monitoring data, and microbial monitoring data. Clustering is performed on multiple historical operation and maintenance strategies of the marine ranch to characterize several historical operation and maintenance tendencies of these strategies, resulting in multiple operation and maintenance tendency clusters. Specifically, this includes: using the TF-IDF algorithm, extracting keywords and performing vector feature generation on the multiple historical operation and maintenance strategies of the marine ranch to obtain corresponding operation and maintenance strategy features; performing density-based clustering on the multiple operation and maintenance strategy features to obtain multiple operation and maintenance tendency clusters; and determining the operation and maintenance tendency of the marine ranch based on the distance between the environmental tendency and the cluster center of each operation and maintenance tendency cluster, specifically including: calculating the distance between the environmental tendency and the cluster center of each operation and maintenance tendency cluster. The distance between the cluster centers of the clusters is used to determine the dominant tendency cluster in the maintenance tendency clusters based on the minimum distance, and the exclusion tendency cluster in the maintenance tendency clusters based on the maximum distance. Based on a preset first major model, the maintenance strategy characteristics of each maintenance cluster in the dominant tendency clusters are summarized to obtain the corresponding dominant maintenance tendency. Based on the first major model, the maintenance strategy characteristics of each maintenance cluster in all maintenance tendency clusters other than the dominant tendency clusters and the exclusion tendency clusters are summarized to obtain the corresponding secondary maintenance tendency. Based on the dominant maintenance tendency and the secondary maintenance tendency, the maintenance tendency of the marine ranch is obtained. Based on the aforementioned multiple monitoring features, multiple corresponding operation and maintenance knowledge nodes are selected from a preset knowledge base, and an operation and maintenance strategy forest is constructed based on these multiple operation and maintenance knowledge nodes. Specifically, this includes: for each monitoring feature, multiple operation and maintenance knowledge nodes corresponding to each monitoring feature are selected from the preset knowledge base; based on the hierarchy of the operation and maintenance knowledge nodes, multiple operation and maintenance knowledge paths corresponding to each monitoring feature are constructed with each monitoring feature as the root node; based on the multiple operation and maintenance knowledge paths corresponding to each monitoring feature, a full combination of the operation and maintenance knowledge paths corresponding to each monitoring feature is constructed to obtain the operation and maintenance strategy forest; wherein, when combining the operation and maintenance knowledge paths, the order of the operation and maintenance knowledge nodes is rearranged according to the execution timing of the operation and maintenance knowledge nodes in the operation and maintenance knowledge paths. Based on the stated operational tendencies, the operational strategy forest is pruned to obtain the target operational strategy. Specifically, this includes: performing a first pruning on the operational strategy forest, during which operational strategy trees with conflicting nodes are removed; and then, based on the stated operational tendencies and a preset second model, performing a second pruning on the operational strategy forest to filter and obtain the target operational strategy; and finally, performing operational management and control of the marine ranch according to the target operational strategy.

2. The method for operation and maintenance management of a marine ranch according to claim 1, characterized in that, The intelligent agent cluster includes weather analysis intelligent agents, equipment analysis intelligent agents, ocean current analysis intelligent agents, and microbial analysis intelligent agents; the pre-set intelligent agent cluster analyzes various monitoring data of the marine ranch to be analyzed, obtaining corresponding monitoring characteristics, specifically including: Based on the visual recognition model, the morphological feature changes of each cloud cluster in the regional weather monitoring cloud map are identified. Then, through the weather analysis agent, the rainfall change time series, wind change time series and morphological feature changes of each cloud cluster in each block of the regional weather monitoring cloud map are correlated and analyzed to obtain the weather time series monitoring features corresponding to the regional weather monitoring cloud map. The device analysis agent extracts features from the device operation data using wavelet transform to obtain device operation features. It then uses isolated forest and single classification algorithms to detect and verify the device operation features, obtaining device anomaly types and device anomaly probabilities. Finally, it combines the device operation features, device anomaly types, and device anomaly probabilities to obtain the device operation monitoring features corresponding to the device operation data. Based on the fluid dynamics model, the regional ocean current monitoring data is visualized to obtain an ocean current monitoring map. The ocean current analysis agent is used to identify the morphological characteristics of each ocean current in the ocean current detection map and to match the data features corresponding to the morphological characteristics of each ocean current in the regional ocean current monitoring data. The morphological characteristics and data features of each ocean current are combined to obtain the ocean current zoning monitoring features corresponding to the regional ocean current monitoring data. Based on the microbial analysis agent, bioinformatics analysis is performed on the microbial monitoring data to obtain regional species distribution and biological indicator characteristics. Species community characteristics are predicted based on the microbial monitoring data. By combining the regional species distribution, the biological indicator characteristics, and the species community characteristics, the corresponding microbial monitoring characteristics are obtained.

3. The method for operation and maintenance management of a marine ranch according to claim 2, characterized in that, The construction process of the intelligent agent cluster is as follows: The agent cluster is subjected to multiple rounds of iterative training. At the start of training, each agent in the agent cluster uses preset construction parameters as initial training parameters. Each agent uses historical monitoring data of the corresponding type of the marine ranch as the training set for training. Until the shared round number is reached, at the end of each training round, each agent uses the weighted average of the training parameters of all agents in the current round as the training parameters of all agents in the next round. After the shared round number is reached, each agent performs iterative training on its own until the training parameters of each agent reach the corresponding threshold and then stop.

4. A marine ranch operation and maintenance management system, characterized in that, The method for implementing the operation and maintenance management of a marine ranch as described in any one of claims 1 to 3 includes a data tendency analysis module, a strategy clustering analysis module, a strategy forest construction module, and a strategy forest pruning module. The data trend analysis module is used to analyze various monitoring data of the marine ranch to be analyzed through a preset intelligent agent cluster, obtain corresponding monitoring features, and search a preset monitoring feature hierarchical classification dictionary based on the monitoring features to determine the environmental trend of the marine ranch; wherein, the various monitoring data include regional weather monitoring cloud maps, equipment operation data, regional ocean current monitoring data, and microbial monitoring data; The strategy clustering analysis module is used to cluster multiple historical operation and maintenance strategies of the marine ranch to characterize several historical operation and maintenance tendencies of the multiple historical operation and maintenance strategies, obtain multiple operation and maintenance tendency clusters, and determine the operation and maintenance tendency of the marine ranch based on the distance between the environmental tendency and the cluster center of each operation and maintenance tendency cluster. The strategy forest construction module is used to select multiple corresponding operation and maintenance knowledge nodes from a preset knowledge base based on the multiple monitoring features, and construct an operation and maintenance strategy forest based on the multiple operation and maintenance knowledge nodes. The strategy forest pruning module is used to prune the operation and maintenance strategy forest in combination with the operation and maintenance tendency to obtain the target operation and maintenance strategy, and to carry out operation and maintenance management of the marine ranch according to the target operation and maintenance strategy.

5. The marine ranch operation and maintenance management system according to claim 4, characterized in that, The strategy clustering analysis module includes a strategy clustering submodule; the strategy clustering submodule includes a feature extraction unit and a feature clustering unit; The feature extraction unit is used to extract keywords and vector features from multiple historical operation and maintenance strategies of the marine ranch based on the TF-IDF algorithm, so as to obtain the corresponding multiple operation and maintenance strategy features. The feature clustering unit is used to perform density-based clustering on the multiple operation and maintenance strategy features to obtain multiple operation and maintenance tendency clusters.

6. The marine ranching operation and maintenance management system according to claim 4, characterized in that, The strategy forest construction module includes a knowledge node filtering submodule and a strategy forest construction submodule; The knowledge node filtering submodule is used to filter out multiple operation and maintenance knowledge nodes corresponding to each monitoring feature from a preset knowledge base, and construct multiple operation and maintenance knowledge paths corresponding to each monitoring feature with each monitoring feature as the root node according to the hierarchy of the operation and maintenance knowledge nodes. The strategy forest construction submodule is used to construct a full combination of operation and maintenance knowledge paths corresponding to each monitoring feature based on multiple operation and maintenance knowledge paths corresponding to each monitoring feature, thereby obtaining an operation and maintenance strategy forest; wherein, when combining operation and maintenance knowledge paths, the order of operation and maintenance knowledge nodes is rearranged according to the execution timing of operation and maintenance knowledge nodes in the operation and maintenance knowledge paths.

Images